Measuring of the dielectric properties of mineral oil
L. V. BADICU1, P. V. NOTINGHER1, L. M. DUMITRAN1,
G. TANASESCU2 AND D. POPA3
1
University Politehnica of Bucharest, 313 Splaiul Independentei St., 060042,
Bucharest, ROMANIA, E-Mail: [email protected]
2
SIMTECH INTERNATIONAL Ltd., 111 Constantin Brancoveanu Blvd., Bucharest 4, ROMANIA
3
ICMET Craiova, 144 Calea Bucuresti, 200515 Craiova, ROMANIA.
Abstract: - Power transformers activity in optimal conditions largely depends on their insulation system
condition. Therefore it lays emphasis on dielectric properties measurement of their basic components (paper
and oil).
In this paper, one of the new methods for dielectric properties measurement of mineral oil, respectively
dielectric response method in frequency domain, is presented.
With the NOVOCONTROL spectrometer, experiments were made on two types of mineral oil (MOL
virgin oil and NYNAS virgin oil) with different water content and the variation curves of relative complex
permittivity components ( ε 'r and ε 'r' ) and of dielectric loss tgδ with frequency and temperature were plotted. It
can be seen that ε 'r , ε 'r' and tgδ increase at low frequencies and their values depend on temperature values and
temperature variation way.
Key-words: - transformers, insulation system, mineral oil, dielectric spectroscopy, relative permittivity, loss
factor.
1 Introduction
The basic components of the insulation systems
of power transformers are cellulosic parts and
mineral oil, the oil representing about 75 – 80 % of
the insulation system.
During transformers operation, a decrease of
dielectric characteristics of the paper and oil
(respectively, their degradation) is registered due to
the thermal stresses, electrical stresses, oxygen, water
etc. to which are submitted. Degradation products
based on cellulose are the result of three chemical
important reactions: pyrolisis (resulting small chains,
CO, CO2, H2O, and furans), oxidation (resulting CO,
CO2, H2O and acids) and hydrolysis (resulting
smaller chains).
Due to the oxygen action, different products
appear inside the oil (water, gases, sludges), which
worsen their dielectric properties and increase their
viscosity (some of these being deposited on the
winding and blocking heat transmission). Under heat
(which causes a chemical degradation) and oxygen
action, oil color changes, the electric strength and
interfacial tension decrease and the dielectric loss
and acidity index increase. Water contamination
leads to a sharp decrease of the dielectric properties
of the oil. Water can appear from external medium
and/or due to the degradation processes of the paper
and oil.
The evaluation of oil condition can be done using
several methods, like water content determination,
dissolved gases analyses, interfacial tension
measuring, the oxygen method, the acidity index
method, the electric strength method, the loss factor
method etc. The new methods used are based on
relative permittivity measurement, loss factor
measurement,
absorption/resorption
currents
measurement, polarization index measurement,
conductivity factor measurement, return voltage
method [1, 2] etc. In previous paper [3] some
experimental results concerning the variations of
electrical properties (resistivity, permittivity, loss
factor, conductivity coefficient, polarization index)
of the power transformer insulation paper
(pressboard) with water content and temperature
were presented. In this work the first results obtained
by dielectric spectroscopy concerning temperature
influence and its way of variation on electric
properties of mineral oil are presented.
2 Dielectric Response Method
Dielectric spectroscopy is based on interactions
between applied electric field and electric dipoles of
the material under test, frequency values being
between 10-6 and 1010 Hz [4].
If an electric field of magnitude E0 (t ) = E0 ⋅ 1(t )
(Fig.1) the response of the dielectric at t ≥ t0 is
polarization P(t):
P (t ) = ε 0 χ (t )1(t )E 0
(1)
where χ(t) is the dielectric susceptivity (dielectric
response function), ε0 – the permittivity of vacuum
and 1(t) – the unit step [5,6].
Considering an ideal step for an electric field, the
adequate polarization for the very fast polarization
processes (electronic, ionic etc.) takes the value
P(t ) = P(t0 ) = P∞ . In case of big values for t, the
polarization finally becomes static, respectively
P(t → ∞) = Ps (Fig.1).
Fig.1 Polarization of a dielectric exposed to a step electric
field of magnitude E0 la t = t0.
For a time dependent excitation E(t), the time
dependent polarization P(t) is done (using the
Duhamel’s integral) by the equation:
t
P(t) = ε0χ∞E(t) + ε0 ∫ f (t − τ)E(τ)dτ,
(3)
where C0 is the geometric capacitance of test object,
σt is the dc conductivity and δ(t ) is the delta function
(which characterizes the voltage step at t = t0).
(5)
Fig.2 Time variation of absorption ia(t)
and resorption ir(t) currents.
Assuming that the polarization period Tc is
sufficiently long, so that f(t+Tc) ≈ 0, the dielectric
response function f(t) is proportional to the resorption
current:
f (t ) = −
where f(t) is the dielectric response function (a
monotonous and increasing function).
If a dc voltage U0(t) = U0δ(t) is applied to the
condenser coating which contains the test object, a
current ia(t) through the test object (Fig.2) appears:
0, if t 0 > t > TC
δ(t ) = 
.
1, if t 0 ≤ t ≤ TC
ir (t ) = −C 0 U 0 [ f (t ) − f (t + Tc )] .
(2)
−∞
σ
ia (t ) = C 0U 0 [ t + ε ∞ δ(t ) + f (t )] ,
ε0
test object and is independent of any polarization
process, the last one represents all the active
polarization processes during the voltage application
and the middle part with the delta function cannot be
recorded in practice due to the large dynamic range
of current amplitudes inherent to the very fast
polarization processes.
If the test object is short-circuited at t = tc, the
resorption current ir(t) can be measured:
i r (t )
.
C 0U 0
(6)
In case of applying an harmonic electric field
E (t ) = E 2 sin ωt and considering that all the
polarization processes are instantaneous, and χ (ω)
(the Fourier Transform of the dielectric response
function f(t)), the complex susceptivity and tgδ
dielectric loss factor result:
∞
F(ω) = χ(ω) = χ'(ω) − iχ"(ω) = ∫ f (t)exp(−iωt)dt ,
(7)
0
ε(ω) = 1 + χ' (ω) − iχ" (ω) = ε 'r − jε"r ,
(8)
(4)
The absorption current contains three terms: the
first is related to the intrinsic conductivity σ t of the
ε "r +
tgδ =
σt
ε 0ω
ε 'r
.
(9)
The equation (7) represents the link between time
and frequency domains. Thus, it is obvious that the
complex susceptivity χ (ω) can be converted to the
effective value 1 V and frequency between 1 mHz
and 10 kHz.
dielectric response function f(t) and vice versa[7,8].
3 Experiments
For experimental study, two types of
transformers oil were used: MOL virgin oil and
NYNAS virgin oil.
Fig.4 Novocontrol spectrometer: 1 – PC, 2 – Control
System, 3 – Modular Measurement System,
4 – Measurement Cell, 5 – Temperature Control System.
3.1 Water content measurement
Water content determination was made by KarlFischer method, using the experimental set-up from
Fig.3.
Fig.5 Cell for dielectric properties of liquids measurement:
1 – inner electrod; 2 – ground connecter; 3 – guard
electrod; 4 – outer electrod; 5 – sample liquid; 6 – teflon;
7 – hole; 8 – metal container; 9 – hole with spring;
10 – teflon isolating rings; 11 – teflon cap.
Fig.3 Karl-Fischer Coulometer.
Table 1 Water content values for oils [ppm].
Oil
NYNAS
MOL
1
15.9
20.8
2
14.3
21.9
Measurement
3
4
5
14.7 14.2 15.3
20.6 20.4 21.2
Medium
14.9
20.9
The results of the measurement (Table 1)
highlighted that the water content is 30 % higher in
MOL oil than NYNAS oil.
3.2 Relative permittivity and loss factor
measurement
To measure the relative complex permittivity ε r
and loss factor tgδ, a NOVOCONTROL
spectrometer (Fig.4) fitted with a liquids cell (Fig.5)
was used.
The three electrode cylindrical sample cell
avoids the errors related to the thermal expansion of
the measured liquid, protects against sample leakage
and prevents evaporation. It also increases the
accuracy of the measurement by decreasing the
influence of fringing fields. The voltage applied had
To check the accuracy and reproducibility of the
experimental results, each sample was measured
three times (the measurements M1…M3 for MOL and
respectively N1…N3 for NYNAS) in the same
condition.
The measurements were carried out at T = 30 ºC
and the obtained values for real component of
relative permittivity for three values of frequency are
presented in Table 2.
Table 2 Real component of relative permittivity ε 'r .
f
1 mHz
10 Hz
10 kHz
M1
4.21
2.36
2.34
M2
4.17
2.35
2.35
M3
4.23
2.36
2.35
N1
3.09
2.43
2.43
N2
2.97
2.43
2.43
N3
2.97
2.43
2.43
It results that, for both samples, values of ε 'r
increase when frequency decreases. This is due, in a
great measure, to the interfacial polarization increase.
3.3 Temperature influence
Both components of complex relative
permittivity ( ε 'r and ε"r ), as well as loss factor tgδ
values are strongly influenced by test sample
of dielectric losses. The increase of dielectric losses
is highlighted by the increase of imaginary
component of relative permittivity values ε"r and of
loss factor tgδ, both at high frequencies as well as
low frequencies (Figs.7, 8, curves 1, 1’, 2, 2’).
4
10
2'
2
10
1'
tgδ
temperature (Fig.6 – 8, 12 – 14) and the temperature
variation way (Fig.9 – 11).
The
temperature
increase
allowed
a
pronouncedly orientation of the electric dipoles and,
so, an increase of polarization and ε 'r too (Fig.6).
This phenomenon is more intense in case of low
frequencies (1 mHz). When an electric field is
applied to a dielectric, heat build-up occurs inside
due to Joule-Lenz effect, because the material
conductivity is not zero. In case of time variable
electric field, losses by dielectric hysteresis are
produced (electric viscosity), (electric polarization
does not vary in phase with the electric field).
0
10
-2
10
1
2
10
-4
1x10
8
2'
-6
10
εr'
30
40
50
60
o
70
80
90
T [ C]
6
1'
Fig.8 Variation of the loss factor tgδ with temperature T
for NYNAS (1 – f = 10 kHz, 1’ – f = 1 mHz), MOL
(2 – f = 10 kHz, 2’ – f = 1 mHz), when temperature T
increases from 30 ºC to 90 ºC.
4
2
2
1
20
0
30
40
50
60
70
80
90
o
T [ C]
10
4
10
2
10
0
15
2
εr'
Fig.6 Variation of the real permittivity component ε 'r with
temperature T for NYNAS (1 – f = 10 kHz, 1’– f = 1 mHz),
MOL (2 – f = 10 kHz, 2’ – f = 1 mHz), when temperature
T increases from 30 ºC to 90 ºC.
10
5
εr"
1
0
2'
10
-2
1x10
-4
10
-6
30
40
50
60
50
60
70
80
90
T [ C]
Fig.9 Variation of the real permittivity component ε 'r with
temperature T: 1 – T increases from 30 ºC to 90 ºC,
2 – T decreases from 90 ºC to 30 ºC (f = 1 mHz).
1
30
40
o
1'
2
70
80
90
o
T [ C]
Fig.7 Variation of the imaginary permittivity component
ε"r with temperature T for NYNAS (1 – f = 10 kHz,
1’ – f = 1 mHz), MOL (2 – f = 10 kHz, 2’ – f = 1 mHz),
when temperature T increases from 30 ºC to 90 ºC.
With temperature increase, the electric dipoles
oscillation and the mobility of charge carriers also
increase, respectively the liquid conductivity
increases, both phenomena leading to the increasing
3.4 The influention of the temperature
variation way
In figures 9 – 11, the influence of the temperature
variation way on the relative permittivity components
ε 'r and ε"r and loss factor tgδ for frequency 1 mHz is
presented. In this case, the temperature was modified
by increasing, respectively decreasing between 30 ºC
to 90 ºC. It noted that, the values obtained for all
three quantities at the temperature decrease are
greater than those obtained at the temperature
increase.
1400
2
1200
εr"
1000
800
600
1
400
200
0
30
40
50
60
70
o
80
90
In figures 12 – 14, variations of ε 'r , ε"r and tgδ
with frequency and temperature for MOL virgin oil
are presented. It can be observed that, because of
temperature increase, important increases in
permittivity and dielectric loss were produced due to
the reduction of liquid viscosity, of increase electric
dipoles mobility and of ionic conductivity increase.
The relative permittivity components and the
loss factor take higher values for MOL than NYNAS
(Figs.15 – 17). This situation is due to the water
content, which is higher for MOL oil.
T [ C]
4
10
Fig.10 Variation of the imaginary permittivity component
ε"r with temperature T: 1 – T increases from 30 ºC to
90 ºC, 2 – T decreases from 90 ºC to 30 ºC (f = 1 mHz).
3
10
o
30 C
o
o
o
70 C
50 C
90 C
2
εr"
10
1
10
140
0
10
120
-1
10
2
-2
10
tgδ
100
-3
10
80
-4
10
-3
60
10
-2
10
-1
10
0
1
10
10
2
3
10
10
10
4
f [Hz]
40
20
1
0
30
40
50
60
70
o
80
90
Fig.13 Variation of the imaginary permittivity component
ε"r with frequency f and temperature T for MOL oil when
temperature T increases from 30 ºC to 90 ºC.
T [ C]
3
10
Fig.11 Variation of the loss factor tgδ with temperature T:
1 – T increases from 30 ºC to 90 ºC, 2 – T decreases from
90 ºC to 30 ºC (f = 1 mHz).
2
10
o
tgδ
o
30 C
1
10
o
50 C
o
70 C
90 C
0
10
10
-1
10
εr'
8
-2
o
10
o
10
90 C
6
-3
70 C
o
50 C
4
-4
10
o
10
30 C
10
-2
10
-1
10
-2
-1
10
10
0
1
10
2
10
3
10
10
4
f [Hz]
2
0
-3
10
-3
0
10
1
10
2
10
f [Hz]
3
10
4
10
Fig.12 Variation of the real permittivity component ε 'r
with frequency f and temperature T for MOL oil when
temperature T increases from 30 ºC to 90 ºC.
Fig.14 Variation of the loss factor tgδ with frequency f and
temperature T for MOL oil when temperature T increases
from 30 ºC to 90 ºC.
4
Conclusions
After experimentes carried out on the two types
of oil (MOL oil and NYNAS oil), it can be noted that
their representative dielectric properties, respectively
complex permittivity (characterized by real and
imaginary components ε 'r and ε"r ) and loss factor are
strongly influenced by samples temperature and by
the voltage supply frequency.
10
εr'
8
6
Another parameter which highly influenced the
experimental results was the temperature variation
way, the values of quantities ε 'r , ε"r and tgδ being
higher when the temperature decreases from 90 ºC to
30 ºC. All the three quantities take higher values for
frequencies near 10-3 Hz. The water content of the
two types of oil is too high than norm imposed limit
of operation and they cannot be used into power
transformers without a previous drying treatment.
2
4
Acknowledgements
This work was supported in part by the
Romanian Ministry of Education, Research and
Youth, PNCDI II Program (Project MIDMIT 22080/2008).
1
2
0
-3
10
10
-2
-1
10
10
0
10
1
10
2
3
10
f [Hz]
4
10
Fig.15 Variation of the real permittivity component ε 'r
with frequency f for: 1 – NYNAS oil, 2 – MOL oil.
4
10
3
10
2
2
1
1
εr"
10
10
0
10
10
-1
10
-2
10
-3
10
-4
-3
10
10
-2
10
-1
0
10
1
10
10
2
3
10
4
10
f [Hz]
Fig.16 Variation of the imaginary permittivity component
ε"r with frequency f for: 1 – NYNAS oil, 2 – MOL oil.
3
10
2
tgδ
10
1
10
0
10
2
1
-1
10
-2
10
-3
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
3
10
4
10
f [Hz]
Fig.17 Variation of the loss factor tgδ with frequency f for:
1 – NYNAS oil, 2 – MOL oil.
References
[1] X. Zhang, E. Gockenbach, “Asset-Management
of Transformers Based on Condition monitoring
and Standard Diagnosis”, IEEE El.Insul. Mag.,
vol 24, 4, pp. 26 – 40 2008.
[2] M. Wang, A.J. Vandermaar, K.D. Srivastava
“Review of Condition Assessment of Power
Transformers in Service” IEEE El.Insul. Mag.,
vol. 24, 18, pp. 12 – 23, 2002.
[3] L.V. Badicu, P.V. Notingher, L. M. Dumitran, G.
Tanasescu, D. Popa “Use of Dielectric
Spectroscopy to Estimate the Condition of
Cellulose-Based Insulation”, JEEE, vol 2, 1, pp.
7 – 12, 2009.
[4] P. Hedvig “Dielectric Spectroscopy of
Polymers”, Publishing House Kiado, Budapest,
1977.
[5] W.S. Zaengl, “Dielectric Spectroscopy in Time
and Frequency Domain for HV Power
Equipment, Part I: Theoretical Considerations”,
IEEE Electrical Insulation Magazine, vol. 19, 5,
pp. 5 – 19, 2003.
[6] A.K. Joncher, The Universal Dielectric
Response: Part I, IEEE Electrical Insulation
Magazine, vol 6, 2, pp. 16-21, 1990.
[7] U. Gafvert, “Condition Assessment of Insulation
Systems: Analysis of Dielectric Response
Methods”, Proc. of Nordic Insulation
Symposium NORD IS 96, Trondheim, Norway,
pp. 1-19, 1996.
[8] D. Linhjell, L. Lundgaard, U. Gafvert,
“Dielectric
Response
of
Mineral
Oil
Impregnated Cellulose and the Impact of
Aging”, Trans. Dielectr. Electr. Insul., vol. 14,
1, pp. 156-169, 2007.