Beruflich Dokumente
Kultur Dokumente
5, OCTOBER 2015
Abstract—Frequency-response measurement offers by far Tank current actually happens to be the current to the tank
the most sensitive known technique to detect any mechanical through all of the ground capacitances of the winding and,
deformation in transformer windings. By default, during this hence, for convenience, this current (i.e., the shunt current)
measurement, the neutral or line current is considered as the
winding response. The overall sensitivity to detect a mechanical henceforth will be referred to as the tank current of the winding
change depends not only on the response quantity and the terminal in this paper, as shown in Fig. 1.
connections chosen, but, also on the arrangement of poles and Frequency-response analysis (often referred to as FRA or
zeros of the system function. Some pole-zero arrangements are SFRA) of the transformer winding introduced around the 1980s
inherently better suited to reveal mechanical changes. Because ([5], [6]) still continues to be a topic of interest [7]–[9]. One
of this, it is imperative to examine whether neutral current is
indeed the best response quantity to consider or if there exists basic task of FRA is to identify winding damage immediately
any other quantity, say, for example, total shunt current of the after it has occurred. So, FRA sensitivity in this regard is obvi-
winding (also referred to as the tank current), to accomplish this ously very crucial. The higher the sensitivity, the possibility of
task better. With this motivation, this paper compares the reso- detecting/identifying winding damage or deformation increases.
nance behavior of neutral and tank current. Based on analytical Once the damage is detected, the diagnosis process can focus on
derivations and actual experiments, it emerges that for detecting
simulated mechanical changes, the tank current seems to be a the next goal of assessing its severity and its localization along
more competent alternative compared to neutral current. It offers the length of the winding, which can eventually provide valu-
enhanced detection sensitivity, especially for interleaved windings. able inputs to schedule maintenance work.
The possible reasons to explain why it is so are presented. The As far as the sensitivity of the FRA measurement is con-
author believes that this finding needs reinforcement by field trials cerned, all possible terminal measurements on the winding, line
to ascertain its viability.
current, neutral current, transferred surge, neutral-end voltage,
Index Terms—Frequency response, resonance, tank current, etc. have been examined [10], [11]. Based on experiments on
transformer winding.
actual transformers, the best terminal connection for the tested
and nontested windings has been identified for achieving high
I. INTRODUCTION AND BACKGROUND sensitivity [12], [13]. But to the best of the author's knowledge,
measurement of tank current has not been examined as an op-
tion for use in FRA measurements nor has the sensitivity it
T HE NEUTRAL current method was proposed by Ha-
genguth [1] in 1944, as a sensitive method to detect
dielectric faults occurring during a high-voltage (HV) impulse
can afford been analyzed so far. A careful perusal of the very
early literature [3], [14] on high-voltage (HV) impulse testing
of transformers reveals that even during those times, there was
test on transformers. Ever since then, it has come to be accepted
considerable debate on which quantity would afford better sen-
as the default quantity to be measured, and more important, is
sitivity and localization of dielectric faults occurring during the
the most sensitive means to detect faults [2], [3]. Later, there
full-wave and chopped-wave tests. It was concluded then that no
was also a suggestion to measure tank current as an alternative
single quantity could be identified as the best for the task, and
[3], [4]. But based on practical experience, it emerges that neu-
simultaneously be suitable for all types of transformer wind-
tral current gained popularity, thereby leading to its widespread
ings [14]. Since contributions to the tank current are distributed
acceptance. Perhaps one can attribute the aforementioned fact
throughout the entire length of a winding, it intuitively seems
as a possible reason to explain why neutral current was an
that tank current would contain identical winding-related infor-
automatic quantity of choice (as the output response) even in
mation similar to any other measured quantity. Hence, it would
low-voltage frequency-response analysis (FRA) measurements.
be interesting to consider the feasibility of tank current for use
in the FRA method and investigate the following matters:
Manuscript received October 26, 2014; revised February 04, 2015; accepted
• resonance behavior of tank current;
April 25, 2015. Date of publication April 29, 2015; date of current version
September 21, 2015. Paper no. TPWRD-01304-2014. • analysis of its sensitivity to fault conditions.
The author is with Corporate Research, ABB AB, Vasteras 72178, Sweden To achieve this objective, the analytical expressions for the
(e-mail: saurav.pramanik@se.abb.com).
neutral and tank current are determined, from which resonance
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. behavior is predicted. The ability of both these currents to
Digital Object Identifier 10.1109/TPWRD.2015.2427772 detect changes in windings is estimated by introducing a small
0885-8977 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
PRAMANIK: RESONANCE BEHAVIOR AND SENSITIVITY TO DETECT MECHANICAL CHANGE 2277
Fig. 1. Ladder network model of the transformer winding showing line , neutral , and tank currents. (Note: For identical sections
and . Furthermore, and so on).
where,
Fig. 4. Computed tank current magnitude response of a five-section ladder net- and
work model: (a) Healthy. (b) Faulty.
(2)
Therefore
3) Tank current, as an outcome of the fault, exhibits a new
pair of resonance peaks in the magnitude response (see
Fig. 4); a feature which is not observable in the neutral cur-
rent (Fig. 3). Besides this, existing peaks also experience a
change in position and magnitude.
In summary, it emerges that the tank current magnitude re-
sponse of a healthy model does not exhibit all of the resonances
in the magnitude response, but interestingly, most of them (if
not all) emerge distinctly as a consequence of a fault intro- (3)
duced into the model. This particular feature, if possible to gen-
eralize, would definitely afford enhanced detection sensitivity From the above, it is observed that the line and neutral current
and, hence, the tank current measurement would perhaps be expressions have four terms as follows.
a more competent alternative for fault detection compared to • The first term is a constant term, comprised of only and
the neutral current. Therefore, in the present context, it would , and devoid of “s”.
be worth investigating the following peculiarities of the tank • The second term is a purely inductive term (i.e., equivalent
current: inductance ).
1) Is the absence of alternate resonance peaks (i.e., dormant • The last two terms are functions of all inductances and
or latent resonances) in tank current magnitude response a capacitances.
generic property observable in any uniformly wound trans- The last two terms are actually the oscillating components
former winding? If so, why? and describe the resonance behavior of the winding. Roots of
2) Do these dormant/latent resonances always reappear in their denominator determine the resonance/natural frequencies
the magnitude response whenever there is a fault in the of the winding. An important observation in this regard is that
winding? the second and fourth term in the line current expression are
In order to answer these generic questions, it will be beneficial exactly equal to the corresponding terms in the neutral current
to switch to analytical derivations and this is continued in the expression and, hence, they are cancelled out while subtracting
next section. one from the other to compute the tank current expression. This
PRAMANIK: RESONANCE BEHAVIOR AND SENSITIVITY TO DETECT MECHANICAL CHANGE 2279
Fig. 5. Neutral and tank current magnitude response and their corresponding pole-zero maps. (a) Pole-zero map and (b) magnitude response. (Note: Only one pole
or zero for each complex conjugate pair has been included here for the reader's convenience. The numbers identify the resonance peaks in the right-hand subplot
and their corresponding poles in the left-hand subplot.)
TABLE I Next, the faulty case was examined with the corresponding
POLE-ZEROS OF TANK CURRENT OF THE
new set of poles and zeros (for tank current) along with the cor-
FIVE-SECTION LADDER NETWORK
responding magnitude response (see Fig. 6). This figure depicts
how the poles and zeros in every alternate pair are separated
(also evident in Table I), and now manifest themselves as new
resonance peaks in the magnitude response. The occurrence of
these new resonance peaks implies that the tank current is more
sensitive to faults in comparison to the neutral current which ex-
periences only a change or shift in existing peaks. When com-
paring frequency responses, the emergence of new peak/trough
is obviously a more significant event than a mere shift in existing
is the reason why alternate natural frequencies are not excited peaks/troughs. Now, the next obvious question which arises is:
in the magnitude response and essentially remain dormant or Is this observation of tank current resonance behavior generic
latent. This fact is also verified analytically up to a six-section for any kind of fault in the winding? If so, then a fault condi-
model and numerically up to 10-section model and found to be tion can be inferred by merely counting only the new resonance
consistent. peaks appearing in the subsequently measured magnitude re-
This fact is also true for a lossy ladder network, but the com- sponse in comparison with the reference or healthy FRA. If not,
putation does not readily permit the factorization of the oscil- then are there any specific conditions which might prevent the
lating components into two parts, and the expressions also run emergence of new resonance peaks in tank current magnitude
into several pages. Hence, they are not included here. But the response?
same has been verified numerically with the 5-section ladder To probe this matter further, a three-section ladder network
network. Poles and zeros for both neutral and tank current re- model with all distinct parameters was considered and the
sponse functions were computed and are plotted in Fig. 5 to- analytical expressions in (1)–(3) were recomputed. The expres-
gether with their corresponding magnitude responses. sions are very long and so are not included here, but they reveal
Numerical values of the poles zeros are also listed in Table I. the fact that there exist few particular fault-condition scenarios
The table clearly shows that for the healthy condition, every where pole-zero cancellations for alternate resonant peaks
alternate pair of poles and zeros (shaded) in the tank current occur, if and only if, the following conditions are satisfied:
response function are exactly equal to each other and, hence, • and/or;
are cancelled out. This explains why only alternate resonance • and/or;
peaks are missing in the tank current magnitude response. • .
2280 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 30, NO. 5, OCTOBER 2015
Fig. 6. Simulation results of (a) the pole-zero map and (b) tank current magnitude response for the five-section ladder network with a fault.
The aforementioned conditions suggest that the introduced ladder networks, each interconnected to the other by appropriate
faults are identical and symmetrically located (i.e., equidistant inter-winding capacitances, reveals identical tank current be-
from the line end and neutral end, respectively) and are of equal havior. This indicates that the tank current feature that is being
magnitude. Under this hypothetical condition (which is easy to dealt with here is not a freak phenomenon, but is a salient char-
implement in simulations), it emerges that the tank current re- acteristic of the winding, and it is worthy of further studies and
sponse fails to exhibit emergence of new resonance peaks, as ob- investigation. Perhaps, in the future, it will emerge as a better
served previously. Only a shift of peaks/troughs is observable, candidate for fault diagnostics, at least, in some select types
and that is very similar to what is revealed by the neutral cur- of windings, if not all. This can only be answered with more
rent response. The author hastens to add that encountering such studies.
a scenario in practice, in all likelihood, is extremely remote, if
not impossible, and so it should not be construed as a limitation
of the detection abilities of the tank current method. Moreover, IV. EXPERIMENTS ON ACTUAL WINDING
all of the aforementioned conditions are further rigidly bound
by another condition, that is, , failing which will re- A. Interleaved Disk Winding
sult in the emergence of new resonance peaks in tank-current
magnitude response. However, if the changes are pertinent to To conduct the experiment initially, a single, isolated,
only one section, then all types of changes will result in per- air-core, fully interleaved-disk winding was considered. The
turbation to the tank-current magnitude response, except, winding had a total of 16 disks with 10 turns per disk. Height,
and/or in the middle section which is further restricted by inner and outer diameters are 225 mm, 260 mm and 350 mm
the condition . Therefore, in final summary, these respectively. Two aluminum foils were concentrically placed
particular fault condition scenarios could be of only theoretical inside and outside the winding to represent the ground planes
interest but are extremely remote in practice. Consequences of similar to core and tank respectively, in an actual transformer.
all the aforementioned conditions were individually verified by Finally, both the inner and outer Al-sheets were connected
performing several simulations and analytical calculations for together to represent a common ground plane. If the core is
several models. These findings were experimentally validated present it may alter the flux distribution ([15], [16]) but is
by conducting measurements on actual transformer windings unlikely to introduce any non-uniformity to the distribution of
by introducing simulated mechanical faults, and the pertinent winding parameters. This means that the core may introduce
results are reported in the following section. a change in inductances and capacitances but the changes will
be uniform throughout the length of the winding. Thus, all the
A. Validity for the Three-Phase Scenario sections of the equivalent model will still remain identical to
This unique resonance behavior of the tank current discussed each other (as in Fig. 1) and, hence, the basic findings of the
before for a single isolated winding representation has been present work will not be altered. Input voltage, neutral current
found to remain unchanged even when a three-phase winding and tank current magnitude responses (i.e., healthy responses)
model is considered. Simulation studies conducted on a three- were measured using sweep frequency method by employing
phase winding modeled as three individual mutually coupled the following setup.
PRAMANIK: RESONANCE BEHAVIOR AND SENSITIVITY TO DETECT MECHANICAL CHANGE 2281
V. CONCLUSIONS
This paper highlighted a unique feature of tank current res-
onance behavior of the transformer winding and its possible
application to improve the fault detection sensitivity in FRA
measurements. This was demonstrated by simulation results and
measurement on actual windings. The behavior was explained
using the analytical expressions for both neutral current and tank
current response functions, followed by their pole-zero distri-
Fig. 12. Measured tank current magnitude response of a continuous-disk
winding. (a) Healthy. (b) Faulty. butions. The difference between the tank current and neutral
current resonance behavior actually lies in the difference be-
tween their pole-zero distributions and this paper highlighted
observed in the interleaved disk winding [Fig. 10(b)]. As an- this particular feature to explain the uniqueness of tank current
ticipated, alternate resonance peaks are missing in the tank cur- resonance behavior. Employing this unique tank current reso-
rent magnitude response in comparison to the neutral current nance behavior, one can easily identify any change in trans-
magnitude response. A small kink is observable in between the former winding in practice by counting only the additional or
nd and th resonance peaks [as numbered in Fig. 10(b)] and new resonance peaks in the magnitude response that emerge
this could be explained due to the inherent nonuniformity in during a subsequent FRA measurement. This seems to be the
the actual winding. Both of the responses were measured again usefulness of tank current measurement over the conventional
after introducing a fault to the winding by adding external ca- neutral current measurement, insofar as the interleaved winding
pacitances, that is, a 0.49-nF capacitor between disks 20 and is concerned. This feature appears to be beneficial, especially for
24 as well as 0.98- and 0.18-nF capacitors between the ground interleaved windings, since they exhibit very few natural fre-
and disks 13 and 21, respectively. The measured responses are quencies in comparison with continuous-disk windings. There-
plotted in Figs. 11 and 12. fore, in summary, the overall sensitivity afforded by the tank
PRAMANIK: RESONANCE BEHAVIOR AND SENSITIVITY TO DETECT MECHANICAL CHANGE 2283
current measurement is found to be better than the conventional [7] D. A. K. Pham, T. M. T. Pham, H. Borsi, and E. Gockenbach, “A new
neutral current measurement to detect a fault in the transformer diagnostic method to support standard frequency response analysis as-
sessments for diagnostics of transformer winding mechanical failures,”
winding. The author opines that these results need to be rein- IEEE Elect. Insul. Mag., vol. 30, no. 2, pp. 34–41, Mar./Apr. 2014.
forced by extensive measurements, and the possibility of con- [8] J. C. G. Arispe and E. E. Mombello, “Detection of failures within trans-
sidering tank current in FRA measurements must be explored formers by FRA using multiresolution decomposition,” IEEE Trans.
Power Del., vol. 29, no. 3, pp. 1127–1137, Jun. 2014.
by power utilities and industry alike. [9] Guide for the Application and Interpretation of Frequency Response
Analysis for oil Immersed Transformers, IEEE C57.149-2012, 2012.
[10] K. Ragavan and L. Satish, “Construction of physically realizable
driving-point function from measured frequency response data on a
ACKNOWLEDGMENT model winding,” IEEE Trans. Power Del., vol. 23, no. 2, pp. 760–766,
Apr. 2008.
[11] S. Pramanik and L. Satish, “Enhancing physical resolution of con-
The author would like to thank to Prof. L. Satish, HV Labo- structed ladder network for an interleaved winding,” in Proc. IEEE Int.
ratory, Department of Electrical Engineering, IISc., Bangalore, Conf., Kolkata, India, Dec. 2013, pp. 60–65.
India, for his valuable comments and suggestions during the [12] L. Satish and A. Saravanakumar, “Identification of terminal connection
and system function for sensitive frequency response measurement on
preparation of the manuscript. transformers,” IEEE Trans. Power Del., vol. 23, no. 2, pp. 742–750,
Apr. 2008.
[13] J. A. S. B. Jayasinghe, Z. D. Wang, P. N. Jarman, and A. W. Darwin,
“Winding movement in power transformers: a comparison of FRA
REFERENCES measurement connection methods,” IEEE Trans. Dielectr. Elect.
Insul., vol. 13, no. 6, pp. 1342–1349, Dec. 2006.
[1] J. H. Hagenguth, “Progress in impulse testing of transformers,” Elect. [14] “CIGRE report on the discussions of transformers group 12,” in Proc.
Eng., vol. 63, no. 12, pp. 999–1005, Dec. 1944. CIGRE Conf, May 31, 1952, pp. 147–203.
[2] H. C. Stewart and J. E. Holcomb, “Impulse failure detection methods [15] E. E. Mombello, “Impedances for the calculation of electromagnetic
as applied to distribution transformers,” Elect. Eng., vol. 64, no. 9, pp. transients within transformers,” IEEE Trans. Power Del., vol. 17, no.
640–644, Sep. 1945. 2, pp. 479–488, Apr. 2002.
[3] E. Stenkvist, “Study of fault detection and failure location during surge [16] K. G. N. B. Abeywickrama, A. D. Podoltsev, Y. V. Serdyuk, and S. M.
testing of transformers,” CIGRE rep. 12–129, 1952, pp. 1–31. Gubanski, “Infuence of core characteristics on inductance calculations
[4] C. K. Roy and J. R. Biswas, “Studies on impulse behavior of a trans- for modeling of power transformers,” in Proc. 1st Int. Conf. Ind. Inf.
former winding with simulated faults by analogue modeling,” Proc. Syst., Sri Lanka, Aug. 2006, pp. 24–29.
Inst. Elect. Eng., Gen. Transm. Distrib, vol. 114, no. 5, pp. 401–412,
Sep. 1994. Saurav Pramanik was born in 1984. He received the B.E. degree in electrical
[5] R. Maleswski and B. Poulin, “Impulse testing of power transformers engineering from the Jadavpur University, Jadavpur, Kolkata, India, in 2006 and
using transfer function method,” IEEE Trans. Power Del., vol. 3, no. the M.Sc.(Engg.) and Ph.D. degrees in electrical engineering from the Indian
2, pp. 476–489, Apr. 1988. Institute of Science, Bangalore, India, in 2010 and 2013, respectively.
[6] E. E. Dick and C. C. Erven, “Transformer diagnostic testing via fre- Since 2014, he has been with Corporate Research, ABB, Vasteras, Sweden.
quency response analysis,” IEEE Trans. Power App. Syst., vol. PAS-97, His research areas include condition monitoring and diagnostics of high-voltage
no. 6, pp. 2144–2153, Nov./Dec. 1978. power apparatus and magnetics of power transformers.