Vous êtes sur la page 1sur 9

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
1

Use of Interfacial Charge for Diagnosis and


Activation Energy Prediction of Oil-paper
Insulation used in Power Transformer
D. Mishra, Student Member, IEEE, S. Dutta, Student Member, IEEE, A. Baral, Senior Member, IEEE,
N. Haque, Member, IEEE, S. Chakravorti, Senior Member, IEEE

 model parameters. Among available R-C network-based


Abstract--Activation energy is popularly used for estimation of insulation models, Classical Debye Model (CDM) is perhaps the
remaining life of transformer insulation. It is defined as the most widely used circuit [3].
average rate of all reactions that happen with cellulose. Existing It is reported that measurement temperature significantly
literature shows that activation energy of oil-paper insulation can
be obtained from Polarization Depolarization Current (PDC) and affects the measured dielectric spectroscopy data [4-5]. PDC data
Return Voltage Measurement (RVM) data that are measured at recorded from the same unit at different time instants might be
a specific temperature. It is practically difficult to ensure the influenced by two distinctly different measurement temperatures.
same measurement temperature for both PDC and RVM data. Analysis of such measurement data separately lead to inaccurate
On the other hand, PDC data and its analysis get influenced by assessment of aging that might have happened between the two
de-trapping current. This de-trapping current is generated by measurements. Research findings show that CDM parameters,
ionic charge carriers that get freed from trap sites during PDC
measurement process. Formation of these trap sites are related to obtained using insulation response (measured at two different
physical, chemical reactions that happens at oil-paper interface. temperatures), are related by Activation Energy Ea [6-8]. This
The present paper proposes a methodology which uses de- implies that PDC data recorded at two different measurement
trapped charge, dislodged from deep and shallow traps, to assess temperatures can be successfully compared if the activation
insulation condition and for prediction of activation energy. energy of the system is known. Activation energy is defined as
Thus, eliminating the need of RVM data. The proposed method is the average rate of all the reaction (physical and chemical) that
tested using data collected from various real-life in-service
transformers. involvessolidinsulation. Satisfactory information regarding
remaining life of transformerinsulationcan also be obtained using
Index Terms—Space charge, Dielectric measurement, Power Ea [8-10]. It is reported by researchers that value of Ea generally
transformer, Dielectric materials, Conductivity, Moisture lies within a range of 85kJ/mol to 120kJ/mol [9]. Normally its
measurement value is considered equal to 111 kJ/mol for real-life transformers
[7], [9].Literature shows that Ea is not constant for a given unit
I. INTRODUCTION but depends on paper moisture (%pm)[8]. Its dependency on

D IELECTRIC spectroscopy based non-invasive insulation


diagnosis techniquesare popular among utility providers. As
far as insulation condition is concerned, available literatures [1-2]
moisture content indicates that assuming Ea to be equal to
111kJ/mol is not always accurate. Unfortunately, very few
literatures exist that tries to evaluate value of Ea using data
show that both Time Domain Spectroscopy (TDS) and Frequency obtained through non-invasive testing. One such technique is
Domain Spectroscopy (FDS) based methods provide same reported by two authors of the present paper [9]. In [8-9], Ea is
information. Measurement of FDS data is time consuming estimated using an iterative technique that requires PDC and
especially in the low frequency region.As a result, PDC analysis RVM data measured at a specific temperature. Both PDC and
based non-invasive diagnosis approach has gained popularity in RVM data measurement are time-consuming processes.
recent times [1-3]. Typical analysis procedure of PDC data Furthermore, it might be practically difficult to ensure same
involves identification and thereafter interpretation of insulation measurement temperature for both PDC and RVM [11]. The
above discussion suggests that it will be beneficial to the utilities
D. Mishra is with the Electrical Engineering Department, Indian Institute if Ea identification and insulation diagnosis can be done using a
of Technology (Indian School of Mines), Dhanbad, India (e-mail: single type of non-invasive measurement, preferably PDC.
deepak2010urs@gmail.com).
S. Dutta is with the Electrical Engineering Department, Indian Institute of Apart from measurement temperature, PDC data and hence, its
Technology (Indian School of Mines), Dhanbad, India (e-mail: interpretation also gets affected by de-trapped charge [12-13]. It is
dutta.saurabh2@gmail.com).
A. Baral is with the Electrical Engineering Department, Indian Institute of reported [3] that polarization current is the combination of dipole
Technology (Indian School of Mines), Dhanbad, India (e-mail: current (generated due to dipole orientation) and conduction
arijit@iitism.ac.in ). current. Similarly, depolarization current is expected to have
N. Haque is with the Electrical Engineering Department, Jadavpur
University, India (e-mail: nasirul07@gmail.com). current only due to the relaxation of dipoles. If the conduction
S. Chakravorti is currently the Director, National Institute of Technology, current is calculated accurately and subtracted from polarization
Calicut, India, on lien from Jadavpur University, Kolkata, India (e-mail: current then the resultant current should represent only dipole
s_chakrav@yahoo.com).

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
2

contribution. It is observed that dipole current contribution in both in shallow traps) require greater energy and hence takes longer
polarization and depolarization are not always equal. This time to get absorbed at the electrodes [12, 24]. The above
difference between thetwo current mainly arises due to de-trapped discussion suggests that different region of depolarization current
current contributed by freed ionic charge carriers (previously is influenced by charges that are de-trapped from different trap
trapped at the interface and solid insulation) at the time of depths. Available literature [2-3] shows that the initial portion of
measurement.The originof this trapped charge lies in charge the relaxation current is sensitive to condition of oil while current
transport process occurring in the insulation under electric at larger value of time is influenced by condition of solid
stresses. insulation. It must be rememberedhere, that any region in the
Available literature depicts that space charge trapping and insulation can contain both shallow and deep traps. It will be later
accumulation is significant in multi-dielectric media under shown in this work that the contribution of charges freed from
high fields due to significant charge injection and conductivity shallow traps often get compounded with the contribution from
mismatch between component dielectrics [14-15]. However, charge careers present in oil. Hence, the initial part of de-trapping
PDC measurements are performed at low electric fields. Under current was found to be sensitive to oil condition. The
such stresses, charge injection can be safely ignored. contribution of charges released from deep traps carry
Therefore, the ionic charge carriers present in mineral oil and information mainly related to paper. As the magnitude of Eais
pressboard plays an important role in the conduction process heavily influenced by cellulosic condition, therefore any
[16-17]. Availability of various polar compounds in oil relationship (if at all it exists) between different sections of de-
coupled with cellulose surface imperfections create ample trapping current and Ea is a matter of considerable interest. It will
scopes for ionic charge to get trapped at the interface even at facilitate ease in insulation diagnosis, Ea estimation and
low fields. Charge carriers are known to drift and get remaining life estimation using only PDC data. Therefore, the
accumulated near the electrodes during polarization [16]. present work investigates the influence offreed charge
Another process which results in ionic charge carrier trapping corresponding to different trap depths on insulation condition and
is flow electrification [18-19].Furthermore, it is understood Ea value. The finding reported here is based on analysis of data
that all these processes are influenced by cellulose surface collected from various real-life transformers (details are given in
conditions. It has been discussed in [20] that the resistive Table 1).Fig.1(a) and Fig. 1(b) show polarization and
contribution in PDC comes from alternate layers of oil and depolarization current recorded from some of the above-
pressboard, spacers and parallel oil channels. During mentioned transformers.
polarization current measurement, after a few hundred seconds
these ionic carriers, reduces the oil conductivity. As a result,
the conduction current starts to flow through spacers (solid
insulation) and parallel oil branches [21]. The ionic charge
carriers present in interface and solid insulation region will
inarguably polarize the molecules and dipoles in its
surrounding. This creates a screen of dipoles surrounding the
ion, producing a potential barrier or ion trap hindering the
motion of the ion [22-23].The trapped ion requires sufficient
energy to overcome this trap which it normally obtains
through thermal vibrations. The ions which are able to escape
or „de-trap‟ from the ion trap during polarization current
measurement contributes to the net conduction current. This
process of charge de-trapping may remain active during (a)
depolarization current measurement, particularly if the insulation
contains large number of trapping sites.Such possibility is really
high for service aged transformers, where generation of aging by
products and bond-breaking considerably modify cellulose
microstructure.In such cases, the ionic charge freed from the
trapping sites (located in different parts of oil-paper composite
insulation) are finally absorbedby the electrodes thus contributing
to the overall depolarization current [12].
It has been already discussed that degradation of cellulosic
insulation impacts the value of activation energy of the
transformer insulation. Hence, value of Ea and other performance
parameters (considered in the field of insulation diagnosis) should
be sensitive to the de-trapped charge. It should be mentioned here
that energy required, to set bound charge free, is not same for all (b)
trapped charges. As both shallow and deep traps are known to Fig. 1(a) Polarization current of few tested transformers (b) Depolarization
exist [24], charge bound in deep traps (compared to those residing current of few tested transformers

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
3

It is worth mentioning here that all the transformers It is reported in [12] that de-trapping current calculated
considered for the present work are in-service transformers. from (2) is decaying in nature. The explanation for this shape
PDC data used in the present analysis were recorded using can be given as follows: It is explained in [23] that a classical
KEITHLEY 6517B Electrometer based setup developed in two potential well model can be used to explain the process of
High Voltage Laboratory of Jadavpur University, Kolkata [5]. ion conduction in a typical solid-liquid insulation system.
According to the model, the ionic conduction takes place as
TABLE I
Details of Tested Transformer the trapped ions move from one site to the other. Assuming
Transformer Approximate that there exist two close-by trap sites, ts1 and ts2, separated
Power Rating by a distance 2aand a barrier energy of ΔHb. Sufficient energy
name Age ( in years)
Tf1 240 MVA/420 kV 29 required for an ion located at ts1, jump to ts2, under applied
Tf2 240 MVA/400 kV 22 electric field Ef, is (ΔHb-qaEf), where q is the charge of the
Tf3 200MVA/400 kV 25 ion. Likewise, the energy required for charge to jump from
Tf4 200MVA/420 kV 11 location ts2 to ts1 is (ΔHb+qaEf) [23]. If the external field is
Tf5 240 MVA/420 kV 23
absent, the probability of an ion to move from ts1 to ts2 can be
Tf6 200 MVA/420 kV 17
Tf7 200MVA/765kV 12
given by (3) [23].
Tf8 167MVA/420kV 19 ∆𝐻𝑏
Tf9 270MVA/235kV 13 𝑃 = 𝑓𝑣 𝑒𝑥𝑝 − (4)
𝑘𝑐 𝑇
Tf10 200 MVA/420 kV 31
Tf11 240 MVA/420 kV 26 Where, fv is the frequency of vibration (number of ion
Tf12 125 MVA/220 kV 29 jumps attempted per unit time), k c is the Boltzmann‟s
II. CHARGE TRAPPING AND DE-TRAPPING PHENOMENON IN constant (1.38×10-23) and T is the temperature. The relaxation
OIL-PAPER INSULATION AND ITS INFLUENCE ON TDS DATA time, τr of this process can be expressed in terms of P as given
by (4).
Free radical and acidic components which get generated 1
over time, speeds up the ionization process [25-28]. Due to 𝜏𝑟 = 2𝑃 (5)
conductivity difference between oil and paper these ionic From (4) and (5), it can be observed that the de-trapping of
charges accumulate at the interfacial region [12].During charge occurs in a way of exponentiallydecreasing nature with
depolarization current measurement, some part of accumulated respect to time. The time constant of this process indicates the
charges gets de-trapped and contributes in depolarization energy of the potential well, in which the charge is trapped. As
current [12-13]. In order to separate the de-trapped charge a result, in the present work, the de-trapping current is
from depolarization current, a method is reported based on assumed to have exponential decaying nature.
some assumptions [13]. The de-trapping current i(dtrapp) is
related to polarization anddepolarization current by 1(a) and For ease of computation, the de-trapping current is
1(b) approximated by a single exponential decay function in [12].In
Ipol= ipol(dipole)+icondn 1(a) [12] it is mentioned that de-trapping current should be
approximated by more than one exponential function for
Idepol= idepol(dipole)+ i(dtrapp) 1(b) detailed analysis. It was further observed that using more than
As dipole current contribution should be same in both three exponential functions did not lead to significant increase
polarization and depolarization current i.e. ipol(dipole)=idepol(dipole). in fitting accuracy. Hence, in the present work, the separated
The de-trapping current i(dtrapp) is related to Ipol and Idepol by (2) de-trapping current (obtained from (2)) is approximated by (6)
for thorough analysis.
Ipol - Idepol = icondn- i(dtrapp) (2)
3 𝑡
Conduction current, evaluated from PDC data at smaller value 𝑖(𝑑𝑡𝑟𝑎𝑝𝑝 ) = 𝑖=1 𝐴𝑖 × 𝑒𝑥𝑝 − ; 𝜏3 > 𝜏2 > 𝜏1 (6)
𝜏𝑖
of time, is reported to be time dependent [1-2]. The initial
portion of conduction current is influenced by high mobility of where𝐴1 , 𝐴2 and 𝐴3 represent the amplitude while 𝜏1 , 𝜏2 and
charge carriers or ions present in oil. After some time, this 𝜏3 represent the time constants of different exponential decay
functions.Furthermore, these time constants also indicate the
transient part dies down and the steady state conduction
trap depth associated with various traps where the charges are
current due to dc insulation resistance (R0) flows. Hence, icondn
bound.It should be mentioned here, that although expression
mentioned in (2) can be calculated using (3)
(6) resembles Debye Modeling, it refers to a fundamentally
icondn= Ipol(tfinal)+ Idepol(tfinal) (3) different physical process.
where, Ipol(tfinal) and Idepol(tfinal) are steady state or final It is worth mentioning here that in (6), τ3representlargest time
values of polarization and depolarization current respectively. constant and hence signifiesthe contribution of trap charge
Thus, if the values of Ipol, Idepol and icondn are known, the de- dislocated from deep traps. Similarly,τ1 and τ2represent lower
trapping current can be estimated using (2). Accurate time constants of de-trapped current corresponding to the
measurement of i(dtrapp) profile can be ensured as long as PDC contribution of charge de-trapped from shallow locations.In
measurement is continued until the insulation response starts the present work, the net de-trapped charge is obtained in two
to vary extremely slowly or becomes almost constant. steps: First i(dtrapp) is fitted with a single exponential function

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
4

i(dtrapp_fit) [12]. Thereafter, this fitted function idetrap_fit is understanding, the process described in [9] is briefly
integrated to obtain the total de-trapped charge.It should be elaborated below. The method detailed in [9] uses both PDC
mentioned here that comparative analysis of interfacial charge and RVM data recorded at a specific measurement
(obtained by simple integration of i(dtrapp_fit)) may be temperature T1. The method starts by evaluating the value of
misleading. As the de-trapping phenomenon occurs at interfacial Central Time Constant (CTC) at measurement temperature T1
region and solid insulation, the value of de-trapped current is from measured RVM data. It should be mentioned here that
expected to depend on the size and structure of concerned CTC measured at two different temperatures T1, T2 (T2>T1) are
insulation[13]. The surface area of oil-paper interface in different related by (8) [11,29].
transformer insulation may vary due to their different structure.
Therefore, it is difficult to compare the obtained de-trapped 𝐶𝑇𝐶 𝑇2 = 𝑒𝑥𝑝 𝑘1 × 𝑇2 − 𝑇1 × 𝐶𝑇𝐶 𝑇1 (8)
charge unless the influence of geometry is reduced. It is reported The value of constant ‘k1’ in (8) is generally considered as -
that the influence of geometry can be reduced by considering dc 0.0915 [9,29]. Next, the recorded PDC data is used to find the
insulation resistance R0 as the normalizing factor [13]. Hence, in RVM data at temperature T2 (T2>T1). This is done as follows:
the present work, the de-trapped charge obtained from different The depolarization current is used to identify the CDM
real-life power transformers are normalized by its respective parameters (𝑅𝑖𝑇1 , 𝐶𝑖𝑇1 ) at measurement temperature T1. CDM
R0. Furthermore, three types of normalized de-trapped charge parameters at temperature T1 and T2 are related to Ea as per (9)
(Qdenorm, QD, QS) are considered in the present paper. The and (10)
normalized charges Qdenorm, QD, QS are related to i(dtrapp_fit) by 𝐸𝑎 1 1
𝑅𝑖𝑇2 = 𝑅𝑖𝑇1 𝑒𝑥𝑝 − − ; 𝐶𝑖𝑇1 = 𝐶𝑖𝑇2 (9)
7(a), 7(b) and 7(c). 𝑘 𝑇1 𝑇2


𝑄𝑑𝑒𝑛𝑜𝑟𝑚 =
1
× 𝑖 𝑑𝑡 7(a) where, „k‟ is the Boltzmann constant (8.617385×10 -5eV/K).
𝑅0 𝑡=0 (𝑑𝑡𝑟𝑎𝑝𝑝 _𝑓𝑖𝑡 )
Assuming the value of Ea to be 111 kJ/mol, CDM parameters
1 ∞ 𝑡 𝑅𝑖𝑇2 , 𝐶𝑖𝑇2 at temperature T2 are computed using (9) and (10).
𝑄𝐷 = × 𝐴3 × 𝑒𝑥𝑝 − 𝑑𝑡 7(b)
𝑅0 𝑡=0 𝜏3 The identified parameters of CDM (containing n branches) are
then used to model the RVM spectrum of the insulation at

𝐴1 × 𝑒𝑥𝑝 −
𝑡
𝑑𝑡 + temperature T2. The relation between return voltage Urvm and
1
𝑡=0 𝜏1 voltage UC in ith capacitor of CDM is given by (10).
𝑄𝑆 = × 7(c)
𝑅0 ∞
𝐴2 × 𝑒𝑥𝑝 −
𝑡
𝑑𝑡 Urvm s n N n,i  s n 1 N n 1,i    N 0
𝑡=0 𝜏2  n 1 (10)
Uc s Dn 1  s n Dn   s n 1Dn 1    D0
As per 7(a), Qdenorm stands for total (normalized) de-
trapped charge. On the other hand, normalized charge QD The zero (z) and poles (p) of transfer function given by (10)
(corresponding to largest time constant) represents charges can be obtained by solving numerator and denominator
that are dislodged from deep traps. Similarly, the normalized polynomial. The peak value corresponding to a RV cycle with
charge QS corresponding to lower (lowest and medium) time tc charging time is evaluatedusing (11) through (14)[3,11].
constants represent contribution of shallow traps.It is reported
   t 
U r max t p   U 0 Fmax 1 t p  1  exp  T 2 c T 2  
that under normal condition, two types of traps exist in the
interfacial region, one is shallow trap and another is deep trap
[24,27]. Charges present in shallow traps require   R1 C1  
comparatively less amount of energy to change their location
  tc 
while those residing in deep traps need higher amount of  exp  T2 T2 
   U 0 Fmax n t p  (11)
excitation for their movement from one location to other [24].  2  R1 C1 
III. ESTIMATION OF ACTIVATION ENERGY FROM TDS DATA    t    tc 
 1  exp  T 2 c T 2    exp  
 
 2  Rn C n 
T 2 T 2
It is difficult to estimate the value of activation energy for any
  Rn C n  
given system using data obtained through any one non-
invasive technique. One method for finding Ea is reported by Fmax i (t p )  Fi ,1e p
t m1
   Fi ,n1e p
t mn 1
(12)
Koufakis et al [10] which requires measurement of insulation
resistance at multiple temperatures. Measurement temperature
   
is controlled by atmospheric conditions. After shutdown, a Fi , j  bi  ( p j  zi ,l )   p j  ( p j  pk )  (13)
power transformer is generally provided sufficient cooling  l   k j 
time before PDC, IR measurements are carried out. Utilities
tend to favor minimization of overall shutdown time for asset N n ,i
utilization. Hence, for in-service power transformer, which is bi  ; j, k  1,.., n  1; i, l  1,.., n (14)
Dn 1
not frequently shutdown for testing, it is not always practically
possible to measure IR at different temperatures.In [9] it is In (11), Urmax represents the maximum return voltage, U0 is
reported that activation energy can be estimated for a given the charging voltage applied to the system and tp is the time to
transformer insulation using PDC and RVM data. For easier reach the peak value. Once, the RVM spectrum is evaluated, it

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
5

is used to calculate the CTCCDM(T2) at measurement V. RELATION BETWEEN QS ,QD AND INSULATION PARAMETERS
temperature T2. It can be understood that CTCCDM(T2) and As discussed from Section II, QS and QD represents the
CTC(T2) must match as they represent the same information. charges released from shallow and deep traps, respectively.It
Hence, the difference of CTCCDM(T2) and CTC(T2) is used to has been already explained that activation energy of the oil-
adjust the initially assumed value of Ea and the entire process paper insulation maintains a well-defined relationship with
is repeated. The iterative process is stopped once the total normalized de-trapped charge 𝑄𝑑𝑒𝑛𝑜𝑟𝑚 .
difference becomes less than a pre-defined small number.
IV. VARIATION OF ACTIVATION ENERGY WITH DE-TRAPPED
CHARGE
With continued operation, conductive and polar compounds
get formed within the insulation in addition to surface
degradation of cellulose. These unwanted bi-products
participate in various physio-chemical reactions and results in
formation of trap locations due to broken bonds and fold
chains at interface [26-28]. This accelerates the aging process
and results in formation of even more by-products.
Availability of excess conductive aging by-products escalates
interaction between cellulose and generated polar compounds.
This in turn increases the activation energy of the system. The
above discussion suggests that the normalized de-trapped Fig.2. Variation of activation energy Ea with normalized de-trapped
charge should be influenced by Ea. In order to have a clear charge(R2=0.9211)
understanding about the influence of de-trapped charge on Ea, As the aging increases, more degradation of cellulosic
value of Qdenorm and Ea for all the tested transformers are surface takes place which ultimately results in generation of
presented in Table II. new traps and changes previously created shallow traps into
TABLE II deep traps. Therefore, analyzing the effect of deep and shallow
Normalized Charge and Ea for Different Transformers traps on activation energy and other insulation sensitive
Transformer name Qdenorm %pm Ea (eV) performance parameters is an interesting task. In the case of a
real-life in-service transformer, such behavior of deep and
Tf1 3.585e-14 1.80 1.36
Tf2 1.413e-14 1.60 1.30
shallow traps can be verified either by carrying out invasive
Tf3 3.039e-16 1.10 0.87 testing or by analysis of insulation responses that are measured
Tf4 1.850e-15 1.20 1.21 at regular interval. Utilities prefer minimization of overall
Tf5 2.124e-15 1.30 1.04 shutdown time for maintaining service continuity. This means
Tf6 9.130e-16 1.00 0.96 for an in-service power transformer; frequent testing is not
Tf7 5.785e-15 1.40 1.26 practically feasible. This is also true in the present case. It is
Tf8 1.873e-16 1.10 0.84 not possible to emulate every single process (that is
Tf9 1.012e-16 1.20 0.91 responsible for charge trapping in transformer) in laboratory
Tf10 2.433e-14 1.70 1.36 sample. However, some quantitative information regarding
Tf11 4.720e-15 1.50 1.26 behavior of traps with aging and activation energy can be
Tf12 2.021e-14 1.70 1.34 obtained by studying effect of flow electrification in scaled
It should be mentioned here that value of Ea given in Table II down laboratory sample.
is obtained using technique described in section III while,%pm Two structurally identical samples (S1 and S2) were
is obtained using IDAX 300. Thevariation of normalized de- constructed in the laboratory. Sample S1 was constructed
trapped charge with Ea is shown by scatter plot given in Fig. 2. using new Kraft paper and pressboard. While, sample S2 was
It can be observed from Fig. 2, that the normalized de-trapped constructed using old materials that were available in the
charge indeed influences the value of Ea. Equation (15) is laboratory. The steps involved in construction of the samples
found to best describe the relationship that exists between the are detailed in [5]. Before immersion in transformer oil for
data shown in Fig.2. In order to improve readability, the fitted impregnation, S2 was heated at 120°C for 592h. This was
curve is also included in Fig.2. done to emulate thermal aging in S2. Thereafter, both dry
𝑄𝑑𝑒𝑛𝑜𝑟𝑚 samples were immersed in equal volumes of old transformer
𝐸𝑎 = 0.8374 + 0.5623 × 2.5284 +𝑄𝑑𝑒𝑛𝑜𝑟𝑚
(15) oil. After proper impregnation, oil circulation was
implemented in both the samples. For each sample, oil-
In (15), Qdenorm represents result obtained using 7(a) scaled circulation (to facilitate charge separation at the interface) was
by 1015.Fig. 2 along with (15) show that Ea maintains a well- continued for 96h prior to PDC measurement. Thereafter, the
defined relationship with normalized de-trapped charge. recorded PDC data was analyzed to identify de-trapping
Hence, Ea can be easily estimated using normalized de-trapped current. Table III shows the magnitude of total normalized de-
charge without restoring to iterative process given in section trapped charge (Qdenorm) obtained in the case of sample S1 and
III. S2. In order to improve readability, magnitude of charge freed
from shallow (QS) and deep traps (QD) (evaluated using (7b)

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
6

and (7c)) are also included in Table III. The last column of the current with small time constant (compared to pressboard). It
table presents the activation energy (evaluated using [9]) is very much possible that this phenomenon, which is due to
associated with the samples.It can be observed from the data charge carriers present in mineral oil, might impact the initial
presented in Table III that compared to S1, S2 is associated part of depolarization current measured in the power
with high value of Ea as well as Qdenorm. Furthermore, in the transformers mentioned in Table 1. As this contribution in
case of S1, magnitude of QS and QD are of almost similar depolarization current is distinct from idepol(dipole), this will
magnitude. impact the initial part of de-trapping current and thereby the
Table III. De-trapped charge and activation energy of laboratory value of QS.Therefore, the parameter QS takes account for not
sample only charges de-trapped from shallow traps, but also charge
Sample Qdenorm QD QS Ea carriers present in oil. Keeping this in mind, the observed
S1 2.9244e-16 1.4935e-16 1.4309e-16 0.89 results in Fig.3(b) and (16) suggests that the increase in oil
S2 5.6305e-15 4.2901e-15 1.3404e-15 0.99 conductivity with QS can be attributed with the generation of
new charge carriers ( N c ), which in turn increases QS..
This denotes almost same concentration of deep and shallow
traps. On the other hand, in the case of S2, QD is found to be
significantly larger than QS. This proves that over time, as the
insulation ages and cellulose degrade, new shallow traps get
created and previous shallow traps get converted into deep
traps. This, in turn increases the contribution of QD and
Qdenorm.
It has been experimentally observed thatthe charge traps in
mineral oil are primarily shallow in nature [30], whereas
cellulosic insulation contains both shallow and deep traps
[31].The experimental investigations in [30, 31] were
performed at high electric fields, and the traps considered in
those works were electron traps. It would be therefore
interesting to see, if same trend persists in present case.In that
case, the parameter QS is less likely to be influenced by %pm
(a)
and paper conductivity, as they are more related to the
condition of cellulose. The value of activation energy also
strongly depends on the condition of cellulosic insulation;
hence it is expected that there should not be much correlation
between QS and activation energy. Fig.3(a) show that this is
indeed the case as no relation exists between the normalized
charge corresponding to lower time constant with %pm and
activation energy.
As shallow traps dominate charge de-trapping behavior in
mineral oil, it will be interesting to identify any relation
between QS and oil condition, if exists.Available literature
shows that the conductivity of mineral oil is highly affected by
charge trapping and de-trapping [30]. The conductivity of oil
 oil can be given through the following equation,
 oil  N c q (16) (b)
Fig. 3(a) Variation of %paper moisture and Activation energy with Qs, (b)
where, N c , q and  are charge carrier density, charge of Variation of oil conductivity(measured using IDAX 300) with Qs (R2=0.9612)

individual carrier and mobility of the charge, respectively. As QD is directly influenced by number of deep traps present
From the discussion so far, it can be inferred that the (and hence by cellulose surface), it is likely to affect
parameter N c q is proportional to normalized de-trapped parameters that indicates insulation condition (like %pm)
[15,25]. In order to investigate if this is indeed the case,
charge from shallow traps QS. The variation of QS with oil
variation of QD with %pm is plotted as scatter diagram (Fig.4).
conductivity is plotted in Fig. 3(b). The straight-line nature of
It can be observed from Fig. 4 that QD value can indeed be
the variation in oil conductivity resembles (16) well.In [16]
used to predict the value of %pm.The relationship that exists
detailed PDC analysis was performed on mineral oil, and the
between data illustrated in Fig. 4 is observed to be best
response was explained through ionic charge transport and
described by (17).In order to improve readability, the fitted
diffusion. It was observed in [16] that during polarization
curve is also included in Fig.4.
period the charge carriers drift and get accumulated near the
0.0975
electrodes. During the depolarization period, these charge %𝑝𝑚 = 1.3114 × 𝑄𝐷 (17)
layers return back to the bulk quickly, mainly due to charge In (17), QD represents result obtained using 7(b) scaled by
diffusion. This process ultimately leadsto a depolarization 1015.Majority of available non-invasive moisture

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
7

determination techniques primarily depend on branch the other hand, Fig.3(a) suggest that cellulose properties do
parameters of linear insulation models [5,32]. Branch not significantly influence the magnitude of QS.As Ea
parameter identification of popular insulation models (like represents the average rate of all reactions that involves the
MDM [32]) is a computationally intensive process. cellulose, it is expected that QD will also maintain a well-
Furthermore, it is a known fact that insulation model defined relation with Ea. Fig.5 shows the variation between QD
parameters for a given transformer is not-unique. and activation Energy Ea.Scatter plot shown in Fig.5 show that
there exists a specific relationship between QD and Ea.

Fig.4. Variation of %pm with normalized de-trapped charge QD(corresponding


to higher time constant)(R2=0.9398)
Fig. 5. Variation of Activation Energy with QD (charge corresponding to
Variation of branch parameters affect the accuracy of %pm largest time constant)(R2=0.9327)
value. Hence, paper moisture predicted using techniques
always involves a degree of doubt. Zaengl et al [33] reported a This nature of data can be explained by the fact that
method for estimating %pm using minimum value of tanδ vs deterioration of solid insulation leads to break down of
frequency profile. Though relatively simple to apply, %pm cellulose bonds which in turn serve as additional trap site. As
value estimated using this equation [33] is reported to contain a result, various physio-chemical reactions that happenwith
high error. Method reported in [35] is dependent on tan solid insulationalso increases. This in turn increases the value
(measured at 1Hz) and is also not tested on real-life of QD and hence activation energy increases cumulatively. The
equipment. On the other hand, QD for a given PDC data is variation of de-trapped charge with activation energy in Fig.5
always unique. This fact coupled with the high value of is similar to that shown in Fig.2 where variation of Ea is
correlation factor between QD and %pm ensures reliable and plotted with total de-trapped charge. Equation (18) is found to
accurate prediction of %pm. In order to establish the be best suited for fitting of data shown in Fig.5.
efffectiveness of using QD for %pm estimation, performance of 𝑄𝐷
(17) is compared with other available non-invasive techniques 𝐸𝑎 = 0.8274 + 0.5608 × (18)
0.9566+Q 𝐷
[5, 32-35]. Table IV shows the result of such comparison. It
can be observed from Table IV that compared to available In (18), QD represents result obtained using 7(b) scaled by
techniques, (17) offerbetter accuracy in determination of %pm. 1015. The above discussion suggests that charges dislodged
from deep trap can be used for identification of solid
It is clear from Table II that the general trend of %pm is insulation while, charge carriers de-trapped from shallow traps
increasing with Qdenorm. It is also observed from Fig.4 that QD are sensitive to oil properties. In order to decipher which
have similar trend with %pm. Hence, it can be said thatQDis relation (out of (15) and (18)) is best suited for estimating the
significantly influenced by properties of solid insulation. On value of Ea (obtained using [9]), a performance comparison is

Table IV: Performance comparison of (17) in estimating %pm


Transformer Measured %pm Zaengl et Baral et al Kumar et Baral et al Fofana et al Proposed
Name (Using IDAX) al [33] [5] al [34] [34] [35] method (17)
Tf1 1.80 1.17 1.52 1.66 1.61 1.26 1.8
Tf2 1.60 1.11 1.47 1.51 1.49 1.20 1.58
Tf3 1.10 0.79 0.98 1.03 1.02 0.92 1.0
Tf4 1.20 0.91 1.06 1.04 1.06 1.03 1.3
Tf5 1.30 1.03 1.14 1.18 1.10 1.09 1.31
Tf6 1.00 0.67 0.91 0.94 0.93 0.86 1.12
Tf7 1.40 1.08 1.12 1.23 1.21 1.12 1.44
Tf8 1.10 0.76 1.02 1.04 1.02 0.89 1.04
Tf9 1.20 0.83 1.08 1.12 1.09 0.98 1.2
Tf10 1.70 1.15 1.51 1.59 1.42 1.24 1.69
Tf11 1.50 1.13 1.22 1.39 1.34 1.12 1.43
f12 1.70 1.16 1.50 1.57 1.53 1.24 1.67

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
8

done between them. Table V shows the result of such and overall shutdown time of transformer can be significantly
investigation. In Table V, Ea computed using methodology [9] reduced as well as meaningful information regarding oil
is considered as accurate value. Table V shows that %error condition and moisture content in paper can also be obtained.
involved in estimation of Ea using (18) is either equal or less It should be mentioned here that present work does not
than that obtained using (15). This is expected as (15) is includes the effect of measurement temperature on the
affected by QS which is not sensitive to Ea (Fig. 3(a)). Hence, evaluation of de-trapped charge. Another issue is that the
it can be opined that (18) is best suited for estimating Ea.It is presence of oil-ducts has strong effect on interfacial charge
practicallyrelevantto monitor condition of transformers which accumulation. The distribution of oil-ducts is related to oil-
have been in service for significant time duration as these are pressboard ratio, or X-Y ratio of the transformer insulation.
more likely to suffer insulation degradation/failure. Therefore, future research work will be conducted to study
Table V. Performance comparison of (15) and (18) in estimating Ea impact of measurement temperature and oil-pressboard ratio
on de-trapped charge evaluation.
Transformer Ea % error in estimating Ea
name (Using [9]) using (15) using (18)
VII. REFERENCES
Tf1 1.3600 0 0
Tf2 1.3000 0.76 0.76 [1] W.S. Zaengl, “Dielectric Spectroscopy in Time and Frequency Domain
for HV Power Equipment, Part I: Theoretical Considerations,”IEEE
Tf3 0.8700 8.04 2.29 Electr. Insul. Mag., vol. 19, no. 5, pp. 5-19, 2003.
Tf4 1.2100 11.57 9.09 [2] S. Chakravorti, D. Dey and B. Chatterjee, Recent Trends in the
Tf5 1.0400 6.73 6.73 Condition Monitoring of Transformers ,London: Springer, 2013.
Tf6 0.9600 3.12 3.12 [3] T.K. Saha, P. PurkaitandF. Muller, “Deriving an Equivalent Circuit of
Tf7 1.2600 3.17 1.58 Transformers Insulation for Understanding the Dielectric Response
Measurements,” IEEE Trans. Pow.Del., vol. 20, no. 1,pp. 149-157, Jan.
Tf8 0.8400 2.38 2.38
2005.
Tf9 0.9100 6.59 1.09 [4] T.K. Saha and P.Purkait, “Investigations of Temperature Effects on the
Tf10 1.3600 0.73 0.73 Dielectric Response Measurements of Transformer Oil-Paper Insulation
Tf11 1.2600 2.38 2.38 System,” IEEE Trans. Pow. Del., vol. 23, no. 1, pp. 252-260, 2008.
Tf12 1.3400 0.74 0 [5] A. Baral and S. Chakravorti, “Condition Assessment of CellulosicPart in
Power Transformer Insulation using Transfer Function Zero of Modified
Hence, from utility point of view, knowledge about insulation Debye Model,” IEEE Trans. Dielectr. Electr. Insul.,vol. 21, no. 5,
condition of old units are more crucial compared to those pp.2028-2036, Oct. 2014.
[6] L.E. Lundgaard (Convenor), “Ageing of Cellulose in Mineral-oil
which are recently commissioned. As reported in the paper, Insulated Transformers,”Cigre Report 323 of CIGRE Task Force
insulation response of transformers with high operational age D1.01.10, 2007.
tends to have significant quantities of de-trapped charges. [7] S.M. Gubanski (chair), P. Boss, G. Csépes, V.D.Houhanessian, J.
These charges can be easily identified (from PDC data) and Filippini, P. Guuinic, U. Gäfvert, V. Karius, J. Lapworth, G. Urbani, P.
Werelius andW. Zaengl., “Dielectric Response Methods for Diagnostics
used for monitoring different crucial parameters like Ea, %pm, of Power Transformers,” Report of the TF D1.01.09.
𝜎𝑜𝑖𝑙 . [8] N. Abeywickrama, C. Ekanayake, Y.V. Serdyuk and S. M. Gubanski,
“Effects of the Insulation Quality on the Frequency Response of Power
Transformers”, Jour.Electr.Engg. Tech., vol. 1, no. 4, pp. 534-542,
VI. CONCLUSIONS 2006.
Determination of activation energy for power transformer [9] H.C. Verma, A. Baral, A.K. Pradhan and S. Chakravorti, “A Method to
insulation is a difficult task. It requires either dielectric Estimate Activation Energy of Power Transformer Insulation Using
Time Domain Spectroscopy Data,” IEEE Trans.Dielectr. Electr. Insul.,
response measurement at different temperatures (which is not vol. 24, no. 5, pp. 3245-3253, 2017.
practically feasible) or back-to-back PDC and RVM [10] E.I. Koufakis, C.D. Halevidis, A.D. Polykrati and P.T. Tsarabaris,
measurement of a transformer unit (which is time “Calculation of the Activation Energy of Oil-Paper Insulation in a
consuming).Non-invasive technique reported in [9] does Distribution Transformer”, IEEE Electr.Insul. Mag., vol. 28, no. 2, pp.
52-58, 2012.
provide a way to estimate Ea. However, its application [11] P.R.S. Jota, S.M. Islamand F.G. Jota, “Modeling the Polarization
requires implementation of computationally intensive time- Spectrum in Composite Oil/Paper Insulation Systems,” IEEE Trans.
consuming iterative process. In the present work, a Dielectr. Electr. Insul., vol. 6, no. 2, pp. 145-151, 1999.
methodology is outlined that is capable of providing the same [12] D. Mishra, N. Haque, A.Baral and S. Chakravorti, “Assessment of
Interfacial Charge Accumulation in Oil-Paper Interface in Transformer
result as that would have been obtained using [9] from a single Insulation from Polarization-Depolarization Current Measurements,”
PDC measurement without restoring to any computationally IEEE Trans. Dielectr. Electr. Insul., vol.24, no. 3, pp.1665-1673, 2017.
intensive time-consuming iterative process. [13] D. Mishra, N. Haque, A. Baral and S. Chakravorti, “Effect of Charge
Accumulated at Oil-Paper Interface on Parameters Considered for Power
The present study incorporates the effect of ionic charges Transformer Insulation Diagnosis,” IET Sci. Meas. Tech.,vol. 12, no. 3,
freed from various trap depths during PDC measurements. pp. 411-417, 2018.
Further, qualitative analysis revealed that charge which was [14] S. Dasand N. Gupta, “Effect of Ageing on Space Charge Distribution in
Homogeneous and Composite Dielectrics,”IEEE Trans. Dielectr. Electr.
freed quickly since the beginning of depolarization current Insul., vol. 22, no.1, pp. 541-547, 2015.
measurement is particularly sensitive to oil condition and does [15] F. Rogtiand M. Ferhat, “Effect of temperature on trap depth formation in
not bear correlation with activation energy. The charge multi-layer insulation: Low density polyethylene and fluorinated
released during later part of depolarization current ethylene propylene,”Appl. Phys. Letters, vol. 104, no. 3, pp. 031605,
2015.
measurement is more sensitive towards moisture content [16] Y. Zhou, M. Hao, G. Chen, G. Wilson and P. Jarman, “Quantitative
present in paper and activation energy of the overall study of electric conduction in mineral oil by time domain and frequency
insulation. Results presented in this work indicate that domain measurement,”IEEE Trans. Electr. Insul, vol. 22, no. 5,
computation required for Ea estimation (in comparison to [9]) pp.2601-2610, 2015.

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892739, IEEE
Transactions on Power Delivery
9

[17] R. Bartnikas, “Dielectric Losses in Solid-liquid Insulating Systems –


Part I,” IEEE Trans. Electr. Insul., vol. 5, pp. 113-121; also Part II, vol. Saurabh Dutta(S2017) received his M.Tech degree from
6, pp. 1421, Dec. 1971. Maulana Abul Kalam Azad University of Technology,
[18] T. Paillat, J. M. Cabaleiro, H. Romat and G.Touchard, “Flow West Bengal in 2015. Currently, he is working as a Senior
Electrification Process: the Physicochemical Corroding Model Research Fellow in the Electrical Engineering Department,
Revisited”. IEEE Trans. Dielectr. Electr. Insul., vol. 16, no. 2, pp.359- Indian Institute of Technology (Indian School of Mines),
363, 2009. Dhanbad. His field of interest includes condition
[19] M. Cabaleiro, T. Paillat, O. Moreau and G. Touchard, “Flow monitoring of high voltage equipment.
electrification of dielectric liquids in insulating channels: Limits to the
application of the classical wall current expression,”J. Electrostatics, Nasirul Haque (M2016) received his Master of
vol. 66, no. 1-2, pp. 79-83, 2008. Engineering degree with specialization in high voltage
[20] A. Küchler, High Voltage Engineering: Fundamentals-Technology- engineering from Jadavpur University in 2014. Currently,
Applications. Springer., 2017, pp. 479. he is working as a senior research fellow in High Tension
[21] M. Koch and T. Prevost, “Analysis of dielectric response measurements Laboratory, Electrical Engineering Department, Jadavpur
for condition assessment of oil-paper transformer insulation”, IEEE University. His current area of interest includes charge
Trans. Electr. Insul, vol. 19, no. 6, 2012. trapping and de-trapping in polymeric insulators and
[22] K.C. Kao, “Dielectric Relaxation in Solids,” Elsevier, 2004,chapter 7, condition monitoring of high voltage power equipment and transformers.
pp. 382-383.
[23] R. Bartnikas, “Performance characteristics of dielectrics in the presence Arijit Baral (M2010 - SM2016) obtained his Ph.D. degree
of space charge,”IEEE Trans. Electr. Insul, vol. 4, no. 5, pp.544-557, from Jadavpur University, Kolkata, India in 2015. He is
1997. presently working as an Assistant Professor in the
[24] G. Mazzanti, G.C.Montanari and J.M. Alison, “A space-charge based Department of Electrical Engineering, Indian Institute of
method for the estimation of apparent mobility and trap depth as markers Technology (Indian School of Mines), Dhanbad, India. His
for insulation degradation. Theoretical basis and experimental current field of interest includes condition monitoring of
validation,”IEEE Trans. Dielectr. Electr. Insul., vol. 10, no.2, pp. 187- high voltage power equipment.
197, 2003.
[25] N. Lelekakis, D.Martin and J. Wijaya, “Ageing rate of paper insulation Sivaji Chakravorti(M1990 – SM2000) obtained his
used in power transformers Part 1: Oil/paper systemwith low oxygen Ph.D. degree from Jadavpur University, Kolkata, India in
concentration,” IEEE Trans.Dielectr.Electr. Insul.,vol. 19, no. 6, pp. 1993. He served as the Professor in Electrical Engineering
1999-2008, 2012. Department of Jadavpur University since 2003. He worked
[26] A.M. Emsley and G.C. Stevens, “Review of chemical indicators of at the Technical University Munich as Humboldt Research
degradation of cellulosic electrical paper insulation in oil-filled Fellow in 1995-96, 1999 and 2007, respectively. He
transformers,”IEE Proc. Sci. Meas. Tech.,vol. 141, pp. 324-334, 1994. served as Development Engineer in Siemens AG in Berlin
[27] C. Tang, G. Chen, M. Fu and R.J. Liao, “Space Charge Behavior in in 1998. He worked as US-NSF guest scientist at the
Multi-layer Oil-paper Insulation under Different DC Voltages and Virginia Tech, USA, in 2003. He is a Fellow of Indian National Academy of
Temperatures,” IEEE Trans. Dielectr. Electr. Insul., vol. 17, no. 3, Engineering and Distinguished Lecturer of IEEE Power & Energy Society.
pp.775-784, 2010. His current fields of interest are numerical field computation, condition
[28] C. Tang, R.J. Liao, G. ChenandL.J. Yang, “Research on the feature monitoring of transformers, signal conditioning in high voltage systems,
extraction of DC space charge behavior of oil-paper insulation,” Sci. application of artificial intelligence in high voltage systems and life-long
China Tech. Sci., vol.54, no.5, pp 1315–1324, 2011. learning techniques. Currently, he is the Director of National Institute of
[29] S. Dutta, A. Baral, A.K. Pradhanand S. Chakravorti, “Effect of Technology, Calicut, India.
measurement temperature on power transformer insulation diagnosis
using frequency-domain spectroscopy”, IET Sci. Meas. Tech.,vol.11, no.
6, pp- 773-779, 2017.
[30] Y.F. Du, Y.Z. Lv, C.R. Li, M.T. Chen, J.Q. Zhou, X.X. Li, Y. Zhou
andY.X. Zhong, “Effect of electron shallow trap on breakdown
performance of transformer oil-based nanofluids,”J. Appl. Phys., vol.
110, no. 10, pp. 104104(1)-104104(5),2011.
[31] J. Hao, R. Zou, R. Liau, L. Yang and Q. Liau, “New Method for Shallow
and Deep Trap Distribution Analysis in Oil Impregnated Insulation
Paper Based on the Space Charge Detrapping,”Energies ,vol. 11, no.2,
pp. 271, 2018.
[32] A. Baraland S. Chakravorti, “Prediction of Moisture present in
Cellulosic Part of Power Transformer Insulation using Transfer Function
of Modified Debye Model,”IEEE Trans. Dielectr. Electr. Insul., vol. 21,
no. 3, pp. 1368-1375, 2014.
[33] W.S. Zaengl, “Applications of dielectric spectroscopy in time and
frequency domain for HV power equipment,”IEEEElectr. Insul. Mag.,
vol. 19, no. 6, pp. 9-22, Nov./Dec. 2003.
[34] A. Kumar, H.C. Verma, A. Baral, A.K. Pradhanand S. Chakravorti,
“Estimation of paper-moisture in transformer insulation employing
dielectric spectroscopy data,”IET Sci. Meas. Tech.,vol. 12, no. 4, pp.
536-541, 2018.
[35] Y. Hadjadj, F. Meghnefi, I. Fofana and H. Ezzaidi, “On the Feasibility
of Using Poles Computed from Frequency Domain Spectroscopy to
Assess Oil Impregnated Paper Insulation Conditions,”Energies, vol. 6,
no. 4, pp. 2204-2220, 2013.

Deepak Mishra(S2017) received his M.Tech degree from


Indian School of Mines, Dhanbad in 2015. Currently, he is
working as a Senior Research Fellow in the Electrical
Engineering Department, Indian Institute of Technology
(Indian School of Mines), Dhanbad. His field of interest
includes investigating the influence of trapped charge on
insulation diagnosis of high voltage equipment.

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Vous aimerez peut-être aussi