You are on page 1of 5

TRANSMISSION

Transformer sweep frequency


response analysis (SFRA)
by G M Kennedy, A J McGrail and J A Lapworth, Doble Engineering,

Sweep frequency response analysis (SFRA) is one of the most powerful diagnostic tools for assessing mechanical damage to a transformer
winding. Analysis of the results, which are in the form of frequency response traces can, however, be daunting to new users.

One approach is to attempt automated and these are to be used after routine or has gained acceptability with a much
analysis, and this paper presents an approach type testing. Two relevant to this discussion larger and rapidly expanding group (and
using crosscorrelation coefficients. This is a include turns ratio to detect if the winding including the recent-entr y suppliers of
power signal processing tool that can simplify has faulted, and leakage reactance to equipment). But this has created some
result interpretation in some cases and allow identify deformation. Over the years the problems for these new users and suppliers
limits to be created. measurement of leakage reactance relating to the interpretation of the results.
A common failure mode for power transformers (or short-circuit impedance) has proven The experience is that this is a transitional
is consequent to mechanical deformation valuable, particularly when used during with- phase and new users can acquire the skill
of the core or windings. Core damage is stand evaluation in high power laboratories. within a short time.
more likely as a result of transportation, while The impedance change allowed post fault
But for that first phase there is an interest
winding damage is more likely to be caused can be defined as being less than a 2%
in having an analytical method, ideally
by short circuit type forces. Consequently, change (see IEC60076-5). Provided the
producing a number, ideally as simple as
the requirements are for transformer to be test is done correctly this provides a clear
checked before and after a new delivery or that from leakage reactance.
definition for acceptability. However, for
re-location, and also after any major fault. in-service assessment it is considered by The nature of SFRA results
The latter might be after faults on a cable some that this method is insensitive. In the
connected to the low voltage winding, or The sweep frequency response analysis (SFRA)
1990s a group from the major utilities in
on a measurement transformer on the high test involves injecting a signal at one end
northern Europe evaluated an alternative of a winding and measuring the response
voltage, or a tap changer or bushing failure.
method involving the injection of a low at the other end. Responses with large
These close-up faults may be so severe as
voltage swept frequency sine wave into variations in attenuation over the measured
to initiate protection and then the need is to
each winding in turn. They met as a working frequency range are obtained as a result of
assess the damage to the transformer.
group of EuroDoble and at the end of the variations in the impedance of the complex
Alternatively, there could be lesser damage 1990s documented their experience in a L-C-R distributions of the windings. Since
in the form of deformation. This would reduce Doble test guide and several papers at capacitances and inductances depend on
the capability of the winding to withstand the annual Doble conferences [1], [2]. detailed winding geometry, any movement
any further faults i.e. the winding is bent but Today, groups within IEEE and Cigr are results in changes in the frequencies at which
not broken. Knowledge of such damage is,
working to introduce the method into IEEE resonances occur. It is the identification
therefore, part of a risk assessment process
and IEC standards, and several companies of changes in frequency response that
to be applied for critical units.
now produce instruments to replace is the essence of analysis and diagnosis
Within IEC and IEEE standards, various the laborator y equipment used earlier. of mechanical integrity. This is currently
diagnostic techniques are described, Over these recent years the technique achieved with the expert eye - but the aim
is to use some processing to yield numerical
evaluation of changes.

But differences in response also occur for


reasons other than deformation. These include
differences between individual phases, tap
changer position and configurations of internal
leads between the bushings and windings- as
well as whether the oil was present or absent.
Unsuitable test equipment, lead connections
and layout can produce repeatability issues.
These issues present considerable challenges
when contemplating any automation of the
analysis.

Fig. 1 shows a typical response for a high


voltage star connected winding. The
frequency range of interest is between
20 Hz and 2 MHz. Experience has shown
Fig. 1: Frequency analysis bands. that different sub-bands are dominated

energize - October 2007 - Page 28


TRANSMISSION
dependent on transformer designs and
Frequency manufacturers, there is a risk that there
Region sub-band Component Failure sensitivity
will be false positives in the analysis.
Main core bulk
Core deformation, open circuits, shorted turns The subtle nature of some failures, which
1 < 2 kHz winding
and residual magnetism mean that the SFRA traces are almost
inductance
acceptable, leading to a possible false
Bulk component
2 kHz to Bulk winding movement between windings and negative in the analysis.
2 shunt
20 kHz clamping structure
impedances The traditional view of the Doble team, some
of whom have been reviewing data for almost
20 kHz to
3 Main windings Deformation within the main or tap windings 20 years, is that where unusual results are
400 kHz
obtained, any decision must always be made
400 kHz to Main windings,
~1 MHz Movement of the main and tap windings, by the expert to evaluate the existence of
4 tap windings
ground impedances variations interferences from issues listed above. Where
and internal leads
automated systems can help is as a short term
Table 1: Frequency sub-band sensitivity. crutch for the new user. Most applications will
be for fingerprinting sound transformers where
there should be good overlay between other
by different internal components of the lie not in the capability of the calculation or phases, sister units, etc.
transformer and are subsequently more the application, but in three main areas:
sensitive to different types of failures, as Here some comfort factor will be a help to
Large generic differences between the new user since it will be able to identify
summarized in Table 1. Measurements above
responses of different winding types.
2 MHz tend to be dominated by variations in the normal low level differences between
This means that setting cross-correlation
grounding practices for test leads. traces being compared. The method can
bands universally is difficult, if not
impossible; hence the recommendation act as a filter so that the engineer can use
Normal practice is to separate the response their limited resource of time looking at the
in this paper to make sure the user is
into frequency bands and relate the analysis most important traces. The output should
able to set the band limits to match the
to features in each band, as indicated in not be pass/fail but pass/investigate. The
transformer and winding under test.
Table 1. The first resonance, here at around goals are to express results as pass where
The fact that some differences between
600 Hz, is where the change from inductive variations are within normal bounds for the
responses are inevitable and must be
to capacitive impedance occurs. The allowed e.g. due to measurement configuration and a means to communicate
resonances at higher frequencies relate limitations or manufacturing tolerances. results to end-customers who are not so
to the winding configuration. The above Unless appropriate allowances are made, familiar with SFRA.
regions are a general rule of thumb
dependent on transformer design and test
conditions.

Sub-bands should be modified and adjusted


by the user based on the observed waveform.
For instance, the low frequency Region 1
could be changed to include all frequencies
between the lowest and the first dominant
pole, to completely encapsulate the core
effect, or a sub band within Region 3 could
be chosen that includes a series of dominant
resonances.

In the analysis a response may be compared


with one of the following:

An earlier result for the same phase tested


with the same tap changer position.
If no earlier result is available then another
phase of the same transformer, tested at
the same occasion.
The same phase, same tap changer
position but on a unit believed to be of
the same design group and made at the
same factory.
The preference for selection of options is in
the order as given.

Cross correlation analysis

Several attempts have been made over


the years at creating automated or semi-
automated SFRA analysis tools. Nearly all
have been unsuccessful. The reasons for this

energize - October 2007 - Page 29


TRANSMISSION

CCF
Good match 0,95 1,0
Close match 0,90 0,94
Poor match < 0,89
No or very poor match < 0,0

Table 2: Example CCFs explained.

CCF
Good 0,95 1,0
Marginal 0,90 0,94
Investigate <0,90
Fig. 2: Cross correlation analysis of the H2-H0 winding
to benchmark results. Table 3: Phase A/C or sister unit
(same phase) limits.

Cross correlation coefficients (CCFs) are CCF


researched in a variety of mathematical texts
already used in a variety of industries, primarily on the subject of signal processing [3]. Good 0,98 1,0
telecom, where knowing the quality of signals Marginal 0,96 0,97
Normalizing the results to the individual
is important. They are already well understood Investigate <0,96
power spectrums is what allows this resulting
and provide the type of analysis needed
waveform to be expressed in a simple single Table 4: Benchmark limits.
without blurring the analysis with go/no-go
coefficient. Table 2 helps provide a rough
type outputs.
estimate of what the CCF means in simple
In simplest terms, cross-correlation takes two internal leads. This type of variation could be
language.
sets of numbers and looks at how similar they normal between phases of the same unit.
Calculating cross-correlation coefficients:
are. If two series of numbers such as an SFRA Phase to phase comparison can often be
trace perfectly or nearly match, they would Using CCFs to analyze SFRA data first requires misleading and must be done with care.
have a CCF very close to 1,0. If two traces an understanding of what frequency sub- There are expected variations between
have absolutely no correlation, in other words bands tell us about the physical health of phase and correlation coefficients limits
are completely random, they would have a a transformer. Once the appropriate bands should be selected that draw attention to the
CCF of 0,0. If the two traces are related by are selected the CCFs can be evaluated analysis, and not necessarily call for failure
are diametrical, they would have a negative in the context of the individual parts of a of the unit. Based on experience, phase to
CCF. transformer. The end results will be a series of phase analysis would need to have the most
In SFRA analysis negative CCF are not CCFs, evaluating different components of forgiving limits. Phase to phase analysis of
common, but they do occur on occasion. the transformer. the outside core legs will also be the most
Regardless, negative correlation coefficients credible as these two win dings display the
The following traces were obtained using the
are not considered acceptable when trying most geometry symmetry. Comparison to the
Doble SFRA software for M5200/M5300. By
to look for deviations between traces. The middle phase will be difficult with correlation
adjusting cursor positions to frequency sub-
CCF is defined [2] as: coefficients and should be analyzed using
bands, new auto-correlation constants can
be calculated. Fig. 2 shows the comparis traditional SFRA methods.
on of two phases of a transformer where the Sister units can also be used to conduct
(1)
cursors were set to the positions shown in correlation coefficient analysis. In this case,
Table 1. Note that the first five cursor positions we would expect some variation between
where Xi and are Yi are the two series (or traces correlate to the regions. The correlation genuine sister units, but the analysis must be
in the case of SFRA) being compared at each coefficients are shown below the cursor done on the exact same phases of both units
individual frequency i and X-bar and Y-bar positions to indicate the correlation between (A to A, B to B, etc.)
are the means. the two above frequency regions.
The following correlation coefficients limits are
Equation 1 assumes two real series. In the By observing the above two phases of the recommended for phase to phase and sister
case of signal processing the math becomes same unit, we can see that there are possible unit comparison.
a little more involved, but the end results is still problems noted with the lower correlation
a coefficient between 1 and -1. coefficients of 0,9177 and 0,1906. The The most reliable method of conducting
variations can be seen in the overlay of the correlation analysis is using a benchmark
In simplest terms cross-correlation of two results. Similar to sister unit analysis , the same
two phases. Additional analysis and perhaps
traces f and g is:
narrowing the sub-band could help tell us phases of the same windings should always
(f*g)(x) = f*(t)g(x+t)dt (2) more. The Region 4 correlation coefficient be compared. In addition, the analysis should
of 0,1906 is most likely due to the mismatch be done with the transformer in the same
This results in a new correlation trace that is
of resonance in the 600 kHz to 900 kHz tap changer positions. The following table
then power density normalized to the input
region. It is important to note that this does provides recommend correlation coefficients
and output signals. The large * indicates
not necessarily indicate that the unit is bad limits for benchmark comparisons.
convolution. The mathematics of Equation
2 and subsequent CCF calculation are as this highest frequency region is known to These limits should be used as an aid to
beyond the scope of this paper and can be be much more sensitive to tap winding and analysis and reporting results but should not

energize - October 2007 - Page 30


TRANSMISSION
be used in isolation to judge the mechanical
condition of a transformer. A SFRA technician
should still analyze the results and use these
limits as an aid for additional investigation.

Case studies

The following case studies show various


scenarios that a SFRA tester may encounter
in the field and demonstrates how to use
correlation coefficients in the analysis.

Case 1: Benchmark comparison

In this case we have factory and initial field Fig. 3: Cross correlation analysis of benchmark results H2 - H0 winding.
results available for a 675 MVA generator step-
up transformer built in 2002. The transformer
suffered from a fire in the connected
isophase bus and was further tested using
SFRA and other electrical diagnostic tests. This
case was presented as a paper at the 2005
Doble Client Conference.

Variations at low frequencies relate to the


magnetic state of the core of the transformer;
the variation is commonly seen, is well
understood and is acceptable. Minor variations
at the highest frequencies are due to minor
differences in test lead grounding due to
Fig. 4: Cross correlation analysis of case 2 (phase to phase) -
bushings which had been damaged during the scan from 20 Hz - 2 MHz.
fire. The main responses overlay very well. This is
strong evidence that nothing has moved within
the transformer. SFRA was used in conjunction
with leakage reactance to propose that the
transformer was mechanically sound and was
worthy of an internal inspection rather than
scrapping. The inspection confirmed the SFRA
diagnosis and the transformer was successfully
returned to service.

Using the cross correlation coefficient analysis


on the H2 - H0 winding and default regions,
we find that the unit was acceptable with the
following cross correlation coefficients.
Fig. 5: Close-up view of LV winding frequency shifts in 100 kHz 1 MHz.
As mentioned above, the slight dip in the
Region 1 CCF is due to the core magnetization magnetization could be easily recognized we look closer, Fig 5, we can see that it is the
and there were some minor variations in the and the minor variation in Region 3 ruled out X3 - X0 phase which is consistently shifted to
Region 3 but the CCF was still very good as an issue. The end result is four figures that higher frequencies. This is an indication of a
at 0,9882. This helps draw the technicians can be used to quantity and report SFRA problem which may relate to axial winding
attention to these two sub-bands. The findings in an easy to interpret format. All movement. Note that we are also using the
findings are above the recommended limit of similarity of the X1 - X0 and X2 - X0 phases
0,98 CCF for benchmark comparison. to act as reference traces; if the center and
Frequency sub band CCF
one outer phase are similar then, generally,
1 0 - 2 kHz 0,9879 Case 2: Bent transformer
the other phase should also be similar! This
2 2 kHz 20 kHz 0,9964 The results here are from a 1960s vintage transformer was removed from service and
3 20 kHz 400 kHz 0,9882 50 MVA distribution transformer. The transformer scrapped.
4 400 kHz 1 MHz 0,9988 had tripped out of service on protection and
Using CCF analysis methods, again the regions
was investigated; no reference results were
Table 5: Case 1 CCF results. were set to the default sub-bands. The phase
available for this unit. The HV results phase-
A and phase B CCFs were then calculated.
to-phase had typical variation for a HV delta
Frequency sub band CCF We can see in Table 6 that there is a large
winding.
1 0 - 2 kHz 0,9831 dip the Region 3 CCF down to 0,8262. Using
2 2 kHz 20 kHz 0,9868 An overall look at the LV windings shows a few the recommended limits for analyzing phase
obvious variations, and a trained eye can to phase (Table 3), we can see that Regions
3 20 kHz 400 kHz 0,8262
see that this is particularly at the resonances 1 and 2 are all within the good limit and do
4 400 kHz 1 MHz 0,9567
between 200 kHz and 2 MHz where there are not warrant continued investigation. Region 4
Table 6: Case 2 CCF results. some shifts which are unexpected, Fig 4. If is below 0,98 but above 0,9 and therefore is

energize - October 2007 - Page 32


TRANSMISSION

Frequency sub band CCF


0 - 2 kHz 0,9898
2 kHz 20 kHz 0,9994
20 kHz 400 kHz 0,9913
400 kHz 1 MHz 0,9923

Table 7: Case 3 CCF results


HV phase A sister units.

marginal and deserves a closer look. Region


3 has dropped into the investigate region
(less than 0,9 CCF). Clearly the region did not
correlate as was discussed earlier.
Fig. 7: Cross correlation analysis of Case 3 (phase to phase
Case 3: Two large sister transformers
comparison of HV phase A sister units)
The following case involves the routine test of
two large transformers of 370 MVA 345/14 kV
CCFs should be used with an understanding starting points for analysis and may change
rating. SFRA tests were used to ascertain the
of traditional SFRA analysis techniques. There as more research is done and CCFs are
physical health of the units individually then
is no replacing the trained eye of an SFRA applied to more transformers.
sister unit comparison was used to ascertain
the consistency of construction between the technician as they can recognize the features
Acknowledgement
two units. By using both of these techniques it and patterns associated with various faulted
wa s determined by both phase to phase and conditions. CCF will not tell us the exact This paper was presented at the IEEE Power
by sister unit comparison that both units were nature of the failure, only draw our attention to Africa 2007 conference at Wits University,
in good physical condition without indications regions of interest for continued analysis and Johannesburg, in August 2007, and is
of either winding or core deformation. consideration. Users of this method should republished with permission
also be aware that it is possible for the CCF to References
Comparing sister units must be done with
not indicate a problem when there could be.
care to ensure that the units are really [1] J A Lapworth and T J Noonan, Mechanical
sisters by design. Experience has shown Inappropriately assigned region boundaries, condition assessment of power transformers
that it is not unusual for functionally identical or single resonance shifts for example, could using frequency response analysis
units to have significant design differences, cause only a minor change to the CCF and Proceedings of the 1995 International
client conference, Boston, MA, USA
particularly if some time has elapsed between in truth indicate a substantial problem.
[2] G M Kennedy, C L Sweetser and A J McGrail,
constructions. Manufacturers may take the
The suggested CCF limits should be used Field Experiences with Sweep Frequency
liberty to tweak the design of a transformer thus Response Analysis, Proceedings of the
as general rules of thumb and can be
creating physical deviations in the structures. 2006 International Conference of Doble
adjusted to fit the scenario in question. For Clients. Sec T-6
In this case, they are not sister units. Looking for
successive serial numbers may not guarantee instance, comparing closely design sister [3] A V Oppenheim and R Schafer, Discrete-
two or more units are genuine sisters; more units may necessitate raising the CCF limits Time Signal Processing, Prentice-Hall Signal
proposed in Table 3. These figures of merit Processing Series, 1989
attention should be paid to the date of
manufacture. As a rule of thumb, if more than can help communicate SFRA findings to Contact Steve Svendsen, Eberhardt Martin,
six months has elapsed between constructions, end customers. Again, these should serve as Tel 033 386-0011 v
there may be some differences.

Table 7 and Fig. 7 show an example analysis of


two sister units. As can be seen from the CCFs,
the results show very good correlation. Using
the recommended limits for sister unit analysis
(Table 3), we can see that all correlation
coefficients are well above the recommended
limits and in fact these two units are extremely
close sister units in design.

The slight dip in the Region 1 CCF is most likely


due to a very slight amount of reluctance
variation in the core and is normal.

Conclusion

CCFs can be a useful tool to conduct


specification based analysis of SFRA traces
under a variety of scenarios to include
benchmark results, phase to phase analysis
and sister unit comparison. Care must be
taken to understand the context of the
trace and know that some variation is
expected based on tap changer position,
test conditions and the transformer design.

energize - October 2007 - Page 33

You might also like