polymers

Article
A Modified Aging Kinetics Model for Aging
Condition Prediction of Transformer Polymer
Insulation by Employing the Frequency
Domain Spectroscopy
Jiefeng Liu , Xianhao Fan , Yiyi Zhang * , Hanbo Zheng , Zixiao Wang and Xixi Zhao
College of Electrical Engineering, Guangxi University, Nanning 530004, Guangxi, China;
liujiefeng9999@163.com (J.L.); xianhao_fan@163.com (X.F.); hanbozheng@163.com (H.Z.);
zixiaowang1996@163.com (Z.W.); blairxixi@163.com (X.Z.)
* Correspondence: yiyizhang@gxu.edu.cn




Received: 12 November 2019; Accepted: 10 December 2019; Published: 12 December 2019


Abstract: The aging kinetics model is of great interest to scholars since it is capable of describing
the variation law between the degree of polymerization (DP) and the aging duration of transformer
polymer (cellulose) insulation. However, it is difficult to determine the moisture content inside
the transformer polymer insulation without destroying it, so that the model parameters cannot be
confirmed. Such limitation greatly restricts its application. It is interesting to note that as long as
the moisture content of the transformer polymer insulation could be characterized (replaced) by a
certain feature parameter, the above issue will be solved naturally. The existing researches indicate
that the Frequency Domain Spectroscopy (FDS) is sensitive to moisture. Consequently, the feature
parameter that could characterize the moisture inside transformer polymer insulation (extracted
from the FDS curve) can be used to report a modified aging kinetics model, which could perform the
aging condition prediction of transformer polymer insulation under various test conditions, including
aging duration, aging temperature, and initial moisture. In that respect, the average relative error of
prediction results of prepared samples is equal to 7.41%, which reveals that the reported model might
be serviced as a potential tool for the aging condition prediction of transformer polymer insulation.

Keywords: transformer polymer insulation; Frequency Domain Spectroscopy (FDS); aging kinetics
model; aging condition prediction; degree of polymerization (DP)

1. Introduction
The aging condition of transformer polymer insulation determines the overall service life of
the transformer. Existing researches show that the degree of polymerization (DP) can be serviced
as the most direct and effective indicator for reflecting the aging condition (mechanical strength) of
polymer insulation [1–3]. Reviewing the published researches, we found out that the method for
predicting (or measuring) the DP value of transformer polymer insulation consists of several prevailing
approaches, such as the direct method (viscosity testing) [4] and indirect methods (including the
Frequency Dielectric Spectroscopy (FDS) [5–9] technique, aging kinetics model [10–12], etc.). Although
the viscosity test method is accurate enough, it is not appropriate for analyzing the field equipment
due to its destructiveness. Consequently, the indirect method is of great interest to scholars in contrast
with the direct method.
It is widely accepted that the aging by-products, such as moisture, acids, alcohols, aldehydes,
etc., produced in the degradation process of transformer polymer insulation could regularly affect its
FDS [13]. Therefore, the qualitative or quantitative analysis of the DP value of polymer materials can be

Polymers 2019, 11, 2082; doi:10.3390/polym11122082 www.mdpi.com/journal/polymers

Polymers 2019, 11, 2082 2 of 15

realized by the relevant feature parameter extracted from the FDS data. In literature [14,15], the integral
value of the FDS curve is computed and applied as the feature parameter for achieving the above
goal. In addition, the equivalent circuit model that is capable of analyzing the corresponding aging
condition is also reported, such as the Cole–Cole model [16], the Havriliak–Negami model [17], the
Davidson–Cole model [18], and the Dissado–Hill model [19]. However, it is a fact that the contribution
of the aging condition on measured FDS data could be easily covered by the moisture effect [13–15].
Therefore, it is rather difficult to accurately analyze the aging condition of the transformer polymer
insulation and merely rely on the FDS technique in case the inside moisture is too high.
The published researches pointed out that the aging kinetics model can be employed to describe
the degradation process of transformer polymer insulation under different conditions, including
temperature, aging duration, and especially the initial moisture content mc% [20–25]. Thus, the
discussion of the aging kinetics model can make up for the limitation of the FDS technique. At present,
the research on the aging kinetics model of polymer insulation is mainly based on the Ekamstam
equation [24]. Furthermore, Emsley et al. found that the degradation rate of polymer (cellulose)
materials will not always have a constant attributed to the leveling-off degree of the polymerization
(LODP) phenomenon [21–23]. A second-order kinetics model is thus proposed [23,25]. The findings
reveal that the second-order kinetics model can accurately describe the variation law of the DP value
of polymer materials under different reaction conditions. Therefore, as long as the model parameters
are determined, the above model can be applied to accurately predict the DP value of the polymer
insulation materials. However, it is difficult to determine the moisture content inside the unknown
sample without destroying it for sampling, which makes the model parameters also uncertain. Such
limitation greatly restricts its application on aging condition prediction.
In view of these issues, this research attempts to report a modified aging kinetics model for
aging condition prediction by employing the FDS technique, which aims to overcome the mentioned
limitations. In the present research, depending on the combination of the advantages of both FDS
(sensitive to moisture content) and the aging kinetics model (capable of describing the variation law of
DP), an available modified model is proposed. Findings indicated that the proposed model could be
utilized to predict the aging condition of transformer polymer insulation under various test conditions,
including diverse aging temperature, aging duration, and initial moisture content. Moreover, the
average relative error of prediction results of oil-immersed polymer (cellulose) pressboards is less
than 10%, which further verify the accuracy and feasibility of the reported model. In that respect, the
present contribution reveals that the reported modified aging kinetics model might be serviced as a
potential and effective tool for aging condition prediction.

2. Sample Construction
The transformer polymer (cellulose) insulation material (pressboard) and mineral oil are utilized in
this paper so as to investigate its frequency response characteristic under different insulation conditions.
Therefore, a series of experiments were carried out under controlled laboratory conditions, and the
oil-immersed polymer (cellulose) pressboards with various insulation conditions were later prepared.
The details of both polymer pressboard and oil are described in Table 1.

Table 1. Primary parameters of both cellulose pressboard and oil used in the experiment.

Transformer Polymer Pressboard Transformer Insulating Oil

T4 Weidmann Karamay No.25 naphthenic
Brand Brand
transformer pressboard mineral oil
Thickness 0.5 mm Dielectric loss 4 × 10−4
MD: 98 MPa,
Tensile strength Pour point ≤−45 ◦ C
CMD: 47 MPa
Density 0.96 g/cm3 Flash point 135 ◦ C
Polymers 2019, 11, 2082 3 of 15
Polymers 2019, 11, 2082 3 of 14

The
The pressboards
pressboards with
with various
various initial moisture contents
initial moisture contents were
were completed
completed by by moisture
moisture absorption.
absorption.
Then, the moisture balance distribution process was later performed at 45 ◦ C and lasted for 48 h so as
Then, the moisture balance distribution process was later performed at 45 °C and lasted for 48 h so
to
as perform
to performthethe
FDSFDStest.test.
Subsequently,
Subsequently, DP DP
the the value andand
value moisture content
moisture obtained
content by DP
obtained by tester (by
DP tester
means of viscosity tester, Sierda scientific instrument co., LTD, Shanghai, China) and moister
(by means of viscosity tester, Sierda scientific instrument co., LTD, Shanghai, China) and moister tester (by
means of Karl Fischer Titration, Wantong co., LTD, Switzerland) can be found in Figure
tester (by means of Karl Fischer Titration, Wantong co., LTD, Switzerland) can be found in Figure 1.1.

FigureFigure
1. The 1.
experimental schedule
The experimental of oil-immersed
schedule polymer
of oil-immersed (cellulose)
polymer pressboard.
(cellulose) pressboard.

3. The Establishment of the Traditional Aging Kinetics Model of Transformer Polymer Insulation.
3. The Establishment of the Traditional Aging Kinetics Model of Transformer Polymer
Insulation.
3.1. The Derivation of the Traditional Aging Kinetics Model
3.1. The Derivation
Cellulose of theisTraditional
material a kind of Aging
naturalKinetics
organicModel
polymer homopolymer. Its basic structural unit is
glucose monomer
Cellulose (C6 H10
material isO ), andofits
a 5kind repeating
natural unitpolymer
organic is cellobiose, its general Its
homopolymer. formula is (C6 H10 Ounit
basic structural 5 )n in
is
which n
glucose monomer (C6H10O5), and its repeating unit is cellobiose, its general formula is (C6H10O5)nare
is known as the degree of polymerization (DP) [11,12]. The adjacent glucose monomers in
linked
which byn is1-4-β
known glycoside bond, the
as the degree degradation process
of polymerization (DP)of cellulose
[11–12]. The materials
adjacentisglucose
actuallymonomers
completedare by
the break
linked by of theglycoside
1-4-β glycosidebond,
bond.the In degradation
the course ofprocess
the operation, the glycoside
of cellulose materials isbond of thecompleted
actually cellulose
materials gradually breaks due to the suffered stresses, including thermal, oxidation,
by the break of the glycoside bond. In the course of the operation, the glycoside bond of the cellulose and hydrolysis
stress.
materials Ingradually
this case,breaks
the average
due to number of glucose
the suffered stresses,monomers in each cellulose
including thermal, oxidation,fiber decreased.
and hydrolysis
Consequently,
stress. In this the case, the average DP
corresponding valueof
number decreased
glucose with the increase
monomers of aging
in each duration.
cellulose fiber decreased.
The kinetics model is widely used to describe the change law of
Consequently, the corresponding DP value decreased with the increase of aging duration. DP of polymer materials under
the thermal stressmodel
The kinetics [20–25]. If N is the
is widely used total numberthe
to describe of glucose
change lawmonomers
of DP ofinpolymer
the polymer insulation,
materials under
and M t is the total number of polymer molecules when the aging duration reaches
the thermal stress [20–25]. If N is the total number of glucose monomers in the polymer insulation, time t, then the
corresponding
and Mt is the total average
number DP of polymer
valueofofpolymer insulation
molecules when the canaging
be regarded
duration DPt . time t, then the
as reaches
corresponding average value of DP of polymer insulation
N can be regarded as DPt.
DPt =N (1)
DP = Mt t (1)
Mt
Chain scission number (CSN) is defined as the times of broken chains per polymer chain during
Chain scission number (CSN) is defined as the times of broken chains per polymer chain during
the aging of polymer insulation, the formula for calculating the CSN is as follows:
the aging of polymer insulation, the formula for calculating the CSN is as follows:
DPDP0
CSN == 0 − 1− 1
CSN (2)
(2)
DP DP
t t

where where
DP0 isDP 0 is defined
defined as theasinitial
the initial DP value,
DP value, which which is assumed
is assumed to be
to be 1100 1100paper.
in this in thisThe
paper. The
scission
scission of
fraction fraction of the polymer
the polymer unit (SFCU)unitis(SFCU) is as
defined defined as the
the ratio ratio ofmonomer
of glucose glucose monomer
broken inbroken in a
a polymer
polymer chain to the total number of glucose
chain to the total number of glucose monomers. monomers.
CSN 1 1
SFCU = CSN= 1− DP 1 (3)
SFCU = DP0 DP
= t −0 (3)
DP0 DPt DP0
The existing research on the kinetics model of polymer degradation is mainly based on the aging
Themodel
kinetics existing
ofresearch on the
linear high kineticspolymers.
molecular model of polymer degradation
These researches is mainly
suggest that based on the aging
the degradation of
kinetics
polymermodel
belongsof to
linear high molecular
a random polymers.
chain scission These
reaction researches suggest
(the probability that each
of breaking the degradation
bond between of
polymer
polymer belongs
fibers is to a random
equal) chain
[20,24]. scission
Further, the reaction (the probability
degradation of breaking
rate k is considered each
to be bond between
a constant over a
certain temperature range. Thus, a kinetics equation for describing the variation law between the DP
value and the aging duration of polymer insulation materials is proposed [24].
Polymers 2019, 11, 2082 4 of 15

polymer fibers is equal) [20,24]. Further, the degradation rate k is considered to be a constant over a
certain temperature range. Thus, a kinetics equation for describing the variation law between the DP
value and the aging duration of polymer insulation materials is proposed [24].

1 1
− = k·t (4)
DPt DP0

However, Heywood [26] pointed out that as long as the DP value of polymer (cellulose) insulation
materials drops to a certain value, its trajectory will gradually deviate from the first-order kinetics
model. In this case, the chemical reaction rate is no longer a constant, which is also known as the
leveling-off degree of polymerization [23]. In other words, the parameter k shown in Equation (4) is a
variable that changes with the alteration of aging time, which can be represented by Equation (5).
mt
k(t) = lim = k1 ·e−k2 ·t (5)
∆t→0 N·∆t

From Equation (5), parameters k1 and k2 are equal to a constant, which are both related to the
initial moisture content inside the polymer insulation materials. If there exists a total of m glycosidic
bond breaks in the polymer molecule per unit time, the average degree of polymerization DPt+∆t at
t+∆t can be written as the following form:

N
DPt+∆t = (6)
Mt + m

Moreover, the change amount in the DP value (4DPt+4t ) is shown in Equation (7).

∆DPt+∆t = MtN+m − M N
t
N·m·∆t (7)
= − (M + m)·M ·∆t
t t

The simultaneity of Equations (1) and (7) provide an expression characterizing the rate of decrease
in the DP value of polymer insulation materials.

∆DPt+∆t N2 m
DP0t = lim ∆t = lim − ( M + m )·Mt ·N·∆t
∆t→0 ∆t→0 t
N2 m (8)
≈ lim − Mt 2 ·N·∆t = −DPt · lim N·∆t = dDP
2 m
dt
t
∆t→0 ∆t→0

Substituting Equation (5) into Equation (8), the Equation (8) can be expressed by another form,
shown in Equation (9).
1
dDPt = −k(t)dt = −k1 ·e−k2 ·t dt (9)
DP2t
From Equation (9), the formula for expressing the relationship among the DP, reaction rate, and
reaction duration is obtained by employing the integral operation.
R DPt
1
DP0 DP2
dDP = −k1 ·e−k2 ·t dt
(10)
⇒ DP1
t
1
− DP 0
= kk12 (1 − e−k2 ·t )

The above formula is known as the second-order aging kinetics model, which is utilized for
describing the quantitative relationship between the DP value and the reaction time of polymer
materials. The second-order kinetics model can accurately reflect the change of the polymerization
degree of polymer insulation overall aging processes, and thus has been widely reported [21,23].
However, since the aging rate k(t) is an attenuation function used to describe the slowing down of
the degradation rate of polymer materials when involving the leveling-off degree of polymerization
(LODP) phenomenon, it inevitably lacks a clear physical meaning [25]. Afterwards, depending on the
Polymers 2019, 11, 2082 5 of 15

analysis of polymer degradation at the molecular level, a kinetics model based on the “percentage DP
loss” of polymer materials [25].

DPt
Polymers 2019, 11, 2082 ω(t) = 1 − = ωDP ·(1 − e−kDP ·t ) (11)
5 of 14
DP0
From
From Equation
Equation(11),
(11),ωωDPDPisisassumed
assumed asas
thethe
amplitude
amplitudeof the attenuation
of the attenuationfunction, parameter
function, kDP
parameter
is named as the degradation rate of polymer, ω(t) represents the capacity of the “DP
kDP is named as the degradation rate of polymer, ω(t) represents the capacity of the “DP loss”. Thus, loss”. Thus,
different aging conditions
different aging conditionscan
canbe berepresented
representedbybythethe diverse
diverse values
values of ω(t).
of ω(t). If the
If the polymer
polymer material
material has
has not been degraded so far, the corresponding ω(t) is equal to 0, and the value is equal
not been degraded so far, the corresponding ω(t) is equal to 0, and the value is equal to 1 in the case to 1 in the
case that polymer materials are completely
that polymer materials are completely degraded. degraded.

3.2. The
The Aging
Aging Kinetics
Kinetics Model
Model for
for Aging
Aging Condition
Condition Analysis
Analysis
In
In this
this part,
part, several
several oil-immersed
oil-immersed polymer
polymer pressboards
pressboards (shown
(shown in in Table
Table 1)
1) with
with different initial
moisture
moisture content
content (including
(including 0.78%,
0.78%, 1.32%,
1.32%, 2.07%,
2.07%, 2.74%,
2.74%, andand 3.82%)
3.82%) are firstly prepared. Then,Then, the
above
above pressboards are utilized to perform the alternate thermal aging experiment at 150 °C. ◦ C. The DP

values of
of the
thesame
samepressboard
pressboardwithwith
thethe different
different aging
aging duration
duration are recorded
are recorded in theincourse
the course
of theof the
aging
aging experiment.
experiment. If the horizontal
If the horizontal axis
axis and and vertical
vertical axis areaxis are represented
represented by agingbyduration
aging duration
(hours) (hours)
and DP
and DPrespectively,
values, values, respectively, thepicture
the resulting resulting
thatpicture
adoptedthat adopted to
to describing thedescribing the between
variation law variationaging
law
between
duration aging
and DP duration
values and DPeasily
can be values can be easily plotted.
plotted.
Moreover, it is clear that Equation (11) can be utilized to describe the degradation degradation process
process of
polymer
polymer insulation. Provided that Equation (11) is adopted to fitting the presented variation law
insulation. Provided that Equation (11) is adopted to fitting the presented variation law of
DP vs. aging duration, the parameters (ωDP DP and
and kkDP
DP) )in
inthe
theformula
formulacan can be
be naturally
naturally computed,
computed, and
results are shown in Figure 2.

1200
Measured value (mc%=0.78) Fitting curve 1 R2=0.992
Measured value (mc%=1.32) DP1 (t )=1100 ⋅ [1 − 0.73 ⋅ (1 − e−0.0034⋅t )]
1000 Measured value (mc%=2.07)
Measured value (mc%=2.74) Fitting curve 2 R2=0.999
Measured value (mc%=3.82) DP2 (t )=1100 ⋅ [1 − 0.75 ⋅ (1 − e−0.0074⋅t )]
800
Fitting curves Fitting curve 3 R2=0.997
DPt

DP3 (t )=1100 ⋅ [1 − 0.78 ⋅ (1 − e−0.012⋅t )]

600
Fitting curve 4 R2=0.996
DP4 (t )=1100 ⋅ [1 − 0.79 ⋅ (1 − e−0.016⋅t )]
400
DP1 Fitting curve 5 R2=0.998
DP5 DP4 DP3 DP2 DP5 (t )=1100 ⋅ [1 − 0.81⋅ (1 − e−0.019⋅t )]
200
90 190 290 390 490
Aging duration (hours) Fitting equations
Figure
Figure 2.
2. The
The variation
variation law
law of
of the
the degree
degree of
of polymerization
polymerization (DP)
(DP) value
value vs.
vs. aging
aging duration.
duration.

It is obvious that
that the
the higher
higher the
the initial
initial moisture
moisture content
content inside
inside the
the polymer
polymer insulation,
insulation, the
the faster
faster
the DP value decreases, and the less time it takes for the curve to become flat. Meanwhile,
Meanwhile, the
the impact
impact
generated byby the
the initial
initial moisture
moisture content
content can
can be
be also
also observed
observed ininthe
thevalues
valuesofofcomputed
computedparameters.
parameters.
These values and error bursts (EB)
(EB) are
are shown
shown in in Table
Table 2, and the
the significant
significant figures
figures in the
the experiment
experiment
are reserved for 2 digits.

Table 2. The details of the used parameters in the aging kinetics model.

0.78% 1.32% 2.07% 2.74% 3.82%

mc%
value EB (±) value EB (±) value EB (±) value EB (±) value EB (±)
2.36 1.38 4.45 6.13 6.99
ωDP 0.73 0.75 0.78 0.79 0.81
× 10−4 × 10−4 × 10−4 × 10−4 × 10−4
6.34 2.76 1.25 2.36 1.41
kDP 0.0034 0.0074 0.012 0.016 0.019
× 10−4 × 10−4 × 10−3 × 10−4 × 10−3
Polymers 2019, 11, 2082 6 of 15

Table 2. The details of the used parameters in the aging kinetics model.
Polymers 2019, 11, 2082
0.78% 1.32% 2.07% 2.74% 3.82% 6 of 14
mc%
value EB (±) value EB (±) value EB(±) value EB (±) value EB (±)
The values of parameters shown in Table 2 are subjected to present an increasing trend with the
rising 2.36content. Such the 1.38phenomenon has 4.45 also pointed out 6.99
6.13that the quantitative
ωDP initial0.73
moisture 0.75 0.78 0.79 0.81
× 10−4 × 10−4 × 10−4 × 10−4 × 10−4
relationship of mc% vs. ωDP and mc% vs. kDP can be studied by fitting analysis. The fitting curves and
6.34 2.76 1.25 2.36 1.41
kDP equation
fitting 0.0034can be found 0.0074
in Figure 3 and 0.0123, respectively,
Table 0.016
where ×the 0.019R2 represent
RSS and
× 10−4 × 10 −4 × 10−3 10−4 × 10−3
the residual sum of squares and determination coefficient. The closer the value of RSS is to 0 and the
closer the value of R2 is to 1, the better the fitting results.
The values of parameters shown in Table 2 are subjected to present an increasing trend with
the rising initial moisture content.
Table Such equation
3. The fitting the phenomenon
of mc% vs.has ωDP also pointed
and mc% out
vs. kDP. that the quantitative
relationship of mc% vs. ωDP and mc% vs. kDP can be studied by fitting analysis. The fitting curves and
fitting equation can be found in Figure 3 and Fitting Equation
Table 3, respectively, where the RSS and R2 represent
ω DP ( mc % )= −coefficient.
the residual sum of squares and determination 0.16 ⋅ e − mc %/ 2.6The
+ 0.85
closer the value of RSS is to 0 and the
− mc %/3.3
2 k DP ( mc
closer the value of R is to 1, the better the% )= − 0.033
fitting ⋅e
results. + 0.030

Measured value Measusred value

Fitting curve 0.020 Fitting curve
0.80 0.81
0.019
0.016 0.016
0.79
0.78
ω DP

kDP

0.012 0.012
Model ExpDec1 Model ExpDec1
0.75 y = A1*exp(-x/t1) + y0 Equation y = A1*exp(-x/t1) + y0
Equation
Plot W Plot K
0.75 0.84832 ± 0.02964
0.007 y0 0.02959 ± 0.00422
y0
A1 -0.1622 ± 0.01933 0.0074 A1 -0.03313 ± 0.0033
t1 2.55367 ± 1.07421 t1 3.31727 ± 0.81261
0.73 Reduced Chi-Sqr 2.15824E-5 Reduced Chi-Sqr 1.62965E-7
0.004 0.99789
R-Square (COD) 0.99013 R-Square (COD)
Adj. R-Square 0.98027 0.0034 Adj. R-Square 0.99577
0.70
1 2 3 4 1 2 3 4
Moisture content (%) Moisture content (%)
(a) (b)

Figure3.3.The
Figure Thefitting
fittingcurves
curvesof
ofthe
theparameters
parametersin
inthe
thekinetics
kineticsmodel. (a)ωωDP
model.(a) DP..
DP, (b) kDP

3. The fitting equation of mc% vs. ωDP and mc% vs. kDP .
It is obvious that Table
equations shown in Table 3 could describe the functional relationship between
relevant parameters (ωDP and kDP) and initial moisture
Fitting content. It means that the higher the moisture
Equation
content, the larger the value of the ωDP and kDP, and thus the faster the degradation rate of polymer
ωDP (mc%) = −0.16·e−mc%/2.6 + 0.85
insulation materials. By employing kDP (mc%) = −0.033·e−mc%/3.3 + 0.030shown in Equation (11) and Table
a combination of the formulas
3, a universal model for estimating the remaining life L(t) [27,28] of transformer polymer insulation
under any initial moisture content is readily obtained, which can be expressed in the form of Equation
(12). It is obvious that equations shown in Table 3 could describe the functional relationship between
relevant parameters (ωDP and kDP ) and initial moisture content. It means that the higher the moisture
content, the larger the value of the ωDP and kDP , and 1 − DP t / the
thus DPfaster the degradation rate of polymer
Ln(1 − )
insulation materials. By employing a combination of ωDP (mc%) shown in Equation (11) and Table(12)
the formulas 3,
L ( t ) = −
a universal model for estimating the remaining life L(t) [27,28] of transformer polymer insulation under
k DP (mc%)
any initial moisture content is readily obtained, which can be expressed in the form of Equation (12).
However, it is rather difficult to obtain accurate DP t values during the operation. In other words,
1−DPt /DP
it is impractical to predict the remaining life ofLn a (real
1 − transformer
ωDP (mc%)
) by using Equation (12). Therefore,
L(t) = − (12)
an improved equation to predict the aging degree of polymer
kDP (mc%) insulation of the transformer is required.
An improved equation that can predict the DP value of transformer polymer insulation is proposed,
which However,
is shownitinis Equation (13). to obtain accurate DPt values during the operation. In other words,
rather difficult
it is impractical to predict the remaining life of a real transformer by using Equation (12). Therefore, an
− k DP ( mc% )⋅t
DPt (tthe
)=DP 0 ⋅ [1 − ωDP ( mc%) ⋅ (1 − e )] the transformer is required.
improved equation to predict aging degree of polymer insulation of (13)

4. The Improved Aging Kinetics Model by Employing Frequency Dielectric Spectrum

On the condition that Equation (13) is desired to perform the analysis activities of the aging
condition, a crucial basis is that the moisture content inside the polymer insulation materials is
Polymers 2019, 11, 2082 7 of 15

An improved equation that can predict the DP value of transformer polymer insulation is proposed,
which is shown in Equation (13).

DPt (t) = DP0 ·[1 − ωDP (mc%)·(1 − e−kDP (mc%)·t )] (13)

4. The Improved Aging Kinetics Model by Employing Frequency Dielectric Spectrum

On2019,
Polymers the11,
condition
2082 that Equation (13) is desired to perform the analysis activities of the 7aging of 14
condition, a crucial basis is that the moisture content inside the polymer insulation materials is
accurately determined. However, However, the the determination
determination of of the moisture content inside an unknown unknown
cellulose sample is also difficult.
is also difficult.
In view of of the
the above
above issues,
issues, the
the existing
existing researches
researches reveal
reveal that
that dielectric
dielectric frequency
frequency response
response
might be used as an an available
available approach
approach to to achieve
achieve the the moisture
moisture analysis
analysis of of polymer
polymer insulation
insulation
materials
materials [2,6,7,10,11].
[2,6,7,10,11]. Provided
Providedthat thatthe
therelevant
relevantparameter
parameterextracted
extracted from
fromthetheanalysis
analysisof FDS
of FDScancan
be
adopted
be adopted to reflect
to reflectthethe
moisture
moisture content
content inside
insidethethecellulose
celluloseinsulation,
insulation,thethequantitative
quantitative relationship
relationship
between the extracted
extracted frequency
frequencydielectric
dielectricresponse
responseparameter
parameterand andeither
either ω
ωDPDP (mc%) or kDP k(mc%)
(mc%) or DP (mc%)can
can be established.
be established. Consequently,
Consequently, Equation
Equation (13)(13) could
could bebe readilyemployed
readily employedtotocarry carry out
out the aging
aging
condition prediction
prediction without
without thethe determination
determination of of the
the moisture
moisture content.
content.
The purpose of this chapter is to find out the relevant frequency response feature parameter,
which
which could
could characterize
characterize the moisture
moisture content
content but but slightly
slightly affected
affected byby its
its aging
aging condition.
condition. In the
previous
previous researches
researches[14,15],
[14,15],the
theauthors
authorsdemonstrated
demonstratedthat thatwith
withthe aggravating
the aggravating of of
thethe
aging degree
aging degreeof
transformer polymer insulation, the FDS curve changes remarkably in
of transformer polymer insulation, the FDS curve changes remarkably in low-frequency regions, low-frequency regions, while
remaining unchanged
while remaining in high-frequency
unchanged regions. Therefore,
in high-frequency the frequency
regions. Therefore, the response
frequency information
response
extracted
information from the high-frequency
extracted regions can be used
from the high-frequency regionsto characterize
can be usedthe tomoisture content
characterize theinside the
moisture
transformer
content inside polymer insulationpolymer
the transformer without insulation
being affected by its
without aging
being condition.
affected by its aging condition.
In order to to obtain
obtain thethe desired
desired relevant
relevant parameters,
parameters, the frequency dielectric response test is
performed. ItIt isislaunched
launchedininthe thethree-electrode
three-electrodetest testcell
cellunder
under4545°C.◦ C. The test platform is shown
The test platform is shown in
in Figure
Figure 4. 4.
TheTheFDSFDS curvesofofnon-aged
curves non-agedcellulose
cellulosepressboards
pressboardswith withvarious
various mc% mc% (including
(including 0.78%,
1.29%, 3.15%,3.43%,
1.29%, 2.32%, 3.15%, 3.43%,and and4.03%)
4.03%)are aremeasured
measured andand plotted
plotted in Figure
in Figure 5. Whereas
5. Whereas dielectric
dielectric loss
loss (tanδ)
(tanδ) is theis the
ratioratio of imaginary
of imaginary partpart
to to
thethe real
real ε*(ω),which
partofofε*(ω),
part whichisisnot
not affected
affected by insulation
insulation
geometry.
geometry. Thus,
Thus, the
the parameter
parameter extracted
extracted from
from thethe tanδ
tanδ curve is more appropriate
appropriate for various cellulose
insulation materials.

Figure 4.
Figure 4. The
The frequency dielectric response
frequency dielectric response test
test platform.
platform.

mc%=0.78% mc%=0.78% mc%=0.78%

mc%=1.29% mc%=1.29% mc%=1.29%
2
mc%=2.32% 10 mc%=2.32% 10 1
mc%=2.32%
101 mc%=3.15% mc%=3.15% mc%=3.15%
mc%=3.43% mc%=3.43% mc%=3.43%
mc%=4.03% 101 mc%=4.03% mc%=4.03%
100
ε''(ω)

tanδ
ε'(ω)

100

10-1

10-1

10-2
100 10-2
10-3 10-2 10-1 100 101 102 103 10-3 10-2 10-1 100 101 102 103 10-3 10-2 10-1 100 101 102 103
Polymers 2019, 11, 2082 Figure 4. The frequency dielectric response test platform. 8 of 15

mc%=0.78% mc%=0.78% mc%=0.78%

mc%=1.29% mc%=1.29% mc%=1.29%
mc%=2.32% 102 mc%=2.32% 101 mc%=2.32%
101 mc%=3.15% mc%=3.15% mc%=3.15%
mc%=3.43% mc%=3.43% mc%=3.43%
mc%=4.03% 101 mc%=4.03% mc%=4.03%
100

ε''(ω)

tanδ
ε'(ω)

100

10-1

10-1

10-2
100 10-2
10-3 10-2 10-1 100 101 102 103 10-3 10-2 10-1 100 101 102 103 10-3 10-2 10-1 100 101 102 103
Polymers 2019, 11, 2082f (Hz) f (Hz) f (Hz) 8 of 14
(a) (b) (c)
AsFigure
mentioned above,
The ε*(ω) the tanδ curve pressboards
in the high-frequency regionscontents.
could merely reflect the
Figure 5. 5.The ε*(ω) of non-aged
of non-agedpolymer
polymer pressboardswith various
with moisture
various (a) ε‫(׳‬ω),
moisture contents. ε’(ω),
(a) (b)
information related to moisture content, and thus the selected high-frequency regions are defined as
(b)ε‫(׳׳‬ω),
ε”(ω),(c)(c)
tanδ.
tanδ.
feature frequency regions, and a range of 100–1000 Hz is selected for this paper. In order to reduce
errorsAsinmentioned
FDS data due to experimental
above, operations,
the tanδ curve in the the integral operation
high-frequency (in 100–1000
regions Hz) isreflect
could merely therefore
the
utilized to extract
information relatedthe
to expected frequency
moisture content, response
and thus theparameter. Equation (14) regions
selected high-frequency shows the
arecalculation
defined as
formulafrequency
feature of these integral
regions,values
and a(IV). The
range ofcomputed
100–1000 HzIV values are shown
is selected for thisinpaper.
Table 4.
In order to reduce
b f
errors in FDS data due to experimental operations, the1000integral operation (in 100–1000 Hz) is therefore
IV =  tan δ ( f ) ⋅ f =  tan δ ( f ) df (14)
utilized to extract the expected frequency
fa response parameter. Equation (14) shows the calculation
100

formula of these integral values (IV). The computed IV values are shown in Table 4.
Table 4. The parameters of integral value (IV) and mc%.
fb
X Z 1000
Parameters IV = tan δ( f )· Measured
f = tan δ( f )d f
Value (14)
fa 100
mc% 0.78% 1.29% 2.32% 3.15% 3.43% 4.03%
IV 12.9 17.9 24.4 32.6 44.2 85.1
Table 4. The parameters of integral value (IV) and mc%.
An obvious exponential relationship of IV vs. mc% can be observed, and this viewpoint is proved
Parameters Measured Value
by the fitting analysis. Its results are shown in Figure 6 and Table 5, respectively.
mc% 0.78% 1.29% 2.32% 3.15% 3.43% 4.03%
IV 12.9 17.9 24.4 32.6 44.2 85.1
Table 5. The fitting equation of IV vs. mc%.

An obvious exponential relationship Fitting

of IV Equation
vs. mc% can of IV
be observed, and this viewpoint is proved
by the fitting analysis. Its IV (m c%
results are)=shown
21.5+in0 .04
Figurem c % / 0.56
5 e 6 and −Table e −respectively.
3 2.35, m c % / 0.6 0

Model ExpDec2
90 y = A1*exp(-x/t1) + A2*exp(-x/t2) + y0
Equation
B
85.1
Plot
80 y0 21.4509 ± 7.24359
A1 0.04457 ± 0.07888
-0.55463 ± 0.1278
70 t1
Integral value of tanδ

A2 -32.27303 ± 50.84442
t2 0.60188 ± 0.99627
60 Reduced Chi-Sqr 5.31376
R-Square (COD) 0.99848
Adj. R-Square 0.99238
50
Measured value
40 Fitting curve 44.2

30
32.6
20 24.4
17.9
10
12.9
1 2 3 4
Moisture content (%)

Figure 6.
Figure 6. The fitting equation
equation of
of IV
IV vs. mc%.

Form Table 5, the functional relationship between mc% and IV can be expressed by the means of
IV = f(mc%), thus, the inverse function formula of IV could be expressed as mc% = f −1(IV). Furthermore,
from Table 3, provided that ωDP = g(mc%), and kDP = p(mc%), it means that the moisture content
utilized in the formula of ωDP and kDP can be replaced by the inverse function of IV. Furthermore, if
the mc% in the formula is offset and the quantitative relationship between IV and ωDP/kDP is
established directly, the processed formula is written as ωDP = g[f −1(IV)], and kDP = p[f −1(IV)]. Then, the
model for aging condition prediction based on IV can be derived, which is shown in Equation (15).
Polymers 2019, 11, 2082 9 of 15

Table 5. The fitting equation of IV vs. mc%.

Fitting Equation of IV
IV (mc%) = 21.5 + 0.045emc%/0.56 − 32.3e−mc%/0.60

Form Table 5, the functional relationship between mc% and IV can be expressed by the means of
IV = f (mc%), thus, the inverse function formula of IV could be expressed as mc% = f −1 (IV). Furthermore,
from Table 3, provided that ωDP = g(mc%), and kDP = p(mc%), it means that the moisture content
utilized in the formula of ωDP and kDP can be replaced by the inverse function of IV. Furthermore, if the
mc% in the formula is offset and the quantitative relationship between IV and ωDP /kDP is established
directly, the processed formula is written as ωDP = g[f −1 (IV)], and kDP = p[f −1 (IV)]. Then, the model
for aging condition prediction based on IV can be derived, which is shown in Equation (15).
−1 (IV )]·t
DPt (t) = DP0 ·[1 − ωDP ·[ f −1 (IV )]·(1 − e−kDP ·[ f )] (15)

The functional relationship between IV and ωDP /kDP is derived by simultaneous analysis of the
equations shown in Tables 3 and 5. Then, Figure 7 and Table 6 present the corresponding fitting curves
and fitting equations, respectively. From Table 6, ωDP and kDP can be calculated by the integral value
Polymers 2019, 11, 2082 9 of 14
(IV) of the tanδ curve. The modified model introduced in Equation (15) will promote the practice of
aging condition
practice of aging prediction
conditionofprediction
the unknown polymer
of the samples
unknown on the samples
polymer conditionon
that
theitscondition
moisture content
that its
is not determined.
moisture content is not determined.

0.020
0.80

Calculated value Calculated value

0.015
Fitting curve Fitting curve
ω DP

kDP

Model ExpDec1
0.76 y = A1*exp(-x/t1) + y0 Model ExpDec1
Equation
0.010 Equation y = A1*exp(-x/t1) + y0
Plot w
Plot k
y0 0.81472 ± 0.00192
y0 0.01971 ± 3.80441E-4
A1 -0.26745 ± 0.01899
A1 -0.04708 ± 0.00313
t1 11.80572 ± 0.73199
0.005 t1 12.73151 ± 0.80442
Reduced Chi-Sqr 1.79349E-5
0.72 Reduced Chi-Sqr 6.29725E-7
R-Square (COD) 0.97639 0.977
R-Square (COD)
Adj. R-Square 0.97487 0.97551
Adj. R-Square

20 40 60 80 20 40 60 80

Integral value of tanδ curve Integral value of tanδ curve

(a) (b)
Figure 7. The fitting curves. (a) ωDP vs. IV, (b) kDP vs. IV.
Figure 7. The fitting curves. (a) ωDP vs. IV, (b) kDP vs. IV.

Table 6. The fitting equation of ωDP vs. IV and kDP vs. IV.
Table 6. The fitting equation of ωDP vs. IV and kDP vs. IV.
Fitting Equation
Fitting Equation
ωDP (IV ) = −0.27·e−IV/11.8 + 0.81
ωkD P ((IV
IV)) =
= −−0.047·e − IV /1 1 .8
0 .2 7 ⋅ e−IV/12.7 ++0.020
0 .8 1
DP
k D P ( IV ) = − 0 .0 4 7 ⋅ e − IV /1 2 .7 + 0 .0 2 0
5. A Modified Aging Kinetics Model for Aging Condition Prediction Including the
5. A ModifiedEffect
Temperature Aging Kinetics Model for Aging Condition Prediction Including the Temperature
Effect
In the previous chapter, the integral value IV was applied to calculate the value of ω and k DP DP
corresponding to the chapter,
In the previous same moisture content.
the integral Thus,
value the modified
IV was applied tomodel considering
calculate ωDP and kIV
theofparameter
the value DP
can be applied to
corresponding to the
predict
samethe aging condition
moisture of samples
content. Thus, with various
the modified model moisture contents.
considering However,
the parameter IV
the values
can of ωDP
be applied and kDPthe
to predict areaging
obtained by analyzing
condition the with
of samples variation of DP vs.contents.
lawmoisture
various aging duration
However,at
the values of ωDP and kDP are obtained by analyzing the variation law of DP vs. aging duration at 150
°C. Theoretically speaking, the proposed model is only suitable for transformer polymer insulation
operating at 150 °C. This defect will greatly limit its versatility.
Fortunately, the time-temperature superposition theory [29,30] indicates that the effect
generated by increasing temperature and prolonging the reaction duration on the molecular motion
Polymers 2019, 11, 2082 10 of 15

150 ◦ C. Theoretically speaking, the proposed model is only suitable for transformer polymer insulation
operating at 150 ◦ C. This defect will greatly limit its versatility.
Fortunately, the time-temperature superposition theory [29,30] indicates that the effect generated
by increasing temperature and prolonging the reaction duration on the molecular motion is equivalent.
In other words, a similar degradation degree of the transformer polymer insulation under high
temperature can be also observed in other conditions (both a lower temperature and longer duration).
Such property can be expressed as follows, tref ·Tref = t·T, Tref (T) is regarded as the reference temperature
(different operating or thermal aging temperature), tref (t) is the time (hours) required for the chemical
reaction to reach a certain aging degree at the temperature of Tref (T). Review the Arrhenius theory, which
is utilized to interpret the variation law between chemical reaction rate and reaction temperature [17,18]
as is shown in Equation (16).
Ea
− R·T
ki = A·e i (16)

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, R = 8.314 J/mol·K, ki represents the chemical reaction rate of
the same reaction under i-th types of temperature, and Ea is activation energy and assumed to be
103 kJ/mol [27]. In this paper, 423.15 K (150 ◦ C) is regarded as the Tref . Therefore, the ratio of the
reaction rates at the Tref and T can be expressed by Equation (17), which is defined as shift factor αT .
Ea
T k ( 1 −1)
αT = = = e R Tre f T (17)
Tre f kre f

Thus, the formula for calculating the time t required to reach a given degree of reaction at
temperature T can be represented by Equation (18) [25].
Ea 1
R ( Tre f − T1 )
tre f = t·αT = t·e (18)

Obviously, relying on the simultaneous analysis of Equation (15) and Equation (18), the resulting
modified model will be suitable for predicting the aging condition of transformer polymer insulation
operating (or prepared) at different temperatures, which is presented in Equation (19).

DPt (t, IV, T ) = DP0 ·[1 − ωDP (IV )·(1 − e−kDP (IV )·tre f )]




ωDP (IV ) = −0.27·e−IV/11.8 + 0.81





kDP (IV ) = −0.047·e−IV/12.7 + 0.020





 1000 R 1000
(19)

tan δ( f )· f = 100 tan δ( f )d f
P


 IV =
100







 tre f = t·αT (T)
Ea 1 − T1 )

R ( Tre f

αT (T) = e



It is obvious that the aging condition of the transformer polymer insulation in this model relies on
three crucial factors (parameters), including aging duration t in the hour (at temperature T), initial
moisture content (represented by integral value IV), and operating (thermal aging) temperature T.
Moreover, the crucial basis of Equation (19) is that its FDS data (tanδ) is already known, and then,
the prediction results of aging conditions at the reference temperature (150 ◦ C) can be extended to
different operating (thermal aging) temperatures by considering the shift factor shown in Equation (18).
Thus, Equation (19) can be applied to predict the DP value of the transformer polymer insulation with
different initial moisture content after t hours of operation (or thermal aging) at temperature T. The
aging condition (DP value) of transformer polymer insulation at a different initial moisture (integral
value) content and operating (thermal aging) temperature can be predicted by using Equation (19),
respectively, and the resulting picture is presented in Figure 8. Figure 8a plots the DP vs. t and T, in
the case of mc% = 2.0%. Figure 8b plots the DP vs. t and mc%, and T equal to 423.15 K.
(18). Thus, Equation (19) can be applied to predict the DP value of the transformer polymer insulation
with different initial moisture content after t hours of operation (or thermal aging) at temperature T.
The aging condition (DP value) of transformer polymer insulation at a different initial moisture
(integral value) content and operating (thermal aging) temperature can be predicted by using
Polymers 2019, 11, 2082 11 of 15
Equation (19), respectively, and the resulting picture is presented in Figure 8. Figure 8a plots the DP
vs. t and T, in the case of mc% = 2.0%. Figure 8b plots the DP vs. t and mc%, and T equal to 423.15 K.

(a) (b)

Figure
PolymersFigure 8.2082
8.
2019, 11, The
The DP
DP prediction under different
prediction under different conditions.
conditions. (a)
(a) DP
DP vs.
vs. tt and
and T,
T, (b)
(b) DP
DP vs.
vs. tt and
and mc%.
mc%. 11 of 14

6.
6. Feasibility
Feasibility Investigation
Investigation of
of the
the Proposed
Proposed Modified
ModifiedAging
AgingKinetics
KineticsModel
Model
AA novel
novel modified
modified aging
aging kinetics
kinetics model
model for
for predicting
predicting thethe aging
aging condition (i.e., DP
condition (i.e., DP value)
value) of
of
transformer polymer insulation materials has been put forward. It is observed that
transformer polymer insulation materials has been put forward. It is observed that once the FDS once the data,
FDS
data,
agingaging duration,
duration, and
and the the temperature
aging aging temperature is determined,
is determined, EquationEquation
(19) could (19)
be could
employed be employed
to predict
to predict the DP value immediately. In order to demonstrate the feasibility of
the DP value immediately. In order to demonstrate the feasibility of the proposed model, the proposed model,
the
the investigation experiments of DP prediction of transformer polymer insulation
investigation experiments of DP prediction of transformer polymer insulation prepared in lab prepared in lab
condition
condition is
is performed.
performed. The
The prediction
prediction schedule
schedule isis plotted
plotted in
in Figure
Figure9. 9.

Figure 9.
Figure 9. The
The prediction
prediction schedule
schedule of
of the
the DP
DP of
of the
the transformer
transformer polymer
polymer insulation.
insulation.

In
In this
this chapter,
chapter, several
several transformer
transformer polymer
polymer insulation
insulation pressboards
pressboards with
with different
different insulation
insulation
conditions
conditions are
are selected
selected toto verify
verify the
the feasibility
feasibility of
ofthe
theproposed
proposedmodel.
model. The
The preparation
preparation schedule
schedule of
of
these
these pressboards
pressboards is is referred
referred toto in
in Figure
Figure 1,
1, and
and its
its details
details are
are described
described in
in Table
Table7.7. Pressboards
Pressboards 1–4
1–4
show
show low initial mc%,
low initial mc%,while
whilepressboards
pressboards5–8 5–8contain
containhigh initialmc%.
highinitial mc%.

Table7.7. The
Table The details
details of
of the
the prepared
preparedpressboards.
pressboards.

Pressboard
Pressboard DryPressboard
Dry Pressboard Wet
Wet Pressboard
Pressboard
Parameters
Parameters 11 22 33 4 4 5 5 6 6 7 7 8 8
Measured mc%
Measured mc% 0.98%
0.98% 1.02%
1.02% 1.12%
1.12% 1.15%
1.15% 2.31%
2.31% 2.93%
2.93% 3.87%
3.87% 4.12%
4.12%
Measured DP
Measured DPmm 1093
1093 897
897 751
751 548548 832832 644644 472
472 394394
Agingtemperature
Aging temperature (T)/K
(T)/K 423.15
423.15 423.15
423.15 423.15
423.15 423.15
423.15 403.15
403.15 403.15
403.15 403.15
403.15 403.15
403.15
Aging duration
Aging duration (t)/h(t)/h 2424 7272 9696 168168 96 96 168168 264264 336336
Integral value (IV) 14.7 15.0 16.1 16.2 22.8 42.1 60.1 82.0
Integral value (IV) 14.7 15.0 16.1 16.2 22.8 42.1 60.1 82.0

Moreover,
Moreover,the
theFDS
FDStest is performed
test is performed at◦45
at 45 C, the
°C, resulting tanδ curves
the resulting corresponding
tanδ curves to the above
corresponding to the
eight
abovesamples are presented
eight samples in Figurein
are presented 10.Figure
Afterwards, the integral
10. Afterwards, values
the in avalues
integral range of
in100–1000 Hz100–
a range of are
calculated based on these tanδ curves, and the calculated values are shown in Table 3 as well.
1000 Hz are calculated based on these tanδ curves, and the calculated values are shown in Table 3 as
well.

Pressboard 1 Pressboard 5
Pressboard 2 101 Pressboard 6
100 Pressboard 3 Pressboard 7
Pressboard 4 Pressboard 8
 Integral value (IV) 14.7 15.0 16.1 16.2 22.8 42.1 60.1 82.0

Moreover, the FDS test is performed at 45 °C, the resulting tanδ curves corresponding to the
above eight samples are presented in Figure 10. Afterwards, the integral values in a range of 100–
1000 Hz2019,
Polymers are11,
calculated
2082 based on these tanδ curves, and the calculated values are shown in Table
12 3
ofas
15
well.

Pressboard 1 Pressboard 5
Pressboard 2 101 Pressboard 6
100 Pressboard 3 Pressboard 7
Pressboard 4 Pressboard 8

100 Integral value of

tanδ

100-1000 Hz

tanδ
10-1

Integral value of
100-1000 Hz 10-1

10-2

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)

Figure 10. The Frequency

Frequency Domain
Domain Spectroscopy
Spectroscopy (FDS)
(FDS) of polymer
polymer pressboards.
pressboards. (a)
(a) Dry
Dry pressboard,
pressboard,
(b) wet pressboard.

Substituting the parameter (T, t, and IV, shown in Table 7) into Equation (19), the relevant
model parameters (including the ωDP , kDP , and αT ) and predicted DP values of these pressboards
with different initial moisture content, aging duration, and thermal aging temperature are obtained
immediately and shown in Table 8.

Table 8. Predicted results of the DP value of the prepared polymer pressboards.

Parameters Dry Pressboard Wet Pressboard

Pressboard 1 2 3 4 5 6 7 8
Parameters ωDP 0.738 0.740 0.747 0.747 0.776 0.807 0.813 0.815
Parameters kDP 0.0049 0.0052 0.0064 0.0065 0.012 0.018 0.019 0.020
Parameters αT 1.0 1.0 1.0 1.0 0.20 0.20 0.20 0.20
Measured DPm 1093 897 751 548 832 644 472 394
Predicted DPp 1010 845 720 554 924 693 523 438
Relative error 7.59% 5.80% 4.13% 1.09% 11.06% 7.61% 10.81% 11.17%

The formula of relative error is shown in Equation (20). It is found from Table 8 that the relative
error between the predicted results and the measured results keep in 1%–12%, and the average value
of these relative errors is equal to 7.41%. Furthermore, the estimation error in Table 8 is plotted in
Figure 11, which is represented by the distance between the measured value and the predicted value.
The comparison of the actual aging state and the predicted state of the prepared samples are also
shown in Figure 11 according to the aging state division involving the operating experience.

|Predicted value − Measured value|

Relative error = × 100% (20)
Measured value
 The formula of relative error is shown in Equation (20). It is found from Table 8 that the relative
error between the predicted results and the measured results keep in 1%–12%, and the average value
of these relative errors is equal to 7.41%. Furthermore, the estimation error in Table 8 is plotted in
Figure 11, which is represented by the distance between the measured value and the predicted value.
Polymers 2019, 11, 2082of the actual aging state and the predicted state of the prepared samples are
The comparison 13 also
of 15

shown in Figure 11 according to the aging state division involving the operating experience.
1200
Measured DP
1093 Predicted DP
1000 1010
897 New Insulation
924 Early Aging
Mid-term Aging
845 832
DP value 800 End of Aging
751
720
693
644
600 554
548
523
472
438
400 394

1 2 3 4 5 6 7 8
Pressboard label

Figure 11. Comparison of the measured value and predicted value.

It is worth noting that the reported fitting equations, models, and selected parameters significantly
depend on the individual materialsPredicted value−conditions
and material Measured used
valueduring the experiments (both
Relative error = ×100% (20)
for the oil and the pressboard). Fortunately, Measured value for describing the microstructure or
the parameters
physicochemical
It is worth property
noting thatof transformer
the reportedpolymer
fitting (cellulose)
equations, insulation
models, andmaterials follow
selected similar
parameters
variation rulesdepend
significantly (or laws)onin the
the course of the degradation
individual materials and process. Therefore,
material if the proposed
conditions used duringmodelthe
is
supported to perform the aging condition prediction of polymer insulation under other
experiments (both for the oil and the pressboard). Fortunately, the parameters for describing the test conditions
(such as the various
microstructure thickness, density,
or physicochemical etc., of of
property experiment materials),
transformer polymerthe containedinsulation
(cellulose) parameters should
materials
be revised
follow in advance.
similar variation rules (or laws) in the course of the degradation process. Therefore, if the
proposed model is supported to perform the aging condition prediction of polymer insulation under
7. Conclusions
other test conditions (such as the various thickness, density, etc., of experiment materials), the
contained
Since parameters
the moisture should beinside
content revised inunknown
the advance. sample is difficult to determine, the traditional
aging kinetics model is thus not appropriate to perform the aging condition prediction of transformer
7. Conclusions
polymer insulation materials. Fortunately, provided that the moisture content of the polymer materials
couldSince
be characterized
the moisture (replaced) by a certain
content inside feature parameter,
the unknown sample isthe above to
difficult issue will be solved
determine, naturally.
the traditional
According to this attempt, a modified aging kinetics model improved by employing the FDS technique
aging kinetics model is thus not appropriate to perform the aging condition prediction of transformer
is reported.
polymer The present
insulation analysisFortunately,
materials. and contributions havethat
provided led to themoisture
the followingcontent
conclusions.
of the polymer
1. The integral value of the tanδ curve (IV) can be utilized to characterize the moisture content
inside the transformer polymer insulation materials accurately. Therefore, relying on the analysis of
the relationship between the model parameters (ωDP and kDP ) and IV, its functional relationship can
be established;
2. Depending on the simultaneous analysis, this paper reported a modified aging kinetics model,
which could be applied to perform the aging condition prediction of transformer polymer insulation
materials under various test conditions, including aging duration, aging temperature, and initial
moisture content. Moreover, the average relative error of prediction results of oil-immersed polymer
(cellulose) pressboards is equal to 7.41%, which verified its accuracy and feasibility;
3. The reported fitting equations, models, and selected parameters significantly depend on the
individual materials and material conditions used during the experiments. Therefore, if the proposed
model is supported to be applied to other test conditions (such as the various thickness, density, etc., of
experiment materials), the contained parameters should be revised in advance.

Author Contributions: In this research activity, J.L. and X.F. designed the experiments and wrote the paper; Y.Z.
helped writing the paper; all authors have analyzed the data and approved the submitted manuscript.
Funding: This research was supported by the National Natural Science Foundation of China (61473272; 51867003;
51907034), 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
Polymers 2019, 11, 2082 14 of 15

Ability Promotion Project for Yong Teachers in Universities of Guangxi (2019KY0046; 2019KY0022), and the
Guangxi Bagui young scholars special funding in support of this research.
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. [CrossRef]
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. Zhang, Y.Y.; Li, X.; Zheng, H.B.; Yao, H.L.; Liu, J.F.; Zhang, C.H.; Peng, H.B.; Jiao, J. A Fault Diagnosis
Model of Power Transformers Based on Dissolved Gas Analysis Features Selection and Improved Krill Herd
Algorithm Optimized Support Vector Machine. IEEE Access 2019, 7, 102803–102811. [CrossRef]
4. International Electrotechnical Commission. Amendment 1-Measurement of the Average Viscometric Degree
of Polymerization of New and Aged Cellulosic Electrically Insulating Materials; International Electrotechnical
Commission: Geneva, Switzerland, 2007.
5. 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]
6. Liu, J.F.; Fan, X.H.; Zhang, Y.Y.; Zheng, H.B.; Jiao, J. Temperature Correction to Frequency Dielectric Modulus
and Activation Energy Prediction of Immersed Cellulose Insulation. IEEE Trans. Dielectr. Electr. Insul. 2019.
[CrossRef]
7. Zhang, Y.Y.; Liu, J.F.; Zheng, H.B.; 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]
8. Liu, J.F.; Fan, X.H.; Zhang, Y.Y.; Zheng, H.B.; Zhu, M.Z. Quantitative evaluation for moisture content of
cellulose insulation material in paper/oil system based on frequency dielectric modulus technique. Cellulose
2019. [CrossRef]
9. Liu, J.F.; Fan, X.H.; Zhang, Y.Y.; Zheng, H.B.; Yao, H.L.; Zhang, C.H.; Zhang, Y.B.; Li, D.J. A Novel Universal
Approach for Temperature Correction on Frequency Domain Spectroscopy Curveof Transformer Polymer
Insulation. Polymers 2019, 11, 1126. [CrossRef]
10. 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]
11. 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]
12. Emsley, A.M.; Xiao, X.; Heywood, R. Degradation of cellulosic insulation in power transformers. Part 2:
Formation of furan products in insulating oil. IEE Proc. Sci. Meas. Technol. 2000, 147, 110–114. [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. Liao, R.J.; Liu, J.F.; Yang, L.J.; Wang, K.; Hao, J.; Ma, Z.Q.; Gao, J.; Lv, Y.D. Quantitative analysis of insulation
condition of oil-paper insulation based on frequency domain spectroscopy. IEEE Trans. Dielectr. Elect. Insul.
2015, 22, 322–334. [CrossRef]
15. Liu, J.F.; Fan, X.H.; Zhang, Y.Y.; Zhang, C.H. Condition Prediction for Oil-immersed Cellulose Insulation in
Field Transformer Using Fitting Fingerprint Database. IEEE Trans. Dielectr. Electr. Insul. 2019. [CrossRef]
16. Ojha, S.K.; Purkait, P.; Chatterjee, B.; Chakravorti, S. Application of Cole–Cole model to transformer oil-paper
insulation considering distributed dielectric relaxation. High Volt. 2019, 4, 72–79. [CrossRef]
17. Zheng, Y.M.; Wang, Z.J. Study on Broadband Loss Characteristics of Oil-immersed Papers for Fast Transient
Modeling of Power Transformer. IEEE Trans. Dielectr. Electr. Insul. 2012, 20, 564–570. [CrossRef]
18. Morsalin, S.; Phung, B.T. Modeling of dielectric dissipation factor measurement for XLPE cable based on
Davidson-Cole model. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 1018–1026. [CrossRef]
Polymers 2019, 11, 2082 15 of 15

19. Dissado, L.A.; Hill, R.M. A cluster approach to the structure of imperfect materials and their relaxation
spectroscopy Proceedings of the Royal Society A: Mathematical. Phys. Eng. Sci. 1983, 390, 131–180.
[CrossRef]
20. Emsley, A.M. Kinetics and mechanisms of the low-temperature degradation of cellulose. Cellulose 1994, 1,
26–56. [CrossRef]
21. Emsley, A.M. On the kinetics of degradation of cellulose. Cellulose 1997, 4, 1–5. [CrossRef]
22. Calvini, P.; Gorassini, A.; Merlani, A.L. On the kinetics of cellulose degradation: Looking beyond the pseudo
zero order rate equation. Cellulose 2007, 15, 193–203. [CrossRef]
23. Calvini, P. The influence of levelling-off degree of polymerization on the kinetics of cellulose degradation.
Cellulose 2005, 12, 445–447. [CrossRef]
24. Ekenstam, A. The behavior of cellulose in mineral acid solutions: Kinetics study of the decomposition of
cellulose in acid solutions. BER 1936, 69, 540–553.
25. Ding, H.Z.; Wang, Z.D. On the degradation evolution equations of cellulose. Cellulose 2008, 15, 205–224.
[CrossRef]
26. Heywood, R.J.; Stevens, G.C.; Ferguson, C.; Emsley, A.M. Life assessment of cable paper using slow thermal
ramp methods. Thermochim. Acta 1999, 332, 189–195. [CrossRef]
27. Verma, H.C.; Baral, A.; Pradhan, A.K.; Chakravorti, S. 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]
28. Martin, D.; Cui, Y.; Ekanayake, C.; Ma, H. An Updated Model to Determine the Life Remaining of Transformer
Insulation. IEEE Trans. Power Deliv. 2015, 30, 395–402. [CrossRef]
29. Gillen, K.T.; Celina, M. The wear-out approach for predicting the remaining lifetime of materials. Polym.
Degrad. Stab. 2000, 71, 15–30. [CrossRef]
30. Urzhumtsev, Y.S. Time-temperature superposition. Rev. Mech. Compos. Mater. 1975, 11, 57–72.

© 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/).