Sei sulla pagina 1di 5

0609-137-Mep228 10/11/06 16:10 Page 30

WGA2.26
REPORT

Mechanical condition assessment


of transformer windings using
Frequency Response Analysis (FRA)

I n t ro d u c t i o n ✓ March 2006: Northfleet,


Members: Kent (UK)
In recent years, there Patrick PICHER, Convenor (Canada) ✓ August 2006: Paris
have been many papers pub- John LAPWORTH, TF-1 leader (United Kingdom)
lished on the subject of so- Tim NOONAN, TF-2 leader (Ireland) This report presents the
called “Transfer Function” or Jochen CHRISTIAN, TF-3 leader (Germany) WG achievements on the
“Frequency Response Anal- comparison of measurement
ysis (FRA)” methods for Micheal E. ALPATOV (Russia) techniques, the description of
detecting winding move-
Dierk BORMANN (Sweden) typical FRA responses and the
Richard BREYTENBACH (South Africa) modelling to support inter-
ment in power transformers.
Peter DICK (Canada) pretation. A brief description
Essentially, these techniques
Alexandr DROBISHEVSKY (Russia) of the final report content is
involve injecting a high fre-
Lionel DUMBRVA (Romania) presented in conclusion.
quency signal at one termi-
Robin FISHER (Switzerland)
nal of a transformer and
Hans Kristan HOIDALEN (Norway)
measuring the response of
Paul JARMAN (United Kingdom) Experimental
the winding to that signal. Istvan KISPAL (Hungary) comparison of
The CIGRE 2003 Collo- Micheal KRUEGER (Austria) m e a s u re m e n t
quium in Merida, Mexico Thomas LEIBFRIED (Germany) techniques
[1], on Diagnostic Methods Ryszard MALEWSKI (Poland)
and Interpretation, con- Rodrigo OCON (Mexico) Objectives
cluded that such techniques Ernesto PEREZ (Spain)
were widely used and much Johannes RICKMANN (United States) In order to make a start
more sensitive than the tra- Yukiyasu SHIRASAKA (Japan) on standardisation of mea-
ditional and internationally Stephan TENBOHLEN (Germany) surement techniques, a meet-
accepted method of impe- Martin TIBERG (Switzerland) ing was arranged at the
dance measurements, but Zhongdong WANG (United Kingdom) Siemens transformer works
that work was required on Peter WERELIUS (Sweden) in Nuremberg in January
standardisation and inter- Rafal ZALESKI (Poland) 2005 (Figure 1), with the
pretation. intention of:

Accordingly, in 2004, CIGRE SC A2 decided to set • characterising differences in measurement tech-


up a new Working Group on the subject. Experts from niques (impulse, swept frequency, measuring
SC D1 have participated to the WG A2.26. The WG was impedance, test leads setup, etc.)
set up under the following organisation: • comparing types of FRA measurements on a new
✓ Task Force 1: Introduction to FRA transformer (“end to end”,“inter-winding”,“open”
✓ Task Force 2: Measurement Techniques & “short-circuit” tests, etc.)
✓ Task Force 3: Interpretation • identifying good FRA practices and define practi-
cal limitations of FRA measurements
The following meetings have been held:
✓ June 2004: Montreal Summary of test programme
✓ August 2004: Paris
✓ January 2005: Nuremberg A total of nine different measuring systems (labelled
✓ June 2005: Moscow A–I in this report) were used, including both impulse

30 E L E C T R A N° 228 - Octobre 2006


0609-137-Mep228 10/11/06 16:10 Page 31

WGA2.26
REPORT

Fig. 1: Nuremberg attendees and test object

and swept frequency methods and four commercially The impulse methods were unable to reproduce the
available equipments. low frequency response because of digitizers set to acquire
the higher frequencies. Some swept frequency methods
The test object was a single phase 420/21 kV, also did not have sufficient dynamic range to repro-
266 MVA generator transformer designed for transport duce the typical 90 dB minimum obtained with a
via the German railway system, with long horizontally 50-ohms measuring impedance.
installed HV and neutral bushings.

A series of tests was devised to compare measured


responses between the various equipments, to assess
repeatability of measurements and to estimate the use-
fulness of different measurements.

Main results and conclusions

The main results and conclusions from the compar-


ative testing were as follows.

All test equipments produced essentially the same


measured responses for the test object over a mid range
of frequencies from about 10 kHz to 500 kHz, but there
were significant differences above and below this range
(Figure 2). Fig. 2: Measured HV responses

No. 228 - October 2006 ELECTRA 31


0609-137-Mep228 10/11/06 16:10 Page 32

WGA2.26
REPORT

In Figure 3, the repeatability of each system is eval-


uated by repetition of the measurement on the HV wind-
ing after a series of tests. Differences at high frequencies
appeared to be due to differences in test lead setups, in
particular different practices used for grounding the test
leads (Figure 4), and this appeared to affect the repeata-
bility achievable above about 500 kHz. A subsequent
trial was arranged in Northfleet, Kent (UK), to investi-
gate the effects of variations of measurement practices,
e.g., ground loops, use of braids, etc.

Fig. 5: LV1 and LV2 winding responses


(linear frequency scale up to 2 MHz)

As regards the value of different types of measure-


ments, ‘inter-winding’ responses – both capacitive and
inductive (turns ratio) transfers - were found to be insen-
sitive to tap position (Figure 6), so are expected to be less
sensitive to winding movement compared to ‘end to end’
measurements.

Fig. 3: HV winding responses: repeatability test


(linear frequency scale up to 2 MHz)

Fig. 6: Capacitive inter-winding response


at tap Max and Max-1.

Fig. 4: Test lead connections on horizontal HV “Short-circuit” measurements (across the HV wind-
bushing with a large loop in ground connection ing with the LVs shorted, Figure 7) gave simpler responses
below about 20 kHz. Such measurements allow the influ-
Repeatability up to 2 MHz was not a problem for ence of the core to be removed, and are believed to be
measurements on the LV windings with their short bush-
ings (Figure 5). Subtle differences between different mea-
suring systems were suspected to be due to the variation
of input impedance.

Whereas measurement impedance affected the


detailed form of the resonances, there was no evidence
that there was any benefit to any particular impedance
value. More research is needed to determine if the mea-
surement impedance has an influence on the sensitiv-
ity to winding displacement. For the best consistency
between two different sets of results, the same measur-
Fig. 7: Open (red) and short-circuit (blue)
ing impedance should be used. measurements.

32 E L E C T R A N° 228 - Octobre 2006


0609-137-Mep228 10/11/06 16:10 Page 33

WGA2.26
REPORT

capable of detecting circulating current problems between physical dimensions of a particular design, it is observed
sub-conductors of a turn. The fact that open and short- that the same general form is obtained for windings of
circuit measurements are not significantly different above the same type. It is useful to be able to recognise such
20 kHz suggests that at these frequencies, leakage fluxes “typical” responses. It must be kept in mind that there
determining winding impedances are essentially the same. responses were obtained using an “end-to-end” measure-
ment with 50-Ω measuring impedance.

Second FRA Wo r k s h o p For HV windings, particularly those employing tech-


niques such as interleaving to smooth the impulse dis-
Following the test programme carried out at the tribution at the line end, the generic form is an overall
Siemens factory in Nuremberg, a second test session was reducing attenuation with frequency after the first min-
organized in a substation of National Grid, UK, in North- imum until a maximum close to 0 dB at about 1 MHz
fleet, to investigate the effects of variations of cabling (Figure 9), with superimposed resonances of varying
practices and evaluate the value of standardization. The degrees. Above about 500 kHz there is usually less evi-
tests were performed using several instruments on an dence of resonances and internal leads have a significant
auto-transformer 400/275kV, 500MVA. Figure 8 shows influence and can introduce differences between phases.

Fig. 9: Typical HV winding response


(Fully interleaved winding, single phase unit)

LV windings, particularly the simple spiral windings


of large generator transformers, usually have a series of
well defined resonances up to 2 MHz (Figure 10).

Fig. 8: FRA measurements obtained with


various cabling practices (top) and after
standardization (bottom)
the measurements obtained using various lengths and
layouts of the extension grounding leads, compared to a
standardized practice using a short braid. It was demon-
strated that standardized measurement practices extend
Fig. 10: Typical LV winding response
the repeatability to higher frequencies.
(Generator transformer, single layer wound)

In addition to the general form of the response being


Typical FRA responses determined by the type of winding involved, connec-
tion arrangements can also have a significant influence
Although the detailed form of winding responses on the form of the response over a limited range of fre-
depends on the winding arrangement, geometry and quencies.

No. 228 - October 2006 ELECTRA 33


0609-137-Mep228 10/11/06 16:10 Page 34

WGA2.26
REPORT

The model has been used to predict the changes in


responses caused by various winding deformations.
Radial and axial displacements both resulted in frequency
shifts of resonances.
The model has been useful in increasing the under-
standing of what determines the various features of fre-
quency responses. An important discovery has been that
for “end to end” measurements, the main features of a
winding response are determined primarily by the struc-
ture of that winding and that interaction with other wind-
ings has significantly less influence, introducing only
minor features and changes to resonant frequencies.

As far as WG A2.26 is concerned, the main purpose


of modelling is to contribute to the understanding of
generic issues rather than reproduce measured responses
Fig. 11: Measured (top) and calculated (bottom)
HV responses for Nuremberg transformer – of particular windings.
at tap Max (black) and Max-1 (red)

WG A2.26 is collating examples of typical responses Conclusion


for the range of transformer types typically encountered.
The final report of WG A2.26 will provide guidance
on recommended practices, the practical limitations of
M o d e l l i n g t o s u p p o r t i n t e r p re - FRA measurements and how winding movement can be
tation identified from measured responses.
One difficulty in interpreting FRA responses, espe-
In order to be able to confidently interpret any cially for new users, arises from the fact that FRA mea-
change or unusual feature in a measured response, it is surements are graphical rather than numeric, and that
necessary to have an understanding of controlling fac- there is no generally accepted guide for interpretation.
tors. Experience alone can build up knowledge and
insights, but modelling provides the best hope for a thor- The intention of WG A2.26 is to assist by providing
ough understanding. a description of typical responses and controlling fac-
tors, together with a collection of case examples which
The measured responses of the Nuremberg trans- illustrate how various types of winding movement can
former have been successfully reproduced (Figure 11) be detected by FRA.
by a lumped parameter model which includes mutual
inductances and capacitances for each disc pair calcu-
lated using only the winding geometrical dimensions, R e f e rences
together with a representation of the core. This trans-
former has a double layer LV winding, a partially inter- [1] CIGRE A2 2003 Colloquium papers on the
leaved disc HV winding and a single layer HV tap detection of transformer winding movement by FRA and
winding. other techniques, Merida Mexico, June 2003. ■

34 E L E C T R A N° 228 - Octobre 2006

Potrebbero piacerti anche