polymers
Article
Thermal Aging Evaluation of XLPE Power Cable by Using
Multidimensional Characteristic Analysis of Leakage Current
Yong Liu * , Hao Wang, Han Zhang and Boxue Du
Citation: Liu, Y.; Wang, H.; Zhang,
H.; Du, B. Thermal Aging Evaluation
of XLPE Power Cable by Using
Multidimensional Characteristic
Analysis of Leakage Current.
Polymers 2022, 14, 3147. https://
doi.org/10.3390/polym14153147
Academic Editor: Catalin Gainaru
Received: 10 July 2022
Accepted: 29 July 2022
Published: 2 August 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 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 (https://
creativecommons.org/licenses/by/
4.0/).
Polymers 2022, 14, 3147. https://doi.org/10.3390/polym14153147
Tianjin Key Laboratory of Smart Energy and Information Technology, School of Electrical and Information
Engineering, Tianjin University, Tianjin 300072, China; wang_haoovo@163.com (H.W.);
zh605279246@163.com (H.Z.); duboxue@tju.edu.cn (B.D.)
* Correspondence: tjuliuyong@tju.edu.cn
Abstract: Thermal aging is a common form of cable deterioration. In this paper, the effect of thermal
aging on cables is evaluated by analyzing the harmonic characteristics in cable leakage currents.
Cable samples were first fabricated and subjected to accelerated thermal aging tests at 120 ◦ C. The
experimental circuits were built to test the dielectric loss factor and the AC leakage current of the cable
at different aging times. Then, the improved variational modal decomposition (VMD) algorithm
was used for the time–frequency analysis of the leakage current, and the relationship between
thermal aging and leakage current harmonics was investigated. Thermal aging was discovered
to increase the capacitance and dielectric loss factor of the cable as well as generate harmonics in
the leakage current, with harmonics at 150, 450, and 650 Hz being particularly sensitive to thermal
aging. The multidimensional characteristic parameters such as the time-domain, frequency-domain,
and relative energy and the sample entropy of the leakage current harmonics were calculated. The
results demonstrated thermal aging increased the relative energy and power spectrum energy of the
harmonics and increased the disorder of the harmonic sequence.
Keywords: XLPE cable; thermal aging; leakage current; harmonic characteristics; feature extraction
1. Introduction
With the development of urbanization and the increasing demand for reliability of
power supply, XLPE cables have been widely used in power networks because of their
excellent electrical performance, simple laying, and easy maintenance. Usually, the service
life of an XLPE cable is 20–30 years. However, with the extension of operation years, the
problems in the operation and maintenance process of the cable are gradually exposed. The
cables that were put into service earlier have started to show aging and deterioration [1,2].
Among them, thermal aging is considered as one of the most severe factors leading to
the degradation of the insulation performance of cables [3]. Therefore, it is important to
monitor the cable operation status, especially the thermal aging condition, to improve the
stability of the power supply system.
Thermal aging is a typical form of cable deterioration, and many scholars have studied
the mechanism of thermal aging. Kim C et al. [4] investigated the effect of thermal aging
on the insulation performance of XLPE and discovered that the insulation break-down
strength dropped as the aging period increased, while thermal aging caused a breaking of
the molecular chain structure. Kemari Y et al. [5] found that temperature can change the
crystalline state of XLPE, which leads to a decrease in the dielectric constant. Suo C et al. [6]
investigated the thermal aging stages of insulation and discovered that the properties of
the insulation did not change much in the first stage, whereas in the second stage, the XLPE
microscopic morphology was severely damaged due to crystalline disruption inside the
insulation, which reduced the electron binding effect. Andjelkovic D et al. [7] studied the
physicochemical properties of XLPE insulation. The results showed that the tensile strength
https://www.mdpi.com/journal/polymers
Polymers 2022, 14, x FOR PEER REVIEW 2 of 16
Polymers 2022, 14, 3147 2 of 16

the tensile strength and elongation at break of the XLPE cable’s insulation were affected
by theelongation
and crystallinityatunder
break thermal
of the XLPE agingcable’s
conditions, and the
insulation weretrend was the
affected by same as that of
the crystallinity
the crystallinity.
under Boukezzi
thermal aging L et al. [8]
conditions, andcharacterized
the trend was the the
microstructure
same as thatofofXLPE insulation
the crystallinity.
and found that
Boukezzi the[8]
L et al. high-temperature
characterized the environment
microstructure affected the crystallinity
of XLPE insulation and of XLPE
foundbut that
did
thenot change the crystalline
high-temperature phase. Mecheri,
environment affected theY. et al. [9] investigated
crystallinity of XLPEthe butchanges
did not in die-
change
the crystalline
lectric phase. Mecheri,
loss and mechanical Y. et al.
properties [9] investigated
of cross-linked the changes
insulation underinthermal
dielectric
aginglossandand
mechanical properties of cross-linked insulation under thermal aging
the test results showed that the dielectric loss of XLPE specimens became larger and the and the test results
showed that
mechanical the dielectric
properties decreasedloss after
of XLPE specimens
thermal aging. became larger and the mechanical
properties
Condition decreased
monitoringafter refers
thermal aging.
to the online monitoring of cable condition quantities to
obtainCondition monitoring
cable condition refers Common
information. to the online monitoring
methods includeofDC cable condition quantities
superposition method,
DCto obtain
componentcablemethod,
condition information.
dielectric Common
loss angle tangentmethods
methodinclude DC superposition
[10], partial dis-charge
method,
method DC component
[11–13], groundingmethod,current dielectric loss angle
method [14–16], etc.tangent
The DCmethod [10], partial
superposition method dis-
charge method [11–13], grounding current method [14–16], etc.
usually involves adding a DC power supply to the cable system, filtering out the AC com- The DC superposition
methodfinding
ponents, usuallythe
involves adding
insulation a DC power
resistance, supply to the
and evaluating the cable system, filtering
cable condition based on outthethe
AC components,
insulation resistancefinding
value.theThe
insulation resistance,
DC component and evaluating
method, on the otherthe cable
hand,condition
estimatesbased the
on the of
degree insulation resistance
deterioration value. The DC component
of XLPE-insulated method, on
cables by detecting thethe other hand,
magnitude of estimates
the DC
the degree[17].
component of deterioration
Zhu G et al. [18] of XLPE-insulated
proposed a novel cables
method by based
detecting the magnitudesignal
on low-frequency of the
DC component [17]. Zhu G et al. [18] proposed a novel method based
injection. This approach calculated the dielectric loss angle of the cable’s leakage current on low-frequency
signal
and injection. evaluated
consequently This approach calculated
the cable the dielectric
insulation. Rizzi A etloss angle
al. [19] of the cable’s
designed leakage
an automatic
current and consequently evaluated the cable insulation. Rizzi A
diagnosis system for partial discharges in cable accessories, by selecting the optimal sub-et al. [19] designed an
automatic diagnosis system for partial discharges in cable accessories,
set of features in the partial discharge information set and then classifying the complex by selecting the
optimal subset
discharge of features
pulse patterns in the partial
in different powerdischarge information
cable terminals set The
and joints. and grounding
then classifying
cur-
the method
rent complexcan discharge pulse patterns
obtain information on in
thedifferent
operating power cable of
condition terminals
the cableandby joints.
detecting The
grounding current method
changes in the grounding wire current.can obtain information on the operating condition of the cable
by detecting changes in the grounding wire current.
In this paper, the relationship between the thermal aging of cables and harmonic
In this paper, the relationship between the thermal aging of cables and harmonic
characteristics of ground currents was investigated by a time–frequency domain analysis
characteristics of leakage currents was investigated by a time–frequency domain analysis
and feature extraction methods. Cable samples were fabricated and subjected to thermal
and feature extraction methods. Cable samples were fabricated and subjected to thermal
aging tests at 120 °C.◦ The effects of thermal aging on the dielectric properties of cable in-
aging tests at 120 C. The effects of thermal aging on the dielectric properties of cable
sulation were analyzed by capacitance and dielectric loss factor tests. The leakage currents
insulation were analyzed by capacitance and dielectric loss factor tests. The leakage
of the cable samples were measured, and the harmonic characteristics of the leakage cur-
currents of the cable samples were measured, and the harmonic characteristics of the
rents were analyzed using the time–frequency domain method. Furthermore, the time–
leakage currents were analyzed using the time–frequency domain method. Furthermore,
frequency domain characteristic parameters of the harmonics were extracted, and the dis-
the time–frequency domain characteristic parameters of the harmonics were extracted, and
tribution of the characteristic parameters was examined. The work done will provide a
the distribution of the characteristic parameters was examined. The work done will provide
reference for the insulation evaluation of in-service XLPE cables.
a reference for the insulation evaluation of in-service XLPE cables.
2.2.Samples
Samplesand
andExperimentation
Experimentation
2.1.
2.1.Sample
SamplePreparation
Preparation
The
Theexperimental
experimental cable
cabletype
typewas
wasYJLV
YJLV 8.7/15 kV–1
8.7/15 × 70×
kV–1 mm702,mm
with 2 , an aluminum
with con-
an aluminum
conductive
ductive core core
and and an XLPE
an XLPE insulation
insulation layer,
layer, andand
thethe nominal
nominal thickness
thickness ofof insulationlayer
insulation layer
was 4.5 mm. Considering that the conductive core may discharge to
was 4.5 mm. Considering that the conductive core may discharge to the insulation shield the insulation shield
whenan
when anAC
AChighhighvoltage
voltageisisapplied,
applied,6060mmmmofofinsulation
insulationshield
shieldwas
wasremoved
removedfrom fromeach
each
endofofthe
end thesample
sample toto maintain
maintain sufficient
sufficient insulation
insulation spacing.
spacing. TheThe cable
cable sample
sample is shown
is shown in
in Figure
Figure 1. Due1. Due
to thetosize
the limitation
size limitation
of theofair
theblast
air blast
dryingdrying box,total
box, the the length
total length
of theof the
fab-
fabricated
ricated cablecable
samplesample was mm,
was 400 400 mm, of which
of which the length
the length of theofinsulation
the insulation shieldshield was
was 240
240 mm.
mm.

Insulation Insulation shield Conductor shield

Conductor layer layer layer

Figure 1. The cable sample.

Figure 1. The cable sample.
 Polymers 2022, 14, x FOR PEER REVIEW 3 of 16
Polymers 2022, 14, 3147 3 of 16

The aging temperature was set to 120 °C. The HK-450A+ electric thermostat blast
The aging temperature was set to 120 ◦ C. The HK-450A+ electric thermostat blast
drying box was applied to realize the thermal aging of the cable. The aging chamber was
drying box was applied to realize the thermal aging of the cable. The aging chamber was
adjusted to the required aging temperature, the sample was put in after the temperature
adjusted to the required aging temperature, the sample was put in after the temperature in
in the aging chamber reached the set value, and the sample was taken out for testing at
the aging chamber reached the set value, and the sample was taken out for testing at the
the aging time of 12 days, 24 days, and 48 days, respectively.
aging time of 12 days, 24 days, and 48 days, respectively.
2.2. Experimental
2.2. Experimental Setup
Setup
Before conducting
Before conducting the the experimental
experimental test,test, we
we used
used clean
clean alcohol
alcohol cotton
cotton balls
balls toto wipe
wipe
the cable sample surface, we applied a conductive silver paint on
the cable sample surface, we applied a conductive silver paint on the middle part of thethe middle part of the
insulationshield
insulation shieldof ofthe
thecable
cablesample,
sample,and andwe wewound
woundcopper
copperfoil foilas
asthe
thetest
testelectrode.
electrode.
The HY
The HY 1110-type
1110-type measuring
measuring device
device was
was used
used toto test
test the
thecapacitance
capacitance and anddielectric
dielectric
loss factor
loss factor of
of the
the cable.
cable. Each
Each cable
cable sample
sample waswas grounded
grounded and discharged
discharged before testing.
Thetest
The testwiring
wiringisisshown
shown in in Figure
Figure 2. The
2. The HV HV output
output terminal
terminal of theofHY1110
the HY1110 was con-
was connected
nected to the cable conductor through a HV shield wire, and the
to the cable conductor through a HV shield wire, and the sample input terminal of thesample input terminal of
the HY1110
HY1110 was was connected
connected to thetotest
theelectrode
test electrode
of theofcable
the cable
sample sample through
through a shielda shield wire;
wire; then,
then,
the the HY1110
HY1110 was reliably
was reliably grounded.
grounded. After After the wiring
the wiring was completed,
was completed, the testthevoltage
test voltage
was
wastoset
set to 3and
3 kV kV and
eacheach
cablecable
samplesample
waswas tested
tested 10 10 times.
times. Afterthe
After thetest
testofofeach
eachcable
cablewas
was
completed,
completed, the the cable
cable sample
sample was was discharged
discharged using
using aa discharge
discharge rod, rod, and
and then
then thethe cable
cable
sample
sample waswas replaced
replacedforforthe
thenext
nexttest.
test.

HV shielded wire
Test electrode

HV
output

HY1110
Sample Shielded wire
input

GND

Figure2.
Figure 2. The
The cable
cable capacitance
capacitance and
anddielectric
dielectricloss
lossfactor
factormeasurement
measurementdevice.
device.

The leakage
The leakage current
current test
test experimental
experimental setupsetup is
is shown
shown in in Figure
Figure 3. 3. The
The HVAC
HVAC power
power
consisted of
consisted of aa voltage
voltage regulator
regulator andand aa test
test transformer,
transformer,which
whichcould
could generate
generateAC AC voltages
voltages
of up
of up to
to 50
50 kV.
kV.The The protection
protection resistor
resistor had
had aa resistance
resistance value
value ofof 11 MΩ
MΩ to to prevent
prevent short-
circuitovercurrent
circuit overcurrentshocksshockstotothethe test
test system.
system. The
The high-voltage
high-voltage output
output waswas connected
connected to
to the
cable sample
the cable sampleconductor, while
conductor, the test
while the electrode of the
test electrode ofcable sample
the cable samplewas was
connected to the
connected to
sampling
the samplingresistor. The sampling
resistor. The samplingresistor with awith
resistor resistance value of
a resistance 10 kΩ,
value which
of 10 kΩ, was placed
which was
in an electromagnetic
placed shielding
in an electromagnetic box to reduce
shielding box toexternal interference,
reduce external converted
interference, the current
converted the
flowing
current through
flowing the insulation
through layer into layer
the insulation a voltage
into signal to besignal
a voltage collected
to bebycollected
the acquisition
by the
card and stored
acquisition cardon a PC,
and where
stored on athePC,sampling
where thefrequency
sampling range of the acquisition
frequency range of thecard was
acquisi-
adjustable
tion card wasfromadjustable
0 to 250 MHz.
from Meanwhile,
0 to 250 MHz. theMeanwhile,
voltage applied to the cable
the voltage sample
applied during
to the cable
the experiment
sample during the wasexperiment
8.7 kV, andwas this8.7
synchronization signal was collected
kV, and this synchronization signalby means
was of a
collected
voltage divider and an oscilloscope.
by means of a voltage divider and an oscilloscope.
Polymers
Polymers2022,
2022,14,
14,xxFOR
FORPEER
PEERREVIEW
REVIEW 4 4ofof1616
Polymers 2022, 14, 3147 4 of 16

Protective
Protective
resistor Oscillograph
Oscillograph
resistor

Voltage
Voltage
divider
divider
Test
Testelectrode
electrode GND
HVAC Cable
Cable GND
HVAC
power sample
sample
power

Sampling
Sampling Acquisition
Acquisition
resistor
resistor card
card
GND
GND

GND
GND Electromagnetic
Electromagnetic PC
shielding PC
shieldingbox
box
Figure
Figure3. The
TheAC
ACleakage
leakagecurrent
currenttest
testexperimental setup.
Figure 3.3.The AC leakage current test experimentalsetup.
experimental setup.

2.3.
2.3.Experimental
ExperimentalResults
Results
The
The change ofcapacitance
change of capacitanceandanddielectric
dielectricloss
lossfactor
factorofofcable
cablesample
samplewith
withaging
agingtime
time
shown in Figure 4. From the figure, it is seen that the capacitance and dielectricloss
isisshown in Figure 4. From
From the
the figure, it is seen that the capacitance and dielectric loss
factor
factorincrease
increasegradually
graduallywith theaging
withthe agingtime.
time.After
After4848days
daysofofaging,
aging,the
thecapacitance
capacitanceand
and
dielectric
dielectric loss
dielectricloss factor
lossfactor
factorofof the
ofthe cable
thecable sample
cablesample increase
sampleincrease by
increaseby 3.576%
by3.576%
3.576%andand 126.354%,
and126.354%, respectively,
126.354%,respectively,
respectively,
compared with
with those
those of
of the
the unaged
unaged sample.
sample. Therefore,
Therefore, the the thermal
thermal
compared with those of the unaged sample. Therefore, the thermal aging makes agingaging makes
makes the the
thedie-
die-
dielectric
lectric properties
properties of of
XLPEXLPE insulation
insulation deteriorate
deteriorate andand
the the
lossloss
lectric properties of XLPE insulation deteriorate and the loss increase. increase.
increase.

Figure
Figure4.4.The
Thechanges
changesinincapacitance
capacitanceand
anddielectric
dielectricloss
lossfactor
factorofofcable
cablesample
samplewith
withaging
agingtime.
time.

The
Theresults
resultsofofthe
thecable
cableleakage
leakagecurrent
currenttest
testare
areshown
shownininFigure
Figure5.5.ItItcan
canbe
beseen
seenfrom
from
the
the figure
figure that
that for
for the
the unaged
unaged cable,
cable, the
the leakage
leakage current
current is
is approximately
approximately sinusoidal,
sinusoidal, while
while
the figure that for the unaged cable, the leakage current is approximately sinusoidal, while
after the
after the accelerated aging
aging experiment, there there is a significant harmonic distortion in
in the
after theaccelerated
accelerated agingexperiment,
experiment, thereisisaasignificant
significantharmonic
harmonicdistortion
distortion inthe
the
leakage current.
leakage current. In addition, the the leakage current
current amplitude increases
increases gradually with with the
leakage current.InInaddition,
addition, theleakage
leakage currentamplitude
amplitude increasesgradually
gradually withthe the
extension of
extension of the aging
aging time, and
and the amplitude increases by
by 9.09%, 16.36%,
16.36%, and 38.18%
38.18%
extension ofthe the agingtime,
time, andthe theamplitude
amplitudeincreases
increases by9.09%,
9.09%, 16.36%,and and 38.18%
after
after 12, 24, and 48 days of aging, respectively, compared with that of the unaged sample.
after12,
12,24,
24,and
and48 48days
daysofofaging,
aging,respectively,
respectively,compared
comparedwith
withthat
thatofofthe
theunaged
unagedsample.
sample.
Polymers 2022, 14, 3147
x FOR PEER REVIEW 5 5of
of 16

Leakage current (μA)

Leakage current (μA)

Time (s) Time (s)

(a) (b)
Leakage current (μA)

Leakage current (μA)

Time (s) Time (s)

(c) (d)
Figure
Figure 5.
5. The leakage current
The leakage currenttest
testresults
resultsofofcable
cablesamples.
samples.(a)
(a)Unaged;
Unaged;(b)
(b) aging
aging forfor
1212 days;
days; (c)(c)
aging
aging for 24 days; (d) aging for 48 days.
for 24 days; (d) aging for 48 days.

3. Time–Frequency
Time–Frequency Domain Analysis and Multidimensional Characteristic Parameter
Extraction of Leakage Current
3.1. Time–Frequency
3.1. Time–Frequency Domain
Domain Analysis
Analysis of
of Leakage
Leakage Current
Current
In order
In order to
to further
furtherextract
extractthe
theinformation
informationcontained
containedininthe leakage
the current,
leakage thethe
current, leakage
leak-
current was analyzed using the time–frequency analysis method. Dragomiretskiy
age current was analyzed using the time–frequency analysis method. Dragomiretskiy et al. [20]
et
proposed a variational mode decomposition (VMD) as a variational mode-based
al. [20] proposed a variational mode decomposition (VMD) as a variational mode-based signal
processing method that could decompose the signal x(t) into a sequence of intrinsic mode
signal processing method that could decompose the signal x(t) into a sequence of intrinsic
function (IMF(t)) and residuals (res(t)). The original signal x(t) is the sum of IMF(t) and
mode function (IMF(t)) and residuals (res(t)). The original signal x(t) is the sum of IMF(t)
res(t), as shown in Equation (1) [21].
and res(t), as shown in Equation (1) [21].
kk
x (xt()t )=∑ I IMF
MFi (t()t +
) res (t()t ) (1)

i =1
i 1
res
i (1)

Before performing
Before performing the the VMD
VMD on on the
the signal,
signal, two
two parameters,
parameters, the
the modal
modal number
number K K and
and
the quadratic
the quadratic penalty
penalty factor
factor α,
α, need
need toto be
be preset,
preset, where
where K K determines
determines the
the number
number of of IMFs
IMFs
obtained by VMD, and α determines the bandwidth of each IMF. K and
obtained by VMD, and α determines the bandwidth of each IMF. K and α have a signifi- α have a significant
impact
cant on the
impact onVMD
the VMDresults, so this
results, so paper usedused
this paper particle swarm
particle optimization
swarm (PSO)
optimization to find
(PSO) to
the optimal K and α in the VMD. The flow chart of the PSO–VMD algorithm
find the optimal K and α in the VMD. The flow chart of the PSO–VMD algorithm is shown is shown in
Figure 6. The fitness function was the standard deviation of the fuzzy entropy
in Figure 6. The fitness function was the standard deviation of the fuzzy entropy of a set of a set of
IMFs
of obtained
IMFs obtained after the
after theVMD,
VMD,andanditsitsexpression
expressionisisshown
shownininEquation
Equation (2).
(2). The specific
The specific
procedure for calculating the fuzzy entropy is given by Reference
procedure for calculating the fuzzy entropy is given by Reference [22]. The standard[22]. The standard
 deviation can reflect the degree of data dispersion, and the fitness function is minimized
when the optimal combination of K and α is obtained.
Polymers 2022, 14, 3147 6 of 16
 1 K
 Fitness ( K ,  ) 

 ( FuzzyEni   )2
K i 1
 (2)
deviation can reflect thedegree1 Kof data dispersion, and the fitness function is minimized
  = K 
when the optimal combination
FuzzyEni
i 1 K and α is obtained.
of

We set the range of Kto 5~30 and the range ofK α to 1500~6000 and executed the opti-
 s
Fitness(K,
mization process. Figure 7illustrates = K1 process
α)change
the ∑ ( FuzzyEn
of thei − µ)2 of the leakage cur-
fitness


i =1
rent after aging for 12 days. From the figure, it can be seen that the fitness curve shows a (2)
K
= K1 converging,

trend of first decreasing and
 µthen ∑ FuzzyEniand the convergence is completed in the


35th generation. The optimal fitness i =1was 0.186103, and the K and α at that time were 11

and 5070.33, respectively.

Start

Original signal

Initialize PSO
parameters

VMD

Calculate the value of fitness function

Std_FuzzyEn for each particle

No Evaluation of fitness
function values

Update the position and

velocity of particles

Reach the maximum number of

iterations M?

Yes

BestK and Bestα

VMD

End

Figure 6. The
Figure flowflow
6. The chart of PSO–VMD
chart algorithm.
of PSO–VMD algorithm.

We set the range of K to 5~30 and the range of α to 1500~6000 and executed the
optimization process. Figure 7 illustrates the change process of the fitness of the leakage
current after aging for 12 days. From the figure, it can be seen that the fitness curve shows
a trend of first decreasing and then converging, and the convergence is completed in the
35th generation. The optimal fitness was 0.186103, and the K and α at that time were 11
and 5070.33, respectively.
Polymers 2022, 14, x FOR PEER REVIEW 7 of 16

Polymers 2022, 14, x FOR PEER REVIEW 7 of 16

Polymers 2022, 14, 3147 7 of 16

Figure 7. The fitness in the optimization process.

After
Figure
Figure
the
The
7.7.The
VMD,
fitness
fitness
Kthe
ininthe
IMF components
optimization
optimization
can be obtained, and the instantaneous frequency
process.
process.
and instantaneous amplitude of IMF components can be obtained by applying the Hilbert
transform
After on
Afterthe the
theVMD, IMF
VMD, components.
KKIMF
IMFcomponents
components Forcan
the
canbe time signal x(t),
beobtained,
obtained, its
andthe
and theHilbert transform
instantaneous
instantaneous expres-
frequency
frequency
sion
and [23] is
andinstantaneous
instantaneousamplitude
amplitudeofofIMF
IMFcomponents
componentscan canbe beobtained
obtainedby byapplying
applyingthe
theHilbert
Hilbert
transformon
transform onthe
theIMF
IMFcomponents.
components. ForFor
thethe time
time signal
signal x(t),
x(t), its
its Hilbert
Hilbert transform
transform expres-
expres-
 1  x(t )
sion [23]
sion [23] isis x ( t )  H [ x (t )] 
Z 
d  (3)
∧ 1  +∞ t x(t)
x (t) = H [ x (t)] = dτ (3)
1  x(t ) diagram of the ground current
The time-domain diagram xand (t ) the (t )] π −∞ t −dτ
H [ xtime-frequency  (3)
signalThe
after the VMD are
time-domain shownand
diagram in Figure 8. The 
the time-frequency t   diagram
time-domain diagram shows that
of the leakage each
currents
IMF obtained
signal
Theafter after
the VMD
time-domain theareVMD is aand
shown
diagram sinusoidal
in Figure wave,
8. The and thediagram
time-frequency
time-domain
the time-frequency diagram diagram
shows
of the ground further
that each
current
illustrates
IMF that
obtained these
after the IMFs
VMD are
is a actually
sinusoidal fundamental
wave, and and
the odd harmonics,
time-frequency
signal after the VMD are shown in Figure 8. The time-domain diagram shows that each so
diagram that the
further
ground current
illustrates that is
thesecomposed
IMFs are of a fundamental
actually fundamental and a
and series
odd of odd harmonics.
harmonics,
IMF obtained after the VMD is a sinusoidal wave, and the time-frequency diagram further so that theAs aging
leakage
proceeds,
currents isthe
illustrates thatamplitude
composed of aoffundamental
these IMFs each odd harmonic
are actually and increases
a series
fundamental of odd compared
and harmonics.to the
Asunaged
odd harmonics, aging sample,
so proceeds,
that the
the amplitude
which can of each
indicate thatodd harmonic
there is a increases
certain compared
connection to
between the unaged
aging
ground current is composed of a fundamental and a series of odd harmonics. As aging andsample,
harmonicwhich can
gener-
indicate
ation. that there is a certain connection between aging and harmonic
proceeds, the amplitude of each odd harmonic increases compared to the unaged sample, generation.
which can indicate that there is a certain connection between aging and harmonic gener-
Time-frequency diagram
Time-domain diagram
ation.
LC

Time-domain diagram Time-frequency diagram

LC

Frequency (Hz)
Frequency (Hz)

Time (s) Time (s)

(a)
Time (s) Time (s)
Figure 8. Cont.
(a)
 Polymers2022,
Polymers 2022,14,
14,3147
x FOR PEER REVIEW 88 of
of 16
16

Time-domain diagram Time-frequency diagram

LC

Frequency (Hz)

Time (s) Time (s)

(b)
Time-domain diagram
Time-frequency diagram
LC

Frequency (Hz)

Time (s) Time (s)

(c)

Figure 8. Cont.
Polymers 2022, 14, 3147
x FOR PEER REVIEW 9 9of
of 16

Time-domain diagram
LC Time-frequency diagram

Time (s) Time (s)

(d)
Figure
Figure 8.
8. The
Thetime-frequency
time-frequencydistribution
distributionofof
leakage current.
leakage (a) (a)
current. Unaged; (b) aging
Unaged; for 12
(b) aging fordays; (c)
12 days;
aging for 24 days; (d) aging for 48 days
(c) aging for 24 days; (d) aging for 48 days.

3.2. Harmonic Selection and Signal Reconstruction

From thethe analysis
analysisininthe
theprevious
previoussection,section,the thethermal
thermal aging
agingcable ground
cable leakagecurrent has
currents
ahas
series of odd
a series of harmonics,
odd harmonics,whichwhichhold information
hold information about about
the cable’s thermal
the cable’s aging condi-
thermal aging
condition.
tion. A feature
A feature extraction
extraction can calculate
can calculate and select
and select somesome completecomplete and nonredundant
and nonredundant fea-
feature
ture quantities
quantities from
from thethe leakage
leakage current,which
current, whichcan caneffectively
effectivelyindicate
indicatethe
thethermal
thermal aging
aging
condition of cables.
Before feature
feature extraction,
extraction,the thedegree
degreeofofcorrelation
correlation between
between each order
each of odd
order harmon-
of odd har-
ics and the thermal aging state of the cable was evaluated to screen
monics and the thermal aging state of the cable was evaluated to screen out specific har- out specific harmonics
that were
monics more
that weresensitive to changes
more sensitive in the in
to changes thermal agingaging
the thermal state. state.
In this
In paper, Pearson’s
this paper, Pear-
correlation
son’s coefficient
correlation ρ and the
coefficient harmonic
ρ and current content
the harmonic currentrate HRI were
content usedwere
rate HRI as evaluation
used as
indexes. indexes.
evaluation
The expression
The expression ofof Pearson’s
Pearson’s correlation
correlation coefficient
coefficient ρρ is is [24]
[24]
− −
  
n n
  − Y 
∑ XiX−i XX Y Yii  Y 
i =1 i 1   
ρ X,Y = X ,v
Y  s (4)
(4)
 2 n   2 2
−  2
u
u n n  n  −
∑ X iX−i XX   ∑  YY i 
i− Y Y
t
i =1 i 1   ii=11 

where
where ρ ρx,y isisthe
thecorrelation
correlationcoefficient
coefficientbetween
betweenseries
seriesXXand
and Y,
Y, which
which reflects
reflects the
the degree
degree
X,Y
of linear correlation between series X and
of linear correlation between series X and Y. Y.
The
The expression
expression of of the
the harmonic current content
harmonic current content rate
rate HRI
HRIh is
is h
I
HRI h Ih h  100% (5)
HRIh = I1× 100% (5)
I1
where Ih is the effective value of the h-order harmonic current, I1 is the effective value of
where Ih is the effective value of the h-order harmonic current, I1 is the effective value of
the fundamental current, and HRI reflects the content of the harmonics of each order.
the fundamental current, and HRI reflects the content of the harmonics of each order.
Table 1 shows the calculation results of ρ and HRI between each order of harmonics
Table 1 shows the calculation results of ρ and HRI between each order of harmonics
and leakage current. As can be seen from the table, for the unaged cable, the ρ of the 150
and leakage current. As can be seen from the table, for the unaged cable, the ρ of the 150 Hz
Hz harmonic
harmonic is the
is the highest,
highest, but
but it is it is only
only 0.0273,
0.0273, and and its HRI
its HRI is 2.0761%.
is 2.0761%. WithWith the increas-
the increasing of
ing of harmonic order, the ρ and HRI gradually decrease. For the cable after aging for 12
Polymers 2022, 14, 3147 10 of 16

harmonic order, the ρ and HRI gradually decrease. For the cable after aging for 12 days,
the harmonics with a higher ρ and HRI are 450 Hz, 150 Hz, and 650 Hz in order; for cables
with 24 and 48 days of thermal aging, they are 650 Hz, 150 Hz, and 450 Hz and 150 Hz,
450 Hz, and 650 Hz, respectively.

Table 1. The ρ and HRI between each order of harmonics and leakage current.

Unaged Aging 12 Days Aging 24 Days Aging 48 Days

Harmonic Frequency (Hz)
ρ HRI (%) ρ HRI (%) ρ HRI (%) ρ HRI (%)
150 0.0273 2.0761 0.1348 13.2395 0.1445 14.4843 0.1478 14.7071
250 0.0127 0.8584 0.0363 2.6265 0.0586 5.8067 0.0379 3.0394
350 0.0081 0.4936 0.0705 7.5498 0.0469 4.6530 0.0496 4.6078
450 0.0064 0.4367 0.1363 14.8848 0.0984 10.7732 0.1334 14.4038
550 0.0069 0.4311 0.0621 6.6498 0.0540 5.7957 0.0513 5.3084
650 0.0051 0.4188 0.0770 8.4036 0.1424 15.8693 0.0852 9.0979
750 0.0049 0.4198 0.0648 7.0084 0.0784 8.5950 0.0600 6.4019
850 0.0057 0.4246 0.0686 7.4309 0.04132 4.4566 0.0260 2.5617
950 - - 0.0508 5.5170 0.0329 3.5663 0.0497 5.2813

From the above analysis, it can be seen that the harmonics that are sensitive to the
thermal aging state were at 150 Hz, 450 Hz, and 650 Hz. In this paper, the harmonics of
these three frequencies were superimposed to obtain the reconstructed signal, and the
features of the reconstructed signal were extracted.

3.3. Multidimensional Characteristic Parameter Extraction

In this section, for the leakage current data of the unaged cable and the cable after
aging for 12 days, 24 days, and 48 days, 500 groups of reconstructed signals x(i), i = 1, 2, . . . ,
N (N = 4000) were obtained, respectively, according to the above method. The time-domain,
frequency-domain, and relative energy and the sample entropy characteristic parameters
of these reconstructed signals were calculated, and the variation and distribution of these
characteristic parameters were analyzed by using a statistical graph.
(1) Time-domain characteristic parameters
The time-domain characteristic parameters are shown in Table 2 [25].

Table 2. The time-domain characteristic parameters.

Characteristic Parameter Expression Characteristic Parameter Expression

s
N N
Mean value X= 1
∑ x (i ) Root-mean-square value 1
N Xrms = N ∑ x (i )
i =1 i =1
N N 2
Variance 1 2 Square root amplitude
σ2 = ∑ ( x (i ) − X ) Xr = [ N1 ∑
p
N −1 | x (i )|]
i =1 i =1
N Xmax
Absolute mean value 1 Crest factor Cf =
 
X =
N ∑ | x (i )| Xrms
i =1
Maximum value Xmax = max{ x (i )} Waveform factor Xrms
Sf =
|X|
Minimum value Xmin = min{ x (i )} Pulse factor I f = Xmax
|X|
Peak-to-peak value Xpp = Xmax − Xmin Margin factor CL f = XXmax
r

For the reconstructed signal, the maximum value, peak-to-peak value, waveform factor,
peak factor, pulse factor, and margin factor were extracted, where the maximum value and
peak-to-peak value were dimensioned characteristic parameters, and their distributions are
shown in Figure 9. From the figure, it can be seen that thermal aging makes the maximum
and peak-to-peak values of the reconstructed signals increase significantly, and the degree
Polymers 2022, 14, x FOR PEER REVIEW 11 of 16

Polymers 2022, 14, 3147 11 of 16

distributions are shown in Figure 9. From the figure, it can be seen that thermal aging
makes the maximum and peak-to-peak values of the reconstructed signals increase sig-
of increaseand
nificantly, is also greaterof
the degree with the increase
increase of the aging
is also greater time.
with the In addition,
increase thermal
of the aging aging
time. In
makes their
addition, distribution
thermal range their
aging makes larger.
distribution range larger.

(a) (b)
Figure
Figure9.9.The
Thedistribution
distributionofof
dimensioned parameters
dimensioned in the
parameters time
in the domain.
time (a) Maximum
domain. value;
(a) Maximum (b)
value;
peak-to-peak value.
(b) peak-to-peak value.

The
Thedistribution
distributionofofthe
thewaveform
waveformfactor,
factor,peak
peakfactor,
factor,pulse
pulsefactor,
factor,and
andmargin
marginfactor
factor
of the reconstructed signal is shown in Figure 10. It can be seen from the
of the reconstructed signal is shown in Figure 10. It can be seen from the figure figure that that
the
waveform factor shows an increasing-decreasing-increasing trend as the aging
the waveform factor shows an increasing-decreasing-increasing trend as the aging time time in-
creases,
increases, while the peak factor, pulse factor, and margin factor show an increasing trendin
while the peak factor, pulse factor, and margin factor show an increasing trend in
fluctuation.
fluctuation.

(a) (b)

Figure 10. Cont.

Polymers
Polymers 2022,
2022, 14,
14, x3147
FOR PEER REVIEW 1212ofof 16
16

(c) (d)
Figure
Figure 10.
10. The
The distribution
distribution of
of dimensionless
dimensionlessparameters
parametersin
inthe
thetime
timedomain.
domain.(a)
(a)Waveform
Waveformfactor;
factor;
(b) crest factor; (c) pulse factor; (d) margin factor.
(b) crest factor; (c) pulse factor; (d) margin factor.

(2)
(2) Frequency-domain
Frequency-domaincharacteristic
characteristicparameters
parameters
First, the power spectrum P(m)
First, the power spectrum P(m) of the of the reconstructed
reconstructed signal
signal waswas calculated,
calculated, m =m = 1,2,
1, 2, ... ,
..., M, M was the number of spectral lines, and then the mean frequency,
M, M was the number of spectral lines, and then the mean frequency, standard deviation standard devia-
tion frequency,
frequency, centercenter of gravity
of gravity frequency,
frequency, and root-mean-square
and root-mean-square frequencyfrequency of the recon-
of the reconstructed
structed
signal were calculated, and their expressions are given in Table 3 [26]. The mean The
signal were calculated, and their expressions are given in Table 3 [26]. mean
frequency
frequency reflects the magnitude of the power spectrum energy, the standard
reflects the magnitude of the power spectrum energy, the standard deviation frequency deviation
frequency
describes thedescribes
degreetheofdegree of dispersion
dispersion or concentration
or concentration of the spectrum,
of the spectrum, while thewhile the
center
center of gravity frequency and root-mean-square frequency reflect the
of gravity frequency and root-mean-square frequency reflect the variation of the main variation of the
main frequency
frequency band band position.
position.

Table
Table 3.
3. The
The frequency-domain
frequency-domain characteristic parameters.
characteristic parameters.
Characteristic Parameter Expression
Characteristic Parameter Expression
M
 1
Mean frequency
Mean frequency
f −
fM
=

1
m
PM(m)
∑ P(m)
M1
M
s m =1
Standard deviation
Standard frequency
deviation frequency 
n 2
f    M f k -f c   P2( m ) / m
o
f σ = m  1 ∑ [ f k − f c ] × P(m) /m 
m =1
M M
Center of gravity frequency  f∑ [ f P×(Pm(m))]
fc =m 1
k1
m=
M
k

Center of gravity frequency fc  M ∑ P(m)

m =1

Root-mean-square frequency
s
M 2P(m)
h i M
f RMS = ∑ mf 1
k × P(m) / ∑ P(m)
M m =1 M m =1
2
Root-mean-square frequency f RM S    f
m 1
k  P ( m )   P (m)
m 1

As can be seen in Figure 11, thermal aging causes the frequency-domain parameters
of theAsreconstructed
can be seen insignal
Figureto 11,
increase.
thermalThe increase
aging causesof the
the frequency-domain
mean frequency indicates that
parameters
the power spectrum energy of the reconstructed signal increases, and the increase
of the reconstructed signal to increase. The increase of the mean frequency indicates that of the
frequency standard deviation indicates that the frequency distribution of the
the power spectrum energy of the reconstructed signal increases, and the increase of the reconstructed
signal tendsstandard
frequency to be dispersed.
deviationTheindicates
changes in thethe
that center of gravity
frequency frequency and
distribution root-mean-
of the recon-
square frequency
structed indicate
signal tends to bethat the thermal
dispersed. aging shifts
The changes thecenter
in the position
of of the main
gravity band ofand
frequency the
reconstructed signal to the high-frequency direction.
root-mean-square frequency indicate that the thermal aging shifts the position of the main
band of the reconstructed signal to the high-frequency direction.
Polymers2022,
Polymers 14, x3147
2022, 14, FOR PEER REVIEW 1313ofof16
16

(a) (b)

(c) (d)
Figure
Figure 11.
11. The
The distribution
distribution of
of frequency-domain
frequency-domain parameters.
parameters. (a)
(a)Mean
Meanfrequency;
frequency;(b)
(b) frequency
frequency
standard deviation; (c) center of gravity frequency (d) root-mean-square frequency.
standard deviation; (c) center of gravity frequency (d) root-mean-square frequency.

(3)
(3) Relative
Relativeenergy
energyand
andsample
sampleentropy
entropy
The
The expression of the relative energy
expression of the relative energy RE
REk isis k
E
REk Ek k  100% (6)
REk = E0× 100% (6)
E0
where Ek is the energy of the reconstructed signal after 12, 24, and 48 days of aging, and
where Ek is the energy of the reconstructed signal after 12, 24, and 48 days of aging, and E0
E0 is the energy of the reconstructed signal that the unaged.
is the energy of the reconstructed signal that the unaged.
The sample entropy of the reconstructed signal was also calculated, which reflects
The sample entropy of the reconstructed signal was also calculated, which reflects the
the complexity of the reconstructed signal. The procedure for calculating the sample en-
complexity of the reconstructed signal. The procedure for calculating the sample entropy
tropy can be obtained from the reference [27]. The value of the sample entropy is related
can be obtained from the reference [27]. The value of the sample entropy is related to the
to the values
values of theof the pattern
pattern dimension
dimension m and mthe
andsimilarity
the similarity tolerance
tolerance r. Ther. results
The results of pre-
of previous
vious
studies have shown [28] that the sample entropy calculated by taking m = 1 or m =or2 m
studies have shown [28] that the sample entropy calculated by taking m = 1 and=
2r and r = 0.1–0.25 Std (Std is the standard deviation of the original data)
= 0.1–0.25 Std (Std is the standard deviation of the original data) has more reasonable has more reason-
able statistical
statistical properties.
properties. Therefore,
Therefore, wemset= m
we set = 2 and
2 and r = Std.
r = 0.2 0.2 Std.
The
The distributions of the relative energy and sample entropyare
distributions of the relative energy and sample entropy areshown
shownininFigure
Figure12.12.
From
From thethe figure, it can
figure, it can be
be seen
seen that
thatthe
therelative
relativeenergy
energyofofthethereconstructed
reconstructed signal
signal gradu-
gradually
ally increases
increases withwith
thethe extension
extension of of
thethe thermalaging
thermal agingtime,
time,andandthe
the two
two show
show aa positive
positive
correlation,
correlation, which indicates that the thermal aging drives the generation anddevelopment
which indicates that the thermal aging drives the generation and development
of harmonics. In addition, the sample entropy of the reconstructed signal becomes larger
 Polymers 2022, 14, 3147 14 of 16
Polymers 2022, 14, x FOR PEER REVIEW 14 of 16

of harmonics. In addition, the sample entropy of the reconstructed signal becomes larger
afterafter thermal
thermal aging,
aging, which
which indicates
indicates thatthat thermal
thermal aging
aging makes
makes the the reconstructed
reconstructed signal
signal
sequence disorder increase.
sequence disorder increase.

(a) (b)
Figure 12. The
Figure distribution
12. The of relative
distribution energy
of relative and and
energy sample entropy.
sample (a) Relative
entropy. energy;
(a) Relative (b) sample
energy; (b) sample
entropy.
entropy.

4. Conclusions
4. Conclusions
In this paper,
In this cable
paper, samples
cable sampleswere fabricated
were fabricated andand
accelerated
accelerated aging experiments
aging experiments were were
conducted,
conducted, test test
circuits were
circuits built,
were andand
built, the the
dielectric
dielectriclossloss
factor andand
factor leakage
leakagecurrent teststests
current
were performed
were performed on cables under
on cables different
under aging
different agingtimes. Further,
times. a time–frequency
Further, a time–frequency analysis
analysis
of the leakage
of the leakage current signal
current signalwas
was carried
carried out,
out,focusing
focusingon onthethevariation
variationof of harmonics
harmonics in the
the leakage
leakage current,
current,andandthe
themultidimensional
multidimensionalcharacteristic
characteristic parameters
parameters of of
sensitive
sensitiveharmonics
har-
werewere
monics extracted and and
extracted analyzed. The The
analyzed. mainmainconclusions
conclusionsobtained werewere
obtained as follows:
as follows:
(1) (1)
The The capacitance
capacitance of cable
of the the cable sample,
sample, the dielectric
the dielectric lossloss factor,
factor, andandthe the amplitude
amplitude of of
the leakage current of the cable sample gradually increased with the extension of the the
the leakage current of the cable sample gradually increased with the extension of
agingaging
time.time.
ThisThis indicated
indicated that that the thermal
the thermal agingaging
made made
the the dielectric
dielectric properties
properties of of
XLPEXLPE insulation
insulation deteriorate
deteriorate andand the losses
the losses increase.
increase.
(2) (2) Thermal
Thermal aging
aging drove
drove the the generation
generation of odd
of odd harmonic
harmonic componentsininthe
components thecable
cableleak-
leakage
age current
current and andthetheamplitude
amplitudeofofthese theseharmonic
harmonic components
components correlated
correlated with the the
with degree
degree of thermal aging. Among the odd harmonic components, the harmonics at Hz,
of thermal aging. Among the odd harmonic components, the harmonics at 150
450 Hz,
150 Hz, 450 and 650 Hz
Hz, and 650were moremore
Hz were sensitive to thetochanges
sensitive the changesin theindegree of thermal
the degree aging.
of ther-
(3) Thermal
mal aging. aging increased the relative energy and power spectrum energy of the
sensitive harmonics and increased the disorder of the
(3) Thermal aging increased the relative energy and power spectrum energy of the sen-sensitive harmonic sequence.
sitive harmonics and increased the disorder of the sensitive harmonic sequence.
Author Contributions: Conceptualization, Y.L. and B.D.; methodology, Y.L. and H.W.; software, H.W.
and Contributions:
Author H.Z.; validation,Conceptualization,
H.W. and H.Z.; formal analysis,
Y.L. and B.D.; Y.L. and H.W.; investigation,
methodology, Y.L. and H.W.;H.W. and H.Z.;
software,
H.W.resources,
and H.Z.;Y.L.; data curation,
validation, H.W. andH.W.; writing—original
H.Z.; formal analysis,draft preparation,
Y.L. and H.W. and H.Z.;
H.W.; investigation, H.W.writing—
and
H.Z.;review and editing,
resources, Y.L.;
Y.L.; data visualization,
curation, H.W.; supervision,
H.W.; writing—original Y.L.preparation,
draft and B.D.; project
H.W.administration,
and H.Z.; writ-Y.L.;
funding acquisition,
ing—review and editing,Y.L. Allvisualization,
Y.L.; authors have read and
H.W.; agreed to the
supervision, published
Y.L. version
and B.D.; projectofadministra-
the manuscript.
tion, Y.L.; funding acquisition, Y.L. All authors have read and agreed to the published version of
Funding: This research was funded by Key Research and Development Program of Tianjin, grant
the manuscript.
number 19YFHBQY00030.
Funding: This research was funded by Key Research and Development Program of Tianjin, grant
Institutional Review Board Statement: Not applicable.
number 19YFHBQY00030.
Informed Consent Statement: Not applicable.
Institutional Review Board Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
Informed Consent Statement: Not applicable.
corresponding author. The data are not publicly available due to privacy reasons.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author. The data are not publicly available due to privacy reasons.
Polymers 2022, 14, 3147 15 of 16

Conflicts of Interest: The authors declare no conflict of interest.

References
1. Li, J.; Zhao, X.; Yin, G.; Li, S.; Zhao, J.; Ouyang, B.H. The effect of accelerated water tree ageing on the properties of XLPE cable
insulation. IEEE Trans. Dielectr. Electr. Insul. 2011, 18, 1562–1569. [CrossRef]
2. Al-Arainy, A.A.; Ahaideb, A.A.; Qureshi, M.I.; Malik, N.H. Statistical evaluation of water tree lengths in XLPE cables at different
temperatures. IEEE Trans. Dielectr. Electr. Insul. 2004, 11, 995–1006. [CrossRef]
3. Nedjar, M. Effect of thermal aging on the electrical properties of cross-linked polyethylene. J. Appl. Polym. Sci. 2009, 111,
1985–1990. [CrossRef]
4. Kim, C.; Jiang, P.K.; Liu, F.; Hyon, S.; Ri, M.; Yu, Y.; Ho, M. Investigation on dielectric breakdown behavior of thermally aged
cross-linked polyethylene cable insulation. Polym. Test. 2019, 80, 106045. [CrossRef]
5. Kemari, Y.; Mekhaldi, A.; Teyssèdre, Y.; Teguar, Y. Correlations between structural changes and dielectric behavior of thermally
aged XLPE. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 1859–1866. [CrossRef]
6. Suo, C.; Qin, Y.; Li, Z. Effect of thermal aging on dc conductivity of nano-CB/XLPE insulating composites. In Proceedings of the
IEEE International Conference on High Voltage Engineering and Application, Athens, Greece, 10–13 September 2018.
7. Andjelkovic, D.; Rajakovic, N. Influence of accelerated aging on mechanical and structural properties of cross-linked polyethylene
(XLPE) insulation. Electr. Eng. 2001, 83, 83–87. [CrossRef]
8. Boukezzi, L.; Boubakeur, A.; Lallouani, M. Effect of artificial thermal aging on the crystallinity of XLPE insulation cables: X-ray
study. In Proceedings of the Conference on Electrical Insulation and Dielectric Phenomena, Vancouver, BC, Canada, 14–17
October 2007.
9. Mecheri, Y.; Boukezzi, L.; Boubakeur, A.; Lallouani, M. Dielectric and mechanical behavior of cross-linked polyethylene under
thermal aging. In Proceedings of the Conference on Electrical Insulation and Dielectric Phenomena, Victoria, BC, Canada, 15–18
October 2000.
10. Li, L.; Yong, J. A new method for on-line cable tanδ monitoring. In Proceedings of the International Conference on Harmonics
and Quality of Power, Ljubljana, Slovenia, 13–16 May 2018.
11. Densley, J. Review of international standards and practices for medium voltage power cable diagnostics. In Proceedings of the
International Conference on Solid Dielectrics, Winchester, UK, 8–13 July 2007.
12. Haikali, E.N.N.; Nyamupangedengu, C. Measured and simulated time-evolution PD characteristics of typical installation defects
in MV XLPE cable terminations. SAIEE Afr. Res. J. 2019, 110, 136–144. [CrossRef]
13. Chang, C.K.; Lai, C.S.; Wu, R.N. Decision tree rules for insulation condition assessment of pre-molded power cable joints with
artificial defects. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 1636–1644. [CrossRef]
14. He, B.; Zhou, Y.; Li, H.; Ye, T.; Fan, S.; Wang, X. Fault Identification of High-voltage Cable Sheath Grounding System Based on
Ground Current Analysis. In Proceedings of the IEEE 4th Conference on Energy Internet and Energy System Integration, Wuhan,
China, 30 October–1 November 2020.
15. Marzinotto, M.; Mazzanti, G. The feasibility of cable sheath fault detection by monitoring sheath-to-ground currents at the ends
of cross-bonding sections. IEEE Trans. Ind. Appl. 2015, 51, 5376–5384. [CrossRef]
16. Li, Y.; Peng, F.D.; Chen, X.L.; Cheng, Y.H.; Li, X. Study on sheath circulating current of cross-linked power cables. In Proceedings
of the IEEE International Conference on High Voltage Engineering and Application, Chongqing, China, 9–12 November 2008.
17. Iga, A.; Okamoto, T.; Hozumi, N.; Sasaki, S. The Application of On-Line Diagnoses for XLPE Cables by DC Component Method
to On-Site Installed Cables. IEEE Trans. Power Syst. 1991, 111, 1243–1249.
18. Zhu, G.; Zhou, K.; Lu, L.; Li, Y.; Xi, H.; Zeng, Q. Online Monitoring of Power Cables Tangent Delta Based on Low-Frequency
Signal Injection Method. IEEE Trans. Instrum. Meas. 2021, 70, 1–8. [CrossRef]
19. Rizzi, A.; Mascioli, F.M.F.; Baldini, F.; Mazzetti, C.; Bartnikas, R. Genetic optimization of a PD diagnostic system for cable
accessories. IEEE Trans. Power Deliv. 2009, 24, 1728–1739. [CrossRef]
20. Dragomiretskiy, K.; Zosso, D. Variational mode decomposition. IEEE Trans. Signal Processing 2013, 62, 531–544. [CrossRef]
21. Xiangke, Z.; Zhenming, L.; Yajing, W. Residual Current Detection Method Based on Variational Modal Decomposition and
Dynamic Fuzzy Neural Network. IEEE Access 2021, 9, 142925–142937.
22. Chen, W.; Wang, Z.Z.; Xie, H.B.; Yu, W.X. Characterization of surface EMG signal based on fuzzy entropy. IEEE Trans. Neural Syst.
Rehabil. Eng. 2007, 15, 266–272. [CrossRef] [PubMed]
23. Chaudhury, K.N.; Unser, M. Construction of Hilbert Transform Pairs of Wavelet Bases and Gabor-like Transforms. IEEE Trans.
Signal Processing 2009, 57, 3411–3425. [CrossRef]
24. Teng, W. Application of kernel principal component and Pearson correlation coefficient in prediction of mine pressure failure. In
Proceedings of the 2017 Chinese Automation Congress (CAC), Jinan, China, 20–22 October 2017.
25. Wu, F.; Jiang, X.; Ma, W. A Feature Extraction Method of Network Traffic for Time-Frequency Synchronization Applications. In
Proceedings of the 2017 International Conference on Computer Systems, Electronics and Control (ICCSEC), Dalian, China, 25–27
December 2017.
26. Qi, Y.; Xu, Z.; Zhou, C. The Research of Time-Frequency Feature Extraction of Loudspeaker Acoustic Signal by Pre Emphasis in
Time Domain. In Proceedings of the 2008 Second International Symposium on Intelligent Information Technology Application,
Shanghai, China, 20–22 December 2008.
Polymers 2022, 14, 3147 16 of 16

27. Alcaraz, R.; Rieta, J.J. A review on sample entropy applications for the non-invasive analysis of atrial fibrillation electrocardio-
grams. Biomed Signal Processing 2010, 5, 1–14. [CrossRef]
28. Pincus, S.M. Assessing serial irregularity and its implications for health. Ann. N. Y. Acad. Sci. 2001, 954, 245–267. [CrossRef]
[PubMed]