EFFECTS OF TEMPERATURE ON TIME-DOMAIN DIELECTRIC

DIAGNOSTICS OF TRANSFORMERS

T. K. Saha, Senior Member IEEE and P. Purkait, Member IEEE
School oI InIormation Technology and Electrical engineering
University oI Queensland, Brisbane, QLD 4072, Australia


Abstract

Time-domain dielectric testing techniques, namely the Return Voltage Measurement (RVM) and Polarisation
and Depolarisation Current (PDC) measurement are being used by the power utilities Ior assessment oI the
condition oI transIormers oil-paper insulation. However, it has been Iound that the results oI these tests are
highly inIluenced by operating temperature. It requires adequate experience and extreme care to interpret the
RVM and PDC results in the presence oI temperature variations and thermal instability. This paper reports a
detailed study on the eIIects oI temperature on RVM and PDC measurements on a transIormer under
controlled laboratory condition.


1. INTRODUCTION

Return Voltage Measurement (RVM) |1-3| and
Polarisation and Depolarisation Current (PDC) |4-6|
measurement techniques are believed to provide
indication oI the general ageing status and moisture
content oI the oil-paper insulation oI transIormer.
However, the results oI these tests are severely
inIluenced by several environmental Iactors,
predominantly the temperature |7-11|. This temperature
eIIect is more prominent in open substations where the
external environmental conditions are hardly predictable
and controllable.

For both RVM and PDC, the transIormer needs to be
disconnected Irom supply. BeIore starting the test it is
essential to allow suIIicient time Ior the transIormer to
cool down to ambient temperature Irom its initial higher
temperature oI operation. Any tests perIormed during
this temperature transition will corrupt the test results.
Another source oI temperature variation during RVM
and PDC tests is the variation oI ambient temperature
during the day.

In order to interpret the results more accurately in such
cases, it is essential to understand the variations oI the
RVM and PDC measurement results with temperature.
This paper reports laboratory test results oI RVM and
PDC measurement perIormed on a transIormer with
controlled variations oI temperature. A temperature
control cabinet with 1
0
C accuracy was used to vary the
temperature oI the transIormer at discrete steps over a
pre-deIined range. Each time the temperature was set to
a diIIerent value, RVM and PDC tests were perIormed
aIter allowing suIIicient time (around 10 days) Ior the
oil-paper insulation inside the transIormer tank to attain
stable temperature and moisture equilibrium. The
transIormer tank was completely sealed Irom outside
ambient thereby ensuring that the test results are solely
aIIected by the variations oI temperature.

2. DIELCTRIC RESPONSE MEASUREMENTS

2.1 PDC and RVM

For dielectric response (RV and PDC) measurement, a
DC step voltage U(t) with the Iollowing characteristics,
is applied to the initially relaxed insulation system:

0 0 _ t
U(t) U
0
0 _ t _ t
1
(1)
0 t _ t
1

During the initial charging period, the step voltage is
applied Ior time 0 _ t _ t
1
. In this time the charging
current (polarisation current) given by (2) will Ilow.

(

+ = ) ( ) (
0
0 0
t f U C t i
r
pol
c
o
(2)

where C
0
is the geometric capacitance (measured
capacitance at or near power Irequency, divided by the
relative permittivity, c
r
, oI the composite insulation at
power Irequency), c
0
is the vacuum permittivity, and
I(t) is the dielectric response Iunction oI the composite
insulation. The response Iunction f(t) describes the
Iundamental memory property oI the dielectric system
and can provide signiIicant inIormation about the
insulating material |12|. o
r
Is the average conductivity
oI the composite insulation system.

The insulation is then grounded (short circuited) Ior a
subsequent time period t
1
_ t _ t
2
; the magnitude oI the
depolarisation current is given by (3)

| | ) ( ) ( ) (
1 0 0
t t f t f U C t i
depol
+ = (3)

At tt
2
ground (short circuit) is removed Irom the
insulation and a voltmeter is connected across it.
Depending on how long the test object is grounded, t
2
,
some oI the previously polarised molecules get totally
relaxed, but some are not. The polarisation processes
which were not totally relaxed during the grounding
period will relax and give rise to a recovery voltage
across the electrodes oI the insulation. Figure 1 shows
the nature oI the polarization, depolarization current and
the recovery voltage. The test object is charged Irom
0_t_t
1
, grounded Irom t
1
_t_t
2
and Ior t~t
2
the recovery
voltage is measured during open circuit conditions.



Figure 1. Dielectric Response Measurements

2.2 Estimation of Conductivity

From the measurements oI polarisation and
depolarisation currents, it is possible to estimate the DC
conductivity o
r
, oI the test object |12-13|. II the test
object is charged Ior suIIiciently long time so that f(t
t
c
) ~ 0, (2) and (3) can be combined to express the DC
conductivity oI the composite system as:
( ) ( ) ( ) t i t i
U C
depol
ol
p r
~
0 0
0
c
o (4)
The average conductivity Ior a given insulation system
thus, is Iound to be dependent upon the diIIerence
between the polarisation and depolarisation current
values. This composite conductivity, in practice, is the
convolution oI the conductivities oI the oil and the paper
that make up the insulation structure. For estimation oI
the oil and paper conductivities separately it is required
to consider the insulation system to be composed oI a
simple series arrangement oI oil duct and paper
insulation |13, 14| as shown in Figure 2. Each material
is characterised by its conductivity and permittivity
along with the composite dielectric response Iunction
f(t).






Figure 2. Series arrangement structure oI oil and paper

X is the relative amount oI paper in the composite
system. The range oI X is typically 20 to 50 |14| Ior
a transIormer. The value oI X can be calculated more
precisely only when the exact structure oI the insulation
system and all its design parameters are available.

According to |13,14| the eIIective conductivity may be
written in terms oI paper and oil conductivities (o
paper

and o
oil
respectively ) as:
X X
oil paper
oil paper
r
. ) 1 .(
.
o o
o o
o
+
= (5)
The eIIective permittivity c
r
can also similarly be
estimated as:
X X
d p
d p
r
. ) 1 .(
.
c c
c c
c
+
= (6)

where c
p
is the relative permittivity oI paper and c
d
is the
relative permittivity oI oil duct.

Once the values oI the eIIective permittivity c
r
and
hence C
0
is estimated, eIIective conductivity o
r
can be
determined using (4) Irom the diIIerence between the
polarisation and depolarisation currents. The initial
polarisation current (aIter the Iirst transient that is
normally not recorded) can be written as:
( )
d
r oil
pol
U C i
c
c
c
o
. 0
0
0 0
= + (7)
such that
Paper/Pressboard



) 0 ( .
. .
.
0 0
0
+ =
pol
r
d
oil
i
U C c
c c
o (8)

On the other hand, the long-time polarisation current
(steady DC value i
dc
) can be related to the paper
conductivity as:
0
0 0
.
c
o
r
dc
U C i =
(9)
II o
oil
>> o
paper
, then Irom (5) we get
X
paper
r
o
o = (10)
Combining (9) and (10) we get,

dc paper
i
U C
X
.
.
.
0 0
0
c
o = (11)

3. EXPERIMENTAL SET-UP

The insulation system oI the transIormer model under
study consists oI oil and cellulose, as it is in a real power
transIormer. The structure oI the insulation system, the
ratio oI oil/paper, etc. are similar as in a real
transIormer. A temperature sensor oI type 100 Ohm Pt
385 was inserted into the tank to measure the actual
temperature inside the transIormer tank.
The amount oI solid insulation in the model is
approximately 1445g and the oil is 8400g; the ratio oI
oil/solid insulation is 5.8. The ratio oI oil/cellulose
material is about 10/1 to 6/1 in a real transIormer. The
whole model was kept inside a temperature control
cabinet. The temperature oI the control cabinet (and
hence the model transIormer) was set at discrete values
oI 25
0
C (ambient), 30
0
C, 40
0
C, 45
0
C, 65
0
C and 75
0
C.
The temperature sensor placed inside the transIormer
tank was used to make sure that the temperature inside
the tank is equal to that oI the control cabinet. AIter the
temperature inside the transIormer tank was Iound to
reach the temperature set in the control cabinet, it was
allowed to remain like that Ior 10 days. This ensured
that the oil and paper could achieve a new state oI
moisture equilibrium at the elevated temperature. RVM
and PDC tests were perIormed at all the diIIerent
temperatures. The PDC and RV measuring equipment
|2| developed at the School oI ITEE, University oI
Queensland was used Ior all the measurements.






4. ANALYSIS OF RESULTS

4.3 RVM Results

Figure 3 shows the RV spectra obtained at diIIerent
temperatures. It can be seen that the temperature
inIluence cause signiIicant displacements oI the RV
spectra peaks. At high temperatures the spectrum peak
shiIt to smaller times. The variation oI the dominant
time constants with the corresponding temperature is
summarised in Table 1.

Figure 3. RV spectra at DiIIerent Temperatures
Table 1. Dominant time constant vs. temperature

Temperature (
o
C) Dominant Time Constant (sec)
25 82
30 47
40 22
45 13
65 7
75 2

It was reported by Kozlovskis et al |7| that the RV
spectrum Iollows an exponential response depending
temperature and the movement in shape can be
described by the Iollowing equation:

) (
.
1 2
T a
T at d T at d
e t t
A
=
(12)

where t
d at T1
is the dominant time constant at a reIerence
measurement temperature oI T
1
,
t
d at T2
is the dominant time constant at some diIIerent
measurement temperature oI T
2
,
AT is the temperature diIIerence (AT T
2
T
1
)
The constant a is a parameter related to the polarisation
process inside the insulation material.

Figure 5 is the plot between dominant time constant and
the corresponding temperatures. The exponential nature
predicted by (12) is clearly visible in the plot oI Figure
4.

Figure 4. Dominant Time Constant vs. Temperature

4.4 PDC Results

Figure 5 and Figure 6 show the polarisation and
depolarisation currents respectively obtained at diIIerent
temperatures. In each case the transIormer was charged
(polarised) with 500 volts Ior 10,000 seconds and then
discharged (depolarised) Ior 10,000 seconds. It can be
seen that variation in temperature causes signiIicant
displacements oI both the polarisation and
depolarisation currents. It can be seen Irom Figures 5
and 6 that the magnitude oI the polarisation and
depolarisation currents tends to shiIt to higher values
with rising temperature.

Figure 5. Variation oI Polarisation Current with
Temperature

Figure 6. Variation oI Depolarisation Current with
Temperature

It is also noticeable that the shape oI the current also
changes as the temperature rises which indicates that
the insulation system looses its linearity at higher
temperatures. By increase oI temperature, the change in
amplitude and time dependence nature oI the currents is
mainly due to the decrease oI the time constant oI
interIacial polarisation |4|. Oil and paper conductivities
at diIIerent temperatures calculated Irom the
corresponding polarisation currents using (8) and (11)
are presented in Table 2.

Table 2. Oil and paper Conductivity vs. Temperature

Temperature (
0
C) o
oil
(pS/m) o
paper
(pS/m)
25 0.95 0.012
30 1.4 0.025
40 3.4 0.064
45 7.2 0.12
65 14.1 0.7
75 28.0 1.4

Figure 7 and Figure 8 show the nature oI variation oI oil
and paper conductivity with temperature. Both the oil
and paper conductivities are Iound to increase
exponentially with temperature.

Figure 7. Variation oI Oil Conductivity with
Temperature

Figure 8. Variation oI Paper Conductivity with
Temperature

It was reported by |9, 15| that conductivity Iollows an
exponential law:
( )
|
.
|

\
|

~
kT
E
ac
e A T . o (13)
where T is the absolute temperature in Kelvin, A is a
constant related to the mobility oI ions in the insulation,
k is the Boltzman constant and E
ac
is the activation
energy. Equation (13) can be re-written as:

( ) ( )
kT
E
A T
ac
~ ln ln o (14)

Equation (14) represents a linear relation between the
logarithm oI conductivity and inverse oI the absolute
temperature (1/T). The validity oI the above equation
(14) can be demonstrated by the Iollowing two Figures 9
and 10 where the natural logarithms oI oil and paper
conductivities have been plotted against the inverse oI
absolute temperature.

Figure 9. Oil Conductivity vs. Reciprocal oI Absolute
Temperature


Figure 10. Paper Conductivity vs. Reciprocal oI
Absolute Temperature

As predicted by (14) the nature oI the relationships are
almost linear. Similar nature oI variation oI average oil-
paper composite conductivity with temperature was
reported by |9|. It is interesting to note that both oil and
paper conductivities separately demonstrate such linear
nature oI variation with inverse oI absolute temperature.


6. CONCLUSIONS

Interpretation oI RVM and PDC test results still remains
a diIIicult task as it is believed to be inIluenced by
insulation ageing condition, moisture content and also
by environmental condition like the operating
temperature. Results oI RVM and PDC tests perIormed
on a transIormer under laboratory condition with
controlled variation oI operating temperature are
presented in this paper. The dominant time constant oI
the RVM tends to shiIt towards lower values oI time as
the temperature increases. The polarisation and
depolarisation current magnitudes and also the oil and
paper conductivities estimated Irom these currents tend
to be higher at higher temperature. The oil and paper
conductivities calculated Irom the measured polarization
currents are Iound to have a deIinite mathematical
relationship with the temperature. It is thus essential to
take these variations due to operating temperature into
account while interpreting the RVM and PDC results Ior
transIormer insulation condition assessment.

8. REFERENCES

|1| A. Bognar, L. Kalocsai and G. Csepes, E. Nemeth
and J. Schmidt, 'Dielectric Tests oI High Voltage
Oil Paper Insulating Systems (in Particular
TransIormer Insulation) Using DC
Dielectrometrics, in Proc. CIGRE, 33rd Session,
Vol 2, Paris, 1990, pp. 15/33-08.
|2| T. K. Saha, D. J. T. Hill, Z. T. Yao, G. Yeung and
M. Darveniza, 'The Application oI InterIacial
Polarization Spectra Ior Assessing Insulation
Condition in aged Power TransIormers, CIGRE
session, Paris, 1998.
|3| E. Ildstad, U. GaIvert and P. Tharning, 'Relation
between Return Voltage and Other Methods Ior
Measurements oI Dielectric Response, IEEE
International Symposium on Electrical Insulation,
USA, pp.25-28, 1994.
|4| Vahe Der. Houhanessian and W.S. Zaengl, 'On-Site
Diagnosis oI Power TransIormers by Means oI
Relaxation Current Measurements, Proc. IEEE
International. Symposium on Electrical Insulation,
USA, pp.28-34, 1998.
|5| T. LeibIried and A.J. Kachler, 'Insulation
Diagnostics on Power TransIormers Using the
Polarisation and Depolarisation Current (PDC)
Analysis, ConIerence Record oI the 2002 IEEE
International Symposium on Electrical Insulation,
USA, pp.170-173, 2002.
|6| J.J. AlII, Vahe Der. Houhanessian, W.S. Zaengl and
A.J. Kachler, 'A Novel, Compact Instrument Ior
the Measurement and Evaluation oI
Polarisation/Depolarisation Currents Conceived Ior
On-Site Diagnosis oI Electric Power Apparatus,
ConIerence Record oI the 2000 IEEE International
Symposium on Electrical Insulation , USA, pp.161-
167, 2000.
|7| A. Kozlovskis and J. Rozenkrons, 'Temperature
Dependence oI Return Voltage Characteristics,
IEEE Trans. Power Delivery, Vol. 14, Issue: 3, pp.
705-708, 1999.
|8| Y. Sheiretov and M. Zahn, 'A Study oI the
Temperature and Moisture Dependent Dielectric
Properties oI Oil-Impregnated Pressboard,
ConIerence on Electrical Insulation and Dielectric
Phenomena, pp. 487 - 492, 1993.
|9| A. C. Gjaerde, L. Lundgaard and E. Ildstad, 'EIIect
oI Temperature and Moisture on the Dielectric
Properties oI Oil-Impregnated Cellulose, Proc. 9
th

International Symposium on High Voltage
Engineering, Austria, paper No. 1060, 1995.
|10| R.J. Densley and B. K. Gupta, 'EIIect oI
Temperature on Sensitivity oI Diagnostic Tests on
Oil-Impregnated Paper Insulation, Proc. IEEE
Electrical Insulation ConIerence and Electrical
ManuIacturing and Coil Winding ConIerence, pp.
601-604, 2001 .
|11| C. Hongyan and S. Birlasekaran, 'Temperature
Dependent Relaxation Studies on Oil-Filled
TransIormer, IEEE International Symposium on
Electrical Insulation, pp. 174 -178, 2002.
|12| A. K. Jonscher, 'Dielectric Relaxation in Solids,
Chelsea Dielectric Press, 1983.
|13| U. GaIvert and E. Ildstad, 'Modelling Return
Voltage Measurement oI Multi-Layer Insulation
Systems, Proc. 4th International ConIerence on
Properties and Applications oI Dielectric Materials,
AUS, Paper 4234, pp. 123-126, 1994.
|14| G. Frimpong, U. GaIvert and J. Fuhr,
'Measurement and Modeling oI Dielectric
Response oI Composite Oil/Paper Insulation, in
Proc. 5th International ConIerence on Properties
and Applications oI Dielectric Materials, Vol. 1,
pp. 86 -89, 1997.
|15| R. Diabi, J. C. Filippini, C. Marteau and R.
Tobazeon, 'On the Role oI Temperature and
Impurities in the Low Field Conduction oI
Insulating Liquids, Proc. 12
th
International
ConIerence on Conduction and Breakdown in
Dielectric Liquids, pp. 350 -353, 1996.