~

IEEE Transactions on Power Delivery, Vol. 10, No. 2, April 1995 889

Further Studies Of Anomalous Phenomena In Dielectric-Loss

Measurements - Transformer Bushings Model

Raka Levi, Senior Member, IEEE Stan Manifase

Doble Engineering Company
Watertown, MA USA

-
Absfruct The results of ac dielectric-loss measurements on This field measurement serves as an early warning of an
bushings mounted in certain three-phase transformers, if not incipient failure of the element which represents the second
performed in accordance with the prescribed procedure, can most common source of transformer troubles. The
deviate significantlyfrom the correct values. An investigation of measurement technique is widely accepted by utilities and
the incorrect procedure has shown that the capacitive coupling endorsed by bushing manufacturers [3]. Standard
between adjacent windings and the lower portion of the bushings organizations such as IEEE/ASTM [4] and IEC [5] prescribe
influenced the results such that dielectric losses were elevated to
several times their normal values. Experience has shown that this test method and the limiting values for dielectric losses of
bushing test results on all three phases form a pattern in cases new bushings, expressed in percent power factor or tan 6.
where this phenomena is evident. Results of investigative field
measurements showing this phenomena are presented, and an Most bushing nameplates include factory test results. These
explanation of its origin is given. values are used as a benchmark, and any significant changes
from those values may be considered a sign of an internal or
I. INTRODUCTION external problem. Values obtained in excess of the nameplate
are 'indications of main insulation deterioration [6] and/or
Ac dielectric-loss measurements applied to transformer contamination [7], while lower than normal values indicate the
bushings equipped with test or potential taps has proven to be existence of specific conditions influencing the
an effective method of detecting contamination and/or measurements [81.
deterioration of the main insulation (core insulation) without
being influenced by the losses associated with the connected
transformer winding [ 11.

Many early bushing designs did not provide potential or test

taps, and the only means of testing the overall bushing
condition was by removing it from the transformer and
disconnecting it from the windings. Although excessively
expensive to perform, this method remains in use for bushings
equipped with taps as a final investigative test. The majority
of utilities worldwide have strongly relied on the utilization of
bushing taps as test electrodes [l], and the application of the
Ungrounded Specimen Test (UST) method since its inception
in 1948 [2]. Fig. 1. Test Procedure

Tests are performed in factories and in the field following the

same procedure. Factory or laboratory instrumentation allows The standard procedure as described in the test equipment
for applying nominal bushing voltage, while in the field it is user manuals [9] calls for short circuiting windings by
mainly limited to a maximum of 10 kV using portable ac test electrically connecting all the bushings together, shown as
instruments. The simplified test procedure is shown on Fig. 1. dotted line on Fig. 1. This part of the procedure is
occasionally omitted which, in certain transformers leads to
94 SM 415-0 PWRD A paper recommended and approved erroneous results. Since the problem is not apparent on all
by the IEEE Transformers Committee of the IEEE Power transformers, test crews tend to forget this requirement, thus
Engineering Society for presentation at the IEEE/PES errors are sometimes introduced in the measurements.
1994 Summer Meeting, San Francisco, CA, July 24 28, -
1994. Manuscript submitted December 20, 1993; made
available for printing April 18, 1994. 11. BACKGROUND

It is known that incorrect procedure for the bushing tests in

some three-phase transformers may yield anomalous results.

0885-8977/95/$04.00 0 1994 IEEE

890

Proper procedure requires that transformers with three

primary bushings (delta connected windings of phases A, B,
and C, as shown in Fig. 2), must be short circuited for the
bushing tests. This means all three bushings A, B, and C are
electrically connected, short circuiting the transformer IC
windings.

When dielectric-loss or insulation power factor tests are

performed on bushings using taps as test electrodes, all three
test results occasionally appear excessive if the transformer
windings are not short circuited. ,

I I

Fig. 2. Transformer bushings arrangement

Moreover, the power factor measured for the middle

phase (B) bushing is much higher than the other two bushings.
Table 1 shows typical results obtained on three bushings:
results with the windings open circuited (improper technique),
and results once they are properly short circuited. The
improper technique and anomalous results illustrate existence Fig. 3 . Vectorial diagram of voltages and current components
of the symptomatic pattern of two similar high values for the
outside bushings and the highest value for the middle one, bushing in good condition (within the IEC standard limits for
new bushing with 0.7% power factor) . It is important to
Fig. 3 represents a greatly exaggerated vectorial diagram of emphasize this very small angle to demonstrate the minuscule
current components through the bushing insulation, if that value of the in-phase component I,. Test voltage applied is
insulation is represented as a parallel specimen configuration. vectorially represented as V.
Angle 6 between the capacitive component I, and the total test
current IT through the bushing is in the order of 0.4" for a 111. EXPERJMENT
An experiment was conducted in order to detect the cause of
TABLE 1. the subject phenomena. Transformer bushings which exhibited
Measured Percent Power Factor Values the phenomena were tested at several voltages following both
the correct and incorrect procedures of the method shown in

I I Phase A I Phase B I Phase C 1 Fig. 1. The results of the incorrect procedure indicated
significant voltage sensitivity. The percent power-factor values

I Incorrect procedure
I
CASE 1
0.87
I
1.53
I
0.78
I with applied test voltages between 1 and 10 kV changed in the
range from 0.68% to 1.69%, and are shown as the curve A
in Fig. 4. Tests conducted using the correct procedure yielded
Correct procedure 0.26 0.26 0.26 stable results, invariable with voltage, of 0.2% power factor,
and are shown as the flat curve B in the Fig. 4.

Incorrect Procedure 0.82 1.3 0.61 Further tests included voltage amplitude and phase-angle
measurements for all three phases, utilizing the incorrect
Correct Procedure 0.19 0.2 0.2 bushing test technique. Since the test potential was applied to
only one bushing, the other two phases exhibited various
degrees of voltage amplitude and phase discrepancies, as
shown in Fig. 5 and 6. The correct technique, with electrical
 891

1.1 -
1.0 - Anomalous increase in power factor corresponds to an
4.7 - increase in the in-phase current component due to the addition
1.6 -
1.5 - of I,. We will use this information to show the origin of the
1.4 - abnormal readings.
1.3 -

E :::1 The phase relationship between I, and V needs to be defined.

The current 1, is created by the induced voltage V, which lags

!
a
::I:
0.7 -
behind the test voltage V represented in Fig. 3 by 0. The
phase position of the capacitive interference current IE, which
leads its source voltage by 90°, because of its capacitive
0.8 - nature, approaches that of the voltage V and resistive
0.5 - component of the test current, as angle p approaches 90".
0.4 -
D.3 - Flowing through the instrumentation circuitry, a new resistive
Q
0.2 - . I : : : : : : : : + current component I,, is higher than the loss component I,,
0.1 - resulting in an increase in the measured loss angle 6.
D 1 2 3 4 5 6 7 I O 10
Having a similar effect on all three phases, it is obvious that
m w
the middle phase bushing will be affected two-fold. Both
Fig. 4. Bushing percent power factor change with test voltage
A-incorrect, B-comct, C-calculated values for incorrect procedure
adjacent bushings of the outer phases A and C will create a
double impact. In addition, the effect of one outer phase
bushing on the other outer bushing (A phase on C, or C phase
on A) will be diminished due to their distance, and the outer
connection (short circuiting) of all bushings forces the voltage bushings will be subjected to less interference than the center
to be identical in amplitude and phase for all three units. bushing. This explains the difference between phase results
and the existence of an easily observed pattern of high similar
It was observed that a voltage of even greater amplitude than readings for outer bushings and a much higher value for the
the applied test potential was induced in adjacent phases of the middle one.
transformer windings. A 10 kV induced voltage was measured
on bushing A, with a potential of 7 kV applied to bushing C lkeoretical Justification
as shown on Fig. 5.
Without getting involved in the modeling of the transformer
A phase comparison also showed voltage sensitivity. The and the exact parameters of the coupling electrostatic circuit,
phase angle between an applied potential on the bushing under several simplifications and presumptions are introduced:
test and the induced voltage on other bushings increased from
10 to 80 electrical degrees with an applied potential between 1 1. The coupling is purely capacitive.
and 5 kV. The voltage phase difference between bushings A 2. The influence of the remote bushing is neglected.
and B (outer and center) with respect to the potential applied 3. All distributed parameters are lumped.
to the bushing under test is shown in Fig. 6. T h ~ sinduced, 4. Induced interference voltage is represented as a voltage
phase-shifted voltage is shown in the diagram of Fig. 3, as source at the end of the winding.
the V,.

Explanation The source of the higher induced voltage in the adjacent

phases (remote end of the distributed LC circuit) is explained
Measured power factor is obtained as the ratio between the in many text books dealing with the Ferranti Effect and will
in-phase current component I, and the total test current IT. In not be addressed in this paper. Values induced and measured
the case of incorrect procedure, another current component is will be simply represented as a dependent voltage source
added; the interference current I,. This current, through the V,=f(V). Also, the phase angle between induced voltage and
capacitive coupling between adjacent bushings, flows through the test source voltage will be represented as a function
the measurement circuitry and transformer windings, p= f(V).
superimposed to the test current. In this instance the measured
current becomes I, (vectorial sum of I, and 1 3 while its in- The resistive component of the interference current IREcan be
phase component is I,, as per (1) and Fig. 3. obtained as:
892

XI-

M -

UI-

40-

3l-

a-

L
t I '
t 2 3 4 5
TwwLrK€w TEsr WxTL(E w
Fig. 5. Voltage induced on adjacent bushing Pi.6. Phase angle behveen center and outer bushings

IRE = IEcos (90-p) (2) IV. CONCLUSION

The interference current I, can be approximated as: A discrepancy in the measured dielectric-loss values of the
bushings mounted in three-phase transformers was attributed
IE = v, * ac, (3) to interference voltage induced within the transformer. A
phase shift of the voltages induced on the associated windings
where C, is the capacitance of the coupling between adjacent creates a capacitively-coupled current through the insulation
phases. (oil in the case of an oil-filled transformer), between adjacent
phases, shown as I, in Fig. 1, and vectorially in Fig. 3.
Not knowing the value of this capacitive coupling, shown as
C, on Fig. 1, an arbitrary value of 2.5 pF was used to allow The results of the three bushings' loss measurements form a
a comparison of the shape of the measured and calculated pattern of two similar but high values, with the third
power factors as proof for the proposed explanation. significantly higher. This pattern immediately indicates the
presence of such a phenomena and warrants a correction to
Using the correction factors K=f(V), another function of test the procedure.
voltage, which are built into the measuring instrumentation to
convert a measured resistive current component to 10 kV An amendment to the test technique as defined by standard
normalized dielectric losses, the power factor was calculated procedure, where all bushings are electrically connected -
using (2) and (3) and plotted in Fig. 4, as curve C. It is windings are short circuited for the test, eliminates the
obvious that there is similarity between measured and undesired effect, restores the electrostatic field in the domain
calculated values. Considering all the approximations and of the measurement circuitry, and allows for obtaining correct
simplifications made in the calculation, the comparison of the results .
values appears to be successful.
v. ACKNOWLEDGEMENT
Unfortunately, the experiment instrumentation, namely the
phase-angle meter, could not perform above 5 kV test The authors wish to thank the Orange and Rockland
potential. For that reason the end of curve C in Fig. 4, Utilities Co. for allowing us to perform this experiment, as
between 6 and 8 kV, is defined by extrapolation of the well as assisting in the organization and performance of these
measured values for phase angle beyond the 5 kV range. measurements.

Modification of the test procedure, which is included in the VI. REFERENCES:

standard test procedure, eliminates the presence of the
unwanted phenomena. By short circuiting all three bushings 1. A.L. Rickley and R.E. Clark, "Application and
for these tests the potential and phase difference is eliminated, Significance of Ungrounded-Specimen Tests, " Minutes of the
thus no interference is present. 1960 Doble Client Conference, 27AIC60, p. 3-201.
 893

2. E.H. Povey, "Testing Bushings Equipped With Special Mr. Manifase holds a Bachelor of Science degree in electrical
Test Electrodes," Minutes of the 1948 Doble Client engineering from Northeastern University in Boston,
Conference, 51AIC84, p. 3-101. Massachusetts.

3. "Capacitance and Dissipation Factor (tan 6) Test on

Condenser Bushings in Field, Product information,
SECOMXU 1990-10-18, ABB Components AB, Ludvika,
Sweden 1990.

4. IEEE Guide for Loading Power Apparatus Bushings,

(37.19.01 Performance Characteristics and Dimensions.

5. IEC Standard Publication 137, "Bushings for Alternating

Voltages Above lo00 V," Third Edition, 1984.

6. H.A.Ruggles, "Teardown inspection of four high power

factor General Electric Type U bushings," Minutes of the
1983 Doble Conference, 50AIC83, p. 4-801.

7. J.L. Puri, "Power factor of Westinghouse Type "0 Plus"

condenser bushings, Minutes of the I983 Doble Conference,
50AIC83, p. 4-301

8. H.K. Xie, and K.C. Kao, "Further Studies of Anomalous

Phenomena in Dielectric Loss Measurements Using a Three-
Electrode System, " IEEE Transaction on Electrical Insulation,
vol. EI-21 no.1, February 1986.

9. "The Doble Type M2H 10 kV Portable Insulation Test

Set", Instruction Manual, Doble Engineering 1992.

VII. BIOGRAPHY

Raka Levi, (SM) received his Diploma of Engineering degree

in Electric Power Engineering in 1978 at the Electrotechnical
School of the University of Belgrade, Yugoslavia, and Master
of Engineering degree in Electrical Power Engineering at
Rensselaer Polytechnic Institute, Troy, NY in 1979.

Since 1986 he has worked for Doble Engineering Company as

Principal Engineer, consulting on electric power equipment
insulation testing and maintenance.

Stan Manifase has worked for the Doble Engineering Co.

since 1979, and is currently the Assistant Manager of the
Annual International Conference of Doble Clients. Prior to
h s work with the Conference and the technical committees
sponsored by Doble Engineering, Mr. Manifase was a Field
and Principal Engineer, and served several years as the
secretary of the Doble Client Committee on Circuit Breakers
and Bushings.
894
DISCUSSION voltage rise when capacitive current flows through an
L. 0 . WAGENAAR AND J. M. SCHNEIDER, AEP Service Corp., inductance". We must point out that effect caused by the
Columbus, OH. The authors have presented an interesting
paper which gives an empirical explanation for erroneous transformer has a lot of similarities, but also some
dissipation factor readings on bushings tested within fundamental differences or modifications, to the described
a transformer using an incorrect test procedure. The
incorrect procedure yields dissipation factors which are Ferranti effect. In a case of overhead lines, there is no
two to eight times higher than actual. The paper transformer action and/or non-linearities of the iron cores. In
attributes this to the Ferranti effect, which results
in a voltage rise when capacitive current flows through the case of the transformer model, the capacitive current
an inductance. In this case, the capacitances and flowing thorough the inductance of a transformer is creating
inductances appear to be those associated with the
transformer windings. a flux through the iron core which is non-linear, while the
Some of the measurements reported in the paper are not overhead line has a linear inductor, the air. Also, the
consistent with two characteristics of the Ferranti equivalent circuit comprises windings of two phases in series,
effect. First, the measurements shown in Fig. 5 of the
paper indicate a non-linear relationship between the where the voltage is also induced on the adjacent phase due to
magnitudes of induced voltage vs. applied voltage, the transformation effect. In this case the turns ratio is 1:1,
whereas the Ferranti effect would result in a linear
relationship [All. Second, Fig. 6 of the paper shows but the flux distribution depends on the configuration of the
the phase angle of the induced voltage to be dependent
on applied voltage, whereas voltages induced by the
magnetic core - distribution between several legs in parallel.
Ferranti effect would be independent of applied voltage. For these reasons, one would expect non-linear behavior of the
Will authors explain these inconsistencies?
induced voltage on adjacent phases both in magnitude and
It would also be useful to know several ratings and phase relationship.
parameters pertaining to the transformer which was
tested. These include the voltage and kVA ratings, the
connections of the high and low voltage windings, the With regard to the discussers request for clarification of the
capacitances of the bushings, the transformer
inductances (or impedances), capacitances and equivalent inconsistencies in the Ferranti effect explanation, we have
circuit topology necessary to duplicate the measured approached transformer manufacturers for information on
phenomena.
research done in the past, or any existing data supporting these
Finally, we are surprised that the authors referenced findings, but have had no success. This is a very apparatus
IEC Std. 137, instead of IEEE Std. C57.19.01 [A2], for
the standard limit of dissipation factor of new specific problem, and may require further studies, modeling
bushings. The former has a 0.7% limit, whereas the
latter has a 0.55% limit. It is our opinion that both of transformers, and an explanation in a follow up report to
values are too high, but we would have expected to see the IEEE technical audience. Since our paper was intended as
the IEEE limit referenced. Modern design and
manufacturing techniques have enabled bushing a practical presentation of the measurement error, we would
dissipation factors to be in the range of 0.2 to 0 . 4 % , refrain from this examination and invite specialists in the
as indicated by the two examples given in the paper, and
low dissipation factors become increasingly important transformer design field, who are better qualified to explain
in higher voltage bushings when thermal stability this phenomena, to offer their solution and interpretation of the
becomes a factor [ A 3 ] .
origin of this occurrence that we measured in the field.

References For further studies, we have enclosed the data from one
R. D. Begamudre, Extra High Voltage A.C.
transformer in question, in the Table C1.
Transmission Engineering, (John Wiley & Sons,
1986), p. 321.
The opinion of the authors is supportive of the discussers
ANSI/IEEE Std. C57.19.01-1991, IEEE Standard vision that the dissipation factor standard limits should be
Performance Characteristics and Dimensions for
Outdoor Apparatus Bushings. much lower than presently set. It is very common to see new
bushings with paper oil insulating system, tested in the factory
L. B. Wagenaar, "The Significance of Thermal
Stability in EHV Bushings and Current having less than 0.3 % dissipation factor.
Transformers," 1994 Doble Clients Conference Paper.
It is our opinion, however, that the difference between IEC
Manuscript received August 12, 1994. and IEEE standard is not very significant. To justify this
position, let us convey our experience with 50 Hz and 60 Hz
dissipation factor measurements using our new dual frequency
instrument [Cl]. We have found that bushings behave during
the dissipation factor test as an ideal parallel specimen, (ie. the
equivalent schematic diagram could be represented as a
capacitor and resistor in parallel). In this case we have two
current components, capacitive and resistive flowing through
the specimen. A comparison of values obtained at 50 and
Raka Levi and Stan Manifase: 60 Hz on the same bushing shows that the capacitive
component of the test current increases with increase in
The Ferranti effect is a familiar phenomena in open or slightly frequency while the resistive component is constant. This
loaded overhead lines. As the discussors explained, it is "a observation confirms our parallel model.
 895

Table C1 In such case, dissipation factor at 60 Hz has value 0.833 times

The transformer and bushing parameters the 50 Hz value (0.833 = 50160). Since the IEC standards
reflect mainly 50 HZ requirements and IEEE are based on
60 Hz, if the limit at 50 Hz (the IEC standard of 0.7%) is
Characteristics Value Units converted to the 60 Hz base - it would yield dissipation factor
of 0.58 % which is not significantly different from the 0.55 R
Transformer voltage 138113.2 kVlkV
limit of the IEEE standard.
Transformer rating 35 MVA
Reference:
HV winding 3,180 PF [Cl] Brooks, P.R. and Kopaczynski, D.J.: "The Doble
capacitance to M4000, A New-Generation Insulation-Tests Instrument",
ground Minutes of the Fifty-Ninth Annual International Conference of
Doble Clients, Boston, April 1992.
Winding connection DeltalWye
Bushings' A 311 Manuscript received October 25, 1994.
PF
capacitances B 305 PF
C 302 PF
Transformer phase A 735 KQ
impedance at 10 kV B 625 Kdl
C 641 KQ