Sie sind auf Seite 1von 5

2004 Annual Report Conference on Electrical Insulation and Dielectric Phenomena

Dielectric response of oil-impregnatedpaper insulation:


Variation with humidity and ageing level.
D. Linhjell, U. Gafven2 and L.E. Lundgaardl
SINTEF Energy Research, Trondheim, Norway
ABB Corporate Research, Vasteris, Sweden

Abstract: The dielectric response of oil-impregnated


Kraft paper with 3 initial humidity levels (0.17 %, 1.75
% and 3 3 %) has been studied in the frequency range
0.1 mHz - I kHz,before ageing at DP (degree of polymerisation) 1200 and after ageing lo DP 900,600 and
2-300,primarily to investigate if the method can he used
for nondestructive characterization of the state of cellulose insulation in electrical power devices, e.g. transformers. Also, because ageing produces acids, known
acids have been added to unaged paper to investigate
the influence on the dielectric response. For each case,
curves of log Gand log c vs. log frequency obtained
at several temperatures in the range -10 to +80 OC could
readily be shifted to form a master curve at 20 OC, indicating a single Arrhenius-type activation energy. The
master curve covers a frequency range of twelve decades from 10. Hz to I d Hz.The low-frequency par01 of
the curves show a combination of resistive behaviour
and low-frequency dispersion. At 1,756 and 33% humidity a loss-peak appears around IO- Hz. The lowfrequency dispersion and DC conduction also increases
with moisture, and the curves are generally shifted towards higher loss and higher frequencies. The effect of
ageing is small compared to the effect of moisture, hut
the loss peak of the humid samples become more
prominent. Light acids (formic, acetic), which are ahsorbed in the paper, have an effect on the dielectric response similar to water.
Introduction
The electrical insulation in most power transformers
consist of paper on the windings, the windings and/or
core separated with several presshoard cylinders and
ribs making ducts, and all of it embedded in oil which
also impregnates the pressboard and paper and acts as a
coolant. In power transformers in service, both oil and
paper will deteriorate (age). Thus the transformer
owners need to maintain their transformers. What degree of maintenance is considered economically appropriate depends upon the state of the insulation and the
importance of the transformer. Thus somehow the slate
has to be assessed. The slate of the cellulose (paper and
pressboard) is difficult to assess, as taking samples is
usually both impractical and destructive. As long as the

0-7803g584-5/04/$20.00 02004 IEEE

transformer is kept on constant temperature, there will


he an equilibrium distribution of ageing by-products
like water [I-31,acids and furans between the oil and
the cellulose. Thus oil analysis is regularly used for
practical state assessment of both oil and paper, hut as
the partition equilibrium coefficients of the various byproducts are both probably interdependent and for most
of them not well known [4,5],it gives limited information about the state of the cellulose. Further, load varies
and temperatures are therefore not constant.
Thus in recent years there has been considerable interest in the possibility of nondestructive electrical
measurements on the terminals of the transformers, with
3 methods collectively known as dielectric spectroscopy
c6.71. The two time domain methods are polarisation
and depolarisation currents (PDC), and recovery voltage
measurement. The interpretation of the latter has been
shown to be difficult 17-91, The present study uses the
third method, known as frequency domain spectroscopy
(FDS), partly because this is considered less noisesensitive in future field-use than PDC [7,9], and partly
because PDC has difficulties in measuring the fmt few
seconds after switching to polarisation or depolarisation,
limiting its equivalent frequency to helow I Hz. The
information one primarily hopes to get from this is the
moisture of the cellulose, as moisture accelerates ageing
considerably [IO]. and to some extent the reduction in
chain length of the cellulose polymer. The chain length
is usually given as degree of polymerisation (DP), the
average number of monomers in the chains. Ageing of
oil and paper produces acids. In the present study, the
study of effect of acids on the dielectric response is also
under way, and results up to the time of writing are included here.
Several studies have concentrated on pressboard 18,
111 as this constitutes the bulk of the solid insulation
between the low-voltage and high-voltage windings.
This study has so far concentrated on paper, as this is
also unbleached cellulose of the same type as presshoard, and because it is thinner both moistening and
moisture measurement are easier.
In an actual transformer the paper will he in series
with oil gaps and spacers (also cellulose) that will make
spectroscopic results differ from the results obtained
without oil gaps. ?he properties of the oil must be taken

262

into account. There exist computation models to handle


this [6-9,121.

20 C, with the same activation energy effectively used


for both o a n d x.

Theory

Experimental.

It can he shown that under a sinusoidal field, in a linear


and
isotropic
dielectric,
the
current
density J(r) = a E ( t ) + a D / a t , or rather its Fourier m s -

The paper samples and ageing conditions. The paper

form J ( o )= a E ( w ) + i w D ( w ) ,can be written [6,13]:

where o i s the angular trequency, b the DC conductivity, f and f the real and imaginary components of the
complex susceptibility x, E the electric field and D the
displacement. Considering that D = e O E , where
E =I+x is the relative permittivity, we may, for the
purpose of FDS, define a complex & by
J ( w ) = i a E O o E ( wwith
) components
(2)
&(o)=Re(e(a))=I+~(w)
being the capacitive part, and the loss part into which
even the DC conduction has been defined is
a
~ (=oIm(&(o))
)
= -+ ~ ( 0(3)
)
00

It can be shown that f and f are Fourier sine and cosine transforms of the same timedomain function f f r ) .
and are therefore not independent [6,13]. Note that (3)
and this interdependence has the consequence that in a
frequency range where E is constant, a slope of -1 in a
log-log plot of E vs. oindicates pure conduction. The
time-domain functionj7rJ is known as the dielectric response function and is what is measured directly in the
PDC method, except for a contribution from b.
For many dielectrics, the shape of x and b a s functions of frequency does not change with temperature;
changing temperature merely shifts them [6,13]. Then,
forx, and similarly for 0 :

where each A very often can be expressed with a single


Arrhenius-type activation energy: A = exp(-W I kT),
the W s not necessarily being the same in y and o directions or for x and Q The shifts are often negligible in
the ydirection. When recording at different temperatures, but each time within a Limited frequency range,
one can make a resulting extended-frequency-range
masrer curve at a reference temperature by shifting the
curves from the individual temperatures. The activation
energy can he calculated from the shifts.
The results in the present study are presented as
master curves of E and E at a reference temperalure of

used is regular Kraft paper, Munksjo Termo 70. For the


dielectric measurements, paper discs of diameter 76 mm
were used (measurement electrode diameter is 60 mm).
Acidity and moisture content is measured on paper
strips I21 mm long and 12 mm wide. All paper hatches
contained equal dry weight of discs and strips.
Table 1 shaws the ageing conditions and the DP
when samples were taken for dielectric response measurements, both for the ageing experiment and the acid
addition experiment. 0,17 % initial humidity is also referred to as dry. The ageing times needed were calculated from a previous ageing experiment [IO]. Ageing
was done in glass jars with tight screw-cap Lids, containing 70g (dry weight) of papkr and 1,45 1 oil.
In the acid addition experiment, experiments are
done on moist paper (2,6% moisture) and dry paper (not
yet done). Acid combinations are none, light acid mix,
and heavy acid mix, where light acid mix contains formic and acetic acid, heavy acid mix naphtenic and
stearic acid. The paper and oil is kept in glass jars with
tight screw-cap lids, containing 12,2 g (dry weight) of
paper and 250 ml oil.
Table 1. Ageing, acids and papeconditions at sampling.
I%] 0.17 1.75 3.5 2.6 2.6
2,6
Aeeinelemn.IC1
130
110
110
AddedBcid.& .
Light Heavy
DP, 1.sampllng
12W l2W
lux) 1zW
luy)
IZM)
DP, 2.sampling
840
820
loo0
DP, 3.sampling
580
590
DP, 4.sampling
360
340
280
lniL humidity

Preparation of the samples. The paper was first evacuated and then dried for 3-4 days at 100 C and then
cooled to room temperature, all in a vacuum oven. Then
it was quickly transferred to large glass jars with glass
lids and evacuated again. For paper to be moistened, a
calculated amount of water was admitted through a cock
into a glass beaker inside the jars, and the jars were then
kept at 45 OC for 4 days to let the water diffuse into the
paper. Then the paper (still under some kind of vacuum)
was impregnated with dried, degassed new oil. The paper was then transferred to the ageing or storage jars,
and dry, degassed new oil added. In the acidic cases, the
acids had been added to the oil used in this stage. The
acidic oil had a neutralisation value (NV)of 0,4 (unit is
mg KOWg oil), with the acid amounts calculated to let
each of the two acids in each mix contribute NV = 0 2 .
The light acids jars were kept at room temperature for
3 weeks before measuring, while heavy acids got another week and a half in addition.

263

Measurements and test cell. At each sampling time,


moisture in oil and in paper was measured with a coulometric (electrochemical) Karl Fischer litration technique. For the paper an extraction oven at 110 O C was
used, with a pre-dried gas flow carrying the water to the
titration vessel. The acidity of an oil sample was measured by titration with KOH, like in 151, and the acidity
of paper was measured by first extracting the acids with
distilled water and then tiuating the water extract. A
sample consisting of a stack of 3 paper discs was transferred to the dielectric measurement cell. The combined
thickness of the 3 paper layers was 155 - I65 pm. In the
ageing experiments, 70 ml of new, dried oil was added
to replace the oil samples taken.
The dielectric test setup is shown in fig.1. The test
cell consists of an upper HT electrode disc and a lower
measurement electrode disc with guard. The outer pan
of the upper electrode is kept I mm from the lower electrode by a Teflon distance ring. The central pan of the
upper electrode is a piston loaded with 20 kg. The piston diameter is larger than the central measurement area
of the lower electrode. All electrodes are cut from the
same piece of stainless steel. O-rings between the electrodes and the distance-ring and around the piston keep
the cell gas-tight, so that the moisture amount inside
will stay constant. A small amount of oil is added to fill
the spaces around the paper sample, and above there is a
small air volume as expansion volume for the oil at the
elevated temperature measurements. The cell is kept in
a temperature-controlled oil bath. The constancy of the
temperature is usually within 0,1 C. The temperature
sequence used (%) was 2040-60-80-20-0- -10-20.
The dielectric test apparatus is a General Electric
Programma IDA 200 instrument. Frequency range used
at 80 C is I kHz down to 0,l mHz, at lower temperatures the lower end used was I mHz. For each sample
thickness, the geometric capacitance has been calculated, and the results automatically presented as E and
dvs. frequency.

RSult.5
From earlier ageing experiments [IO], the moisture level
of the paper was expected to rise considerably as a result of the ageing, as cellulose chain scissions produce
water. It did not (fig.2). may be except for the initially
dry paper. IiDP is approx. proportional to the number of
chain scissions. The variation in measured moisture
level is hardly significant. It is strange, because the
moisture of the oil did increase somewhat (fig.3). In a
way it is advantageous, because then the other effects of
ageing are more easily seen apart from the effects of
humidity in the dielectric curves.

A
-

..

t.
._

{:j

- *

o.mr

nmz

om

0 ma

l0P

Figure 2. Moisture in paper vs. ageing.

-....--

*------------A
4
5 12

miwm-mp.p.
+
0,IlX

<,,I%
lJII

5 .

0 m,

om

0 -3

om

(mp

Figure 3. Mokmre in oil vs. ageing.

As the gasket of the lid of the jar with dry paper at


130 C did not seem to withstand the temperature, after
a while the paper started ageing much faster than expected. Therefore no results were obtained for dry paper
at medium ageing, i.e. DP 600.
Master curves have k e n made at both 50V and
2oOV measuring voltage. They are so similar that only
50V results are presented here.
In figure 4a-d the three moisture levels are compared at different degrees of ageing. Moisture obviously
has a considerable effect, mainly on E, hut also on E
in the very low frequency region. A small loss peak
seems to be embedded in the othenuise resistive pan of
the curve, the size of the peak increasing with increasing
moisture. In addition to the general shift caused by
moisture, this feature should be easy to use to identify
moisture. It also seems that moisture has a profound

264

,,lo.

1.10.

c)

<,,b ,nolrld
".sUnrWl.

1.10'

2.w

In@

Paper with 3.55 % imual moisture

Figure 5. zcompared across the ageimg levels at


various degrees of initial humidification.

Figure

effect on the frequency for transition between essentially resistive behaviour and the so-called "lowfrequency dispersion" where the curves of &' and &" get
the same slope. For the two moist papers, a small field
dependence was sometimes observed at frequencies
below the frequency of the embedded loss peak, with
200V showing slightly higher losses than SOV.
In figure Sa-c the effect of ageing is compared for
the three moisture levels. The effect is considerably
smaller than the effect of moisture. What more, the initial reduction of DP from about 1200 to about 900 is
accompanied by a small reduction in E'' and in the very

265

-I

*-*
*-----.Iv,_/
_ - - - -I..#,_*r
- ..
-

,m

I_

BI
I

I.__<

..I-^

Irn

10

D,

DO,

?I_ 3r.m

>I_ 1E-1

ramno

11.m

*E*,

increased conductivity of the oil. Apparently it is not the


degree of polymerisation of the cellulose, but the presence of ageing byprocucts that influence the dielectric
response. For a real transformer, it may be impossible to
distinguish between the effect of acids and the effect of
moisture. Both influence diagnostics and ageing in a
similar manner. A complete diagnosis therefore has to
be based on more than one single method. Our results
indicate that dielectric diagnoses have to be augmented
by chemical oil analyses to he able to separate ageing
from humidity. Anyway, both water and acids are ageing accelerators that have to be removed to increase life
expectancy of a transformer.

/I_

Figure 6 . The effect of acids on E

low frequency part of E. Funher ageing causes a slight


increase in both E and the low frequency part of E. In
particular, for the paper with 1.75 9% moisture, the embedded loss peak increases with ageing, even though
there is no early increase in moisture.
Figure 6 shows the effect of acids in a system with
2.6 % moisture in the paper. The light acids cause a
significant increase in E and some change in shape of
E in the transition region to low-frequency dispersion.
The heavy acids curve is almost undistinguishable from
the curve without acids. lhis is in line with the results
that heavy acids favour the oil and hardly enter the paper, while light acids prefer the cellulose [SI.Note that
the curve with light acids has a shape only marginally
different from the curve for 2.6 % moisture alone.
The ageing did produce a small amount of acids.
NV of the oil increased typically from initially 0.005 to
finally 0.04 - 0.05, and there was typically a fivefold
increase in acidity of the paper. Thus some of the produced acids are likely to be light and contribute to the
shifts seen in figure 5 .
The activation energies calculated in the master
curve fining process were in the range 1.01 k 0.08 eV.

Discussion and conclusion


The effect of moisture at the levels used in the present
study should he large enough to be detectable even
when the response is modified by a series oil gap, unless
there is very little cellulose and very much oil in series.
This is similar to what is found in studies involving
pressboard [8,11]. The results regarding moisture are
also similar to results found for mass-impregnated cable
~91.
Acids are produced when ageing cellulose [lo], and
light acids in a moist insulation cause increased losses.
Heavy acids do not cause increased losses in the paper,
simply because they are not found in the paper but in
the oil [51. Of course, the dielectric losses in an insulation system with oil gaps may increase also because of

Acknowledgements
We are grateful for the support from the Norwegian
Research Council, Statnelt SF, Stalkraft SF, Norsk Hydro and Hafslund Nett that made this study possible.

References
J.Fabre and A.ichon.: Deteriorating Rocesses and Roducls
of Paper in Oil. Application to Transformers. 19W I m Cod.
on Large High Voltage Elm. Systems (CIGRE), Paris, Paper
137.
(21 T.V.Oommen: Moisture Equilibrium in Paper-OilSysrems.
Roe.of the ElecUiccaVElecmmics Insulation Gmfwnce. Chicago, oct.3.6.1983, pp. 162-166.
B.C.Lesicutre, A.V.MamiEhev and S.Lindgren:
131 Y.Du,, M.&,
Moisture Equilibrium in Transformer Paper-Oil Systems.
IEES El=. Insd. Magarine,Vol.ll No.l,Jan/Feb 1999.
I41 D.Linhjell, W.Hansen. S.lngebrigtsen and L.Lundgaar(l: Moisture Absorption in Service Aged Transformer OW. Roc.
NORD-IS 03, T a m p , Finland. June 11-13 2W3. pp. 397.401.
151 S.Ingebrigtsen. D.Linhjell, WBansen and L.E.Lundgaard: Solubility of Cabxylic Acids in Poper (I(rafl)-Oil Systems. Roc.
CEIDP 2004, Boulder, O ~ 2004.
L
161 Dielecmc Response Methods for Diagnostics of Power Transformers.ReporioflheCIGREW15.01.09.
171 U.GifveR L.Adeen, M.Tapper. P.Ghasemi, B.J6mson: Dielecm c Specnoscopy in Time and Frequency Domain Applied to
Diagnostics of Pow- Tmformers. RoC. 2000 IEEE 6* 1C
Cod. on Properties and Appl. of Diel. Materials (ICPADM),
Xian. China, June 21-26. ZwO.
[81 V.d.Houhanersian: hlasmment and Analysis of Dielectric
Response in Oil-PaperInsulation Systems. Dissertation, Swiss
F e d 4 Institute of Technology, Ziinch 1998.
191 R.Neimanis: y)n Estimation of Moisture Content in Mass Impregnated Dismbution Cables. Dissertation,Royal Institute of
Technology (KTH), Stockholm 2wO.
[ 101 L . E . L u n d g 4 W.Hansen, D.LinhjeU and TJ.Painter: Aging
Of Oil-Impregnaled Paper in Power Transformers. IEEE Trans.
on Power Delivery, vol. 19. No. I, Jan. 2004, pp. 230-239.
[ I l l C.Ekanayake: Application of Dielectric Specmsmpy for Estimating Moisulre Content in Power Transformers. Dissertation,
Chalmer~Univ. of Technology, Gothenburg, 2003.
[I21 U . G i h e r ~G.Fimpong and LFuk. MModelling of Dielenric
Measuremenls on Power Tramfomers. Paper 151.2, 1998 CIG& S W ~ O D . paris, A U ~ W1998.
1131 A.KJonscher: IlieleemC Relaration in Solids. Chelsea Dielecmcs Press Ltd, London 1983.
Corresp.author: dag.linhjeU@sintef.no
[I]

266

Das könnte Ihnen auch gefallen