Resistivity of transformer oil at low and medium field

strengths
N. L. Harrison

Synopsis
In certain terminal equipment used in high-voltage d.c. transmission schemes, transformer oil is subjected
to a d.c. stress. The behaviour of the oil under this stress is important in the design of equipment. The
volume resistivity p of five samples of transformer oil, ranging from new oil to oil after 10 years' service
in a transformer, was measured over a range of stress levels E and electrode spacings d. The resistivity
increased with increasing field strength up to a critical field strength, above which it fell sharply. Below
this critical value, the resistivity was found to be a function of the ratio Eld. The curve of p against Ejd
consisted of two parts: id) for low values of E/d, p was constant, (6) for higher values, the resistivity followed
/E\"
the equation p = a f — J where a and n are constants. Some potential-distribution measurements were also
made, and the stress in the oil was found to be uniform and approximately equal to the applied stress over
the bulk of the oil in the interelectrode space. The results of the resistivity measurements are interpreted in
terms of electrolytic conduction due to dissociation of weak electrolytes in the oil.

List of symbols
p = volume resistivity of oil
£ = average electric field strength
d = electrode spacing
C, = concentration of positively charged ions
C_. = concentration of negatively charged ions
U.h = mobility of positively charged ions
U. = mobility of negatively charged ions
k — dissociation rate for electrolyte
a = recombination coefficient for positive and negative
ions
e = magnitude of the charge on either ion
J — current density
J+ = current density due to positive ions
J.. = current density due to negative ions
T((/_) = mean lifetime of negative ions of mobility U..
A((/_) = mean free path of ions of mobility U_
Uc = value of U_ for which X(U_) = d
Uo = maximum value of U_
a, n, k, Ko and b are constants which are defined in the text
1 Introduction
In recent years, as a result of the growing number of
d.c. power-transmission schemes, a detailed knowledge of the
behaviour of transformer oil under d.c. stress has become
necessary. In certain components of a d.c. transmission
system, oil is subjected to a d.c. stress in service, and its
behaviour under the influence of the direct field is important
in the design of equipment. In a composite insulation con-
sisting of two or more layers of different materials in series
under the field, the potential distribution across the layers is
determined by the resistivities of the components. Measure-
ment of the d.c. resistivity of oil in the Forrest cell1 has been
an important test for the serviceability of an oil for some
time, but it is now necessary to know what the resistivity will
be under the operating conditions. It is well known2"4 that
the resistivity of insulating liquids is dependent on both
stress and the electrode spacing at which the measurements
are carried out; the purpose of the present work was to
determine systematically the nature of this dependence in
transformer oil.
In order that the results from small-scale experiments could
be applied with confidence to the conditions obtaining in
service, the investigation was also aimed at obtaining funda-
mental data on the conduction mechanism in oil. Conduction
Paper 5518 J, first received 16th November 1967 and in revised form
5th January 1968
Mr. Harrison is with the Nelson Research Laboratories, English
Electric Co. Ltd., Stafford, Staffs., England
736
in electrolytes normally follows Ohm's law. Since the con-
duction in oil is nonohmic, it is either nonelectrolytic, or
else there is some significant difference between electrolytic
phenomena in oil and in aqueous or other solutions for which
Ohm's law is obeyed. Among possibilities in the former case
are cascade or field-emission processes, or conduction by
colloidal or larger particles suspended in the oil.5 Alter-
natively, if the conduction is, in fact, by an electrolytic process,
the apparent resistivity of the oil may possibly be affected
by the presence of space charges which modify the field in
the oil, field-dependent dissociation or recombination of the
ions or ion depletion through neutralisation at the electrodes.
Forster3 explains some of his results on conduction in
benzene in terms of modification of the field in the liquid
by space charges near the electrodes. To test to what extent
his hypothesis is true in oil, some potential-distribution
measurements were made on oil under d.c. stress.
2 Method of test
The test cell designed for this work enabled both
resistivity and potential distribution to be measured, and its
construction is shown in Fig. 1. Tt incorporated a pair of plane-
parallel brass electrodes approximately 10cm in diameter, the
cathode being surrounded by a guard ring to eliminate the
effect of fringing fields. The electrode spacing could be varied
between 0-1 cm by means of a micrometer screw. A vacuum-
tube voltmeter of 108Q input resistance was used for current
measurements on degraded oils, but for new oil, where the
current could be less than 1 nA, a vibrating-reed electrometer
in parallel with a shunt resistor was necessary.
A guarded-probe system, shown diagrammatically in Fig. 2,
was used for the potential-distribution measurements. The
probe itself was a nickel wire, of 0-005 in diameter, stretched
across the interelectrode space, parallel to the faces of the
electrodes to within 001 cm. The ends of the probe wire left
the cell through a pair of copper tubes 0-2cm in diameter.
These formed part of the probe-screen system shown in
Fig. 2. With this arrangement, which was designed to eliminate
entirely errors due to electrical leakages from external con-
nections to the probe, it was necessary to use a null detector,
both terminals of which could be raised to the high potential.
An Ayrton-Mather reflecting electrostatic voltmeter with a
sensitivity of 1 cm/V was found to be suitable. The minimum
detectable probe current with this arrangement was deter-
mined by the resistance between the probe and the probe
screen. This current was always at least two orders of mag-
nitude less than the-current flowing between the electrodes.
The cell was totally enclosed, and during measurements the
space above the oil in the cell was filled with dry nitrogen.
PROC. 1EE, Vol. 115, No. 5, MAY 1968
For measurements at 80°C, the cell was placed in a thermo-
statically controlled oven incorporating forced air circulation.
A Forrest cell1 and a miniature cell with plane-parallel
the cell was allowed to stand for several hours before any
measurements were taken. The oil was stressed for 5min
under any chosen set of conditions before any current value
was recorded, and in all cases it appeared to have effectively
reached equilibrium in this time.
Measurements were made on the four different oil samples
tabulated in Table 1.
All the oils were of sufficiently low moisture content (less
than 40 parts in 106) to ensure that no water was held in
suspension.5

3 Experimental results
3.1 Resistivity measurements
The stress levels studied were limited to below
80kV/cm, as above this value some of the oils were liable to
break down. The general features of the variation of resistivity
with stress and spacing, which were common to all the oils,.
were as follows:
(a) The resistivity of the oil increased with increasing stress
up to a certain critical stress level, above which it fell sharply.
For the degraded oils this critical stress level was about
40kV/cm, but for the new oil it was much lower, about
lOkV/cm. No systematic study was made of this feature or
of the variations in resistivity above this critical stress level.
The generalisations used later refer to oil at stresses below
the critical level.
Fig. 1 (b) The resistivity of a single oil at constant temperature
Construction of cell for the measurement of resistivity and potential varied widely (in one case, by two orders of magnitude) with
distribution in transformer oil changes in electrode spacing d and stress £ up to the critical
Both electrodes are movable with respect to the probe, and are maintained at a level. A typical set of results (taken on oil 3 at 80°C) is
constant spacing by means of the external clamp
shown in Fig. 3, in which each curve corresponds to a con-
stant spacing. Fig. 3 shows that the resistivity increases with
decreasing electrode spacing. In Fig. 4, the resistivity results
probe screen
shown in Fig. 3 and some additional results obtained with

aezs
pro be •

10-0
Fig. 2
Circuit for potential-distribution measurements
h.v. High-voltage supply
v.t.v.m. Vacuum-tube voltmeter, IO8C2 input impedance
e.s.v.m. Reflecting electrostatic voltmeter

stainless-steel electrodes spaced 017cm apart were used to

check the values obtained with the variable-spacing cell.
The cell was cleaned by first immersing all parts in redistilled
petroleum spirit, assembling, and then flushing several times
with petroleum spirit drawn in through a no. 4 sintered-glass 1-0
filter (pore size 5-15/>im). The cell was drained and evacuated
to evaporate any residual solvent. It was then filled with the
oil to be tested. This was drawn in under vacuum in the
same manner as the flushing solvent. The oil received no
treatment apart from the filtration and consequent degassing.
The vacuum in the cell was broken with dry nitrogen, and
2 .
Table 1
DESCRIPTION OF OIL SAMPLES
0-1
0-5 1-5 3-5 4-5
Oil Description Acidity
(mg KOH/g)

I New oil 0 005 Fig. 3

2 New oil degraded at 100°C by bubbling 0 028 Resistivity of oil 3 at 80°C
air through in an open vessel a Electrode spacing = 7 mm
3 Oil after 4£ years' service in a generator O-0I0 b Electrode spacing = 5 mm
transformer. No treatment c Electrode spacing = 3 mm
Oil after 10 years' service in a generator d Electrode spacing = 1 - 5 mm
4 0 067 e Electrode spacing = 10mm
transformer. Meta filtration each year / Electrode spacing = 05mm
g Electrode spacing = 0-25 mm
PROC. IEE, Vol. 115, No. 5, MA Y 1968 737
the Forrest cell and the miniature cell are plotted against the the oil between the electrodes. This then allows a value of
ratio Ejd. It will be noted that all the points lie on the same resistivity independent of electrode effects to be calculated.3
curve. The results obtained with all the oils tested showed The potential distribution in new oil (oil 1) at a range of
100

O10-0
K
s

1-0

0-1 lo
«iof-. v / e - a
log.
Fig. 5
Fig. 4 Resistivity of oils 1,2, 3 and 4, at 23 and 80° C
a Electrode spacing = 7 mm
Resistivity of oil 3 at 80° C e Electrode spacing = 1 mm
a Electrode spacing = 7 mm g Electrode spacing = 0-25 mm
b Electrode spacing = 5 mm
c Electrode spacing = 3 mm
d Electrode spacing = I-5mm
e Electrode spacing = 1-Omm electrode spacings from 0-25 to 0-9cm, and stress levels from
/ Electrode spacing = 0-5mm 0 1 2 to 30kV/cm, was determined; and a single determina-
g Electrode spacing = 0-25 mm
F Forrest cell tion was made on oil 2 at a spacing of 0-5cm and a stress
S Miniature cell
level of 4kV/cm.
Although the field very close to either electrode could be
this feature, and the curves are shown in Fig. 5. In all cases, appreciably greater than the applied field, the field in the
complete curves of resistivity against stress were obtained at bulk of the oil was never significantly different from the
all the electrode spacings given in Fig. 3; however, for the applied field. The maximum ratio of applied field to field in
sake of clarity, only the points corresponding to spacings of the oil was 1-4, and this was obtained at the lowest field
7, 1 and 0-25 mm are plotted in Fig. 5. strength of 012kV/cm. At higher field strengths the ratio
The characteristic for each oil consists of two more or less was very close to ons.
distinct regions: While this modification of the field by space charges
(i) an ohmic region, in which the resistivity is independent of should be taken into account in calculation of the resistivity
both stress and spacing; this occurs at low values of Eld of the bulk of the oil in the interelectrode space, the correction
(ii) a nonohmic region, in which the resistivity depends is very small compared with the changes in apparent resistivity
strongly on the ratio E\d. At sufficiently high values of brought about by variation of Ejd.
Ejd, the resistivity p follows the equation No correction of this kind has been made to the values of
resistivity plotted in Figs. 3-5.
'E
(1) 3.3 Time-dependent effects
When a step voltage is applied to a dielectric, there
a and n are constants which have different values for is an initial pulse of current of short duration (shorter than
different oils. the time of measurement with the instruments used in these
a may vary by several orders of magnitude, from one oil experiments), but even when this has died away the current
to another, but n takes a relatively narrow range of values. flowing in the dielectric usually varies with time, and, in
For the four oils tested, n lay between 0-32 and 0-53. general, its value decreases. This slowly decaying current is
The value of E\d for which the conduction becomes non- referred to as the absorption current.
ohmic depends on the condition of the oil. For a heavily Although no quantitative measurements were made on the
degraded oil it is somewhat higher than for a good oil. The absorption current in oil, two features were apparent during
new oil showed no ohmic region, possibly because it was not the measurements:
possible to obtain results at sufficiently low values of E/d. (a) Under a given set of conditions (temperature, stress,
spacing), the duration of the decaying absorption current was
3.2 Potential-distribution measurements of the same order of magnitude for all the oils. It was not
The main purpose of the potential-distribution strongly dependent upon their resistivities.
measurements was to determine the field in the bulk of (b) For values of E corresponding to the ohmic part of the
738 PROC. IEE, Vol. 115, No. 5, MAY 1968
resistivity curve, there was no absorption current; i.e. the
current flowing was constant from the time of application of
the voltage.
4 Interpretation of results
In some respects, the work described in this paper is
complementary to the work of Forster,3 who measured the
conductivity of benzene at a series of stress levels and electrode
spacings, and also measured the potential distribution in the
liquid. Results taken from Fig. 4 of Forster's paper have been
replotted in Fig. 6 of this paper as a log p against log E\d
2-
•d
5 -
log
iof' v / c m S
Fig. 6
Resistivity of benzene
Values taken from Fig. 4 of Reference 3
a Electrode Spacing = 1 -Omm
b Electrode spacing = 0-8 mm
c Electrode spacing = 0 6mm
(I Electrode spacing = 0-4mm
e Electrode spacing = 0-2mm
curve, i.e. in the same manner as the results obtained in the
present experiments. When plotted in this form, Forster's
results show a similar dependence of resistivity on the ratio
E(d as does a new transformer oil; if the curve is fitted to
eqn. 1, n = 0-30.
Forster also carried out a further series of measurements
in which he compared the conduction current with the field
in the liquid in the interelectrode gap. He found that, over the
central portion of the gap, the field was uniform and less
than the applied field by a factor of up to 2-5 : 1. In this
region he found that the ratio of the conduction current to
the applied field was approximately independent of the voltage
and electrode spacing, indicating, that the true resistivity of
the liquid was constant. However, in his first series of experi-
ments, the results of which are plotted in Fig. 6, Forster
found a resistivity variation of 5 : 1, whereas in his second
series of experiments the corresponding range was only
2-5 : 1.
The variations in resistivity found in the present series of
experiments of up to two orders of magnitude were much
larger than those found by Forster. Such large variations in
resistivity could not be accounted for by the relatively small
modification • in the field observed in the bulk of the oil.
Other published data on space charges in liquids2-6 show
PROC. IEE, Vol. 115, No. 5, MAY 1968
similar small modifications of the internal field, which tends
to support this conclusion.
In any attempt to find a theoretical model which will
account for the results, the number of possibilities is limited
by the fact that increase in field strength is equivalent to
decrease in electrode spacing; this excludes purely field-
dependent effects such asfield-assisteddissociation of netural
molecules. Among recently published theories on conduction
in organic liquids, there are a number7"9 which treat the
liquid as a weak electrolyte,, in which charge carriers are
generated at a constant rate per unit volume in the liquid,
the rate being independent of the electric field. The electrodes
neutralise ions arriving there, but do not supply any charge
carriers themselves.
This model will be used here, with the simplifying assump-
tions that thermal diffusion of the ions can be ignored and
that the field in the liquid is equal to the applied field. These
assumptions are implicit in the theory proposed by Zaky and
House in Reference 9. Tt is also implicit in their equations
that positive and negative ions have equal concentrations.
This latter assumption is justified in the case of relatively
strong electrolytes, where slight differences in the concentra-
tions of positive and negative ions would very radically
modify the field in the liquid. It is not justified in a dielectric
liquid where the ion concentrations are expected to be very
small. In the theory that follows, positive- and negative-ion
concentrations are not assumed equal, and the mathematical
solution of the equations shows that, in fact, they are not
equal.
4.1 Conduction in a liquid containing only two ion
species
We shall consider first a system in which only singly
charged ions of two types, one positive and one negative, are
present. The system is subjected to a field E between parallel
electrodes separated by a distance d. Ion pairs are created at
a constant rate k per unit volume, and recombine at a rate
aC+C_ per unit volume, where a is a constant. At equi-
librium, when all concentrations are constant in time, the
differential equations governing the concentrations of ions
in the liquid are
k-aC+C_ - (2)
dx
k - aC,C_ + U_E r = 0 (3)
dx
x being the distance measured from the anode. The boundary
conditions are:
(4)
x = o C+ = 0
(5)
x = d C_ = 0
since no ions are created at the electrodes. Solving eqns. 2
and 3 using the boundary conditions of exprs. 4 and 5 leads
to the following relationships:
U+C+ + U_C_ = K (6)
where K is independent of x
~1/2
/4k \''2
where KQ = (—U+U_) (9)
and A = 4* <10>
k
K may readily be related to the resistivity of the liquid since
P =
e(U*C+ + U_C_ ^7R
e being the charge on either ion.
739
 We see from eqn. 8 that K, and hence from eqn. 11 that from eqn. 7, with KOIK= 1 0 1 , are plotted. While C + = 0
p, is a function of Eld. at x = 0 and C + = K/U+ at x = d, over the middle portion
Corresponding values of resistivity and Ejd calculated of the curve, C + ~
from eqn. 8 are plotted in Fig. 7 in the same manner as the

IUU

5 -

2-
/

/
5

K.e
P I
I
Fig. 8
2 Positive-ion concentration as a function of distance from anode
KQIK = 1 -01; t/o and K are constants

As KJK becomes closer to unity, the ion concentration

across the gap becomes increasingly more uniform.

5
4.2 Conduction in a liquid containing several ion
species
On the basis of the preceding remarks, we can extend
the theory to cover the condition where ions with a range of
2
mobilities are present, by introducing the assumptions that:
(a) for any one species of ion, the mean free path before
n-i recombination is either much greater or much smaller than
- 4 - 3 - 2 - 1 0 1 2 3 4
the electrode spacing; i.e. we ignore ions in the transition
lofl
io | + lo
«.oA region, since there are relatively few of them compared with
other ions present
Fig. 7 (b) the concentration of ions for which the mean free path is
Theoretical resistivity curve when two ion species only are present shorter than the electrode spacing is uniform across the
e, Ko and A are constants interelectrode gap.
It will be assumed also, purely for the sake of simplicity, that
experimental values shown in Fig. 4 (resistivity of oil 3 at all ions carry the same charge, and that there is only one type
80° C against Eld). The curve consists of a horizontal portion of positive ion present, while the negative ions have a wide
and a rising portion, as with the experimental results. There range of mobilities.
are, however, two important differences: In the experimental results, there is no indication of the
(a) the transition is much sharper in the theoretical than in conduction current tending to a limiting value (i.e. of the
the experimental curves slope of the rising portion of the resistivity curve tending to
(b) the slope of the rising straight-line portion of the curve unity), and hence most of the ions created in the liquid must
is exactly one (corresponding to n = 1 in eqn. 1), whereas recombine before reaching the electrodes. This necessarily
the corresponding slopes in the experimental curves range applies to both types of ions, as they are created and recom-
from 0-32 to 0-53. bine in equal numbers. Hence the mean free paths of the
positive ions are. much shorter than the electrode spacing,
For the rising portion of the curve, eqn. 8 reduces to and so, by assumption (b), their concentration C + is uniform
kd across the interelectrode gap.
(12) Let / + and 7_ be the conduction-current densities due to
positive and negative ions, respectively. Then, at equilibrium,
so that the current density in the dielectric is
74_ = 7_ (15)
£
(13)
~ P~ e since the electrodes do not supply any charge carriers and the
rate of accumulation of charge in the liquid is zero.
This means that all ions created in the interelectrode space Hence
reach the electrodes. We therefore interpret the transition
from the ohmic to the rising portion of the curve as indicating j = 27 . = 2eU,EC, (16)
the stage when the mean distances of travel of the ions before
recombination, under the influence of the field, become equal so that
to, or greater than, the electrode spacing.
1 (17)
P =
As -r decreases, —" 1 2eU,C,
d K
The horizontal portion of the curve is defined by the condition or C\ = (18)
2eU+P
(14) It is reasonable to assume that recombination of ions is
K controlled by their external electric fields, which are the same
In Fig. 8, corresponding values of C+ and x calculated for all species. The mean lifetimes of negative ions are there-
740 PROC. 1EE, Vol. 115, No. 5, MAY 1968
fore proportional to their relative velocities with respect to
the positive ions,
i.e.
where b is a constant, and r{U_) is the mean lifetime of a
negative ion of mobility U_.
This allows calculation of the mean free path A(C/_) in the
direction of the field, of negative ions of mobility U_:
U_Eb
X(U_) = U_ET(U_) = (20)
yv++ U_)C+
We define Uc as the value of £/_ for which A((/_) =
From eqn. 20,
".-"•GSrO"
In accordance with assumption (a), we divide the negative
ions present into two classes:
(i) Those for which U_ < Uc. These ions have mean free
paths which are short compared with the electrode spacing,
and, in consequence, only those ions created within a distance
X(U_) of the anode reach it without recombination.
(ii) Those for which U_ > Uc. All these ions reach the anode
without recombination.
If there are/((/_) ions of mobility U_ created per unit volume
per unit time,
, Uc .Uo
J_ = e\ X(U_)f(U_)dU_ + ed f(U-)dU_
J
o J
U'
the first and second terms being the contribution due to ions
with U_ < Uc and U_ > Uc, respectively. Uo is the maximum
value of negative-ion mobility. Using eqns. 15, 16, 20 and 22,
= The author thanks Dr. Eastwood, Director of Research,
£ nu )du +
- - -E
In principle, this equation could be solved for C+ using the
value of Uc given by eqn. 21, provided that/(£/_) was known.
We note that the value of C + obtained in this way, and
hence of p, depends on the ratio Eld, and not on E or d
separately.
For sufficiently small values of Eld, Uo < Uc, and the
second term of eqn. 23 vanishes, leaving
U_
f(U_)dU_ • • (24)
~U\[ u, u_
i.e. C+ is constant, corresponding to ohmic conduction.
Tn the nonohmic region, the relationship between p and
E/d depends on the form on/(£/_). While there is no informa-
tion available to suggest what the form of /(C/_) should be,
PROC. IEE, Vol. 115, No. 5, MA Y 1968
a form can always be found which will fit the experimental
results on resistivity. For example, the form
(19) /((/_)= (25)
inserted into eqns. 17 and 23 gives, for sufficiently large values
of Eld,
(26)
P=\Te){l) Gc) G) • • • •
This is of the same form as eqn. I, with n — \. While eqn. 23
cannot predict the shape of the resistivity curve, its derivation
shows how the dependence of resistivity on Eld could arise.
5 Conclusions
(a) Space charges do not seriously modify the electric
(21)
field in transformer oil under a d.c. stress, except close to the
electrodes.
(b) The resistivity of oil depends on the ratio offieldstrength
to the electrode spacing. At low values of this ratio, the
resistivity is constant, but at higher values it follows the
equation
a (I)
P= IT
where a and n are constants whose values vary from one
oil to another. This relationship holds only up to a limiting
value of field strength, above which the resistivity falls
sharply.
. (22) (c) The results can be explained in terms of electrolytic con-
duction in the oil, on the assumption that the ions present
have a wide range of mobilities.
6 Acknowledgments
(23)
English Electric Co. Ltd. for permission to publish this article.
References
1
FORREST, j . s.: 'An electrical resistance test for insulating oils',
Proc. IEE, 1948, 95, Pt. II, pp. 337-342
2 WHITEHEAD, J. B.: 'Impregnated paper insulation' (Wiley, 1935)
3 FORSTER, E. o.: 'Electric conduction in liquid hydrocarbons. I
benzene', / . Cliem. Phys., 1962, 37, pp. 1021-1028
4 ZAKY, A. A., TROPPER, H., and HOUSE, H. : 'Electrical conduction in
organic liquids', Brit. J. Appl. Phys., 1963, 14, pp. 651-656
5 STANNETT, A. w.: 'The conductivity of hydrocarbon transformer oil
containing water and solid conducting particles', ibid., 1951, 2,
pp. 110-114
6 ZEIN ELDINE, M. E., ZAKY, A. A., HAWLEY, R., a n d CULL1NGFORD,
M. c.: 'Influence of electrode coatings on space charge distributions
in transformer oil', Nature, 1964, pp. 1309-1311
7 GAVIS, J.: 'Electrical conductivity of low dielectric-constant liquids
by d.c. measurements', / . C/iem. Phys., 1964, 41, pp. 3787-3793
8 SILVER, M.: 'Conduction of electricity in insulating liquids', ibid.,
1965, 42, pp. 1011-1013
9 ZAKY, A. A., and HOUSE, H.: 'Conduction currents in a chlorinated
hydrocarbon (pyrochlor)', Dielectrics, 1963, 1, pp. 140-149
10 CHILDS, D. c , and STANNETT, A. W.: 'Some aspects of the deteriora-
tion of insulating fluids', Proc. /EE, 1953, 100 A, Pt. II, pp.-61-67
741