ET - Jan Feb Issue 1 2/19/07 2:52 PM Page 8

TRANSFORMERS

GETTING VALUE FROM TRANSFORMER SFRA

By G. Matthew Kennedy and Tony McGrail, Doble Engineering Company

INTRODUCTION
Sweep Frequency Response Analysis
(SFRA) is a test for mechanical deforma-
tion in power transformers. It has grown
in usage over the last decade and is now
being standardized by both IEEE and
CIGRE. Doble Engineering has pio-
neered use of the SFRA test over that
time and has gathered a great deal of
experience and value from the test. This
article looks at how you can gain value
from SFRA through a simple approach:
Control-Context-Conclusion.

SFRA – Keep it simple

SFRA provides a fingerprint of a
winding: a signal injected at one end is
measured at the other end of the winding.
Repeat this at many frequencies and we
have a simple response plot which we
can analyze, as shown in Figure 1.
The first thing to note is that the
responses shown are typical for a wind-
ing. The ‘picture’ which has been taken Figure 1 HV, LV and Short Circuit Responses for a single winding
should not vary significantly in future for
the winding – and variation should be
investigated as one possible cause that
winding deformation may have occurred.
(As a note – impulse systems do not mea-
sure the frequency response, they esti-
mate or infer it from impulses, assump-
tions and Fourier Transformer calcula-
tions; they are inherently less repeatable
in practice).

Gaining CONTROL
Experience has taught us at Doble
that keeping the measurement simple is
the best way to ensure that value may be
extracted from it. The Doble M5200 and
M5300 test sets automatically check for
ground loop continuity, for example, as
that is a key component of the high fre-
quency response. The Doble test leads
are simple one-piece design to avoid los-
ing pieces or incorrectly applying them.
Control of the SFRA measurement is
key to gaining good results. We stress the
need for training and support as we know
Continued on page 10 Figure 2 What went wrong?

8 Electricity Today
ET - Jan Feb Issue 1 2/19/07 2:52 PM Page 9

Mikron’s SpyGlass™ Lens and ViewPorts

encourage frequent examinations
of electrical switchgear because
— with cabinet doors closed —
no downtime is required to de-energize
circuits for safety reasons.
• ViewPorts are maintenance-free and
easy to install—20 minutes or less!
• Suitable for both low-voltage and
high-voltage applications (480 volts and up)
• Safer plastic-tipped Spyglass™ Lens
blocks “path to ground,” protecting operator
Raising the safety and convenience standard for thermal inspections.

Mikron Infrared, Inc. • Thermal Imaging Division

Tel: (906) 487-6060 • Fax: (906) 487-6066
e-mail: [email protected] • www.mikroninfrared.com

Call 1-888-506-3900 or visit us at mikroninfrared.com

ET - Jan Feb Issue 1 2/19/07 3:37 PM Page 10

TRANSFORMER SFRA
Continued from Page 8

that it can be quite a challenge to start

taking SFRA measurements without
proper education and back up. We under-
stand the common pitfalls that are
encountered.
Overall – it is necessary to keep it
simple and do it right. For example,
Figure 2, shows two successive SFRA
results for one winding of one trans-
former. The results are different – does
this mean that the transformer is bad?
In fact, the source of the difference
in this case can be traced to a change in
DETC (de-energized tap changer) posi-
tion between the measurements. A small
thing, but a key element of control.
Good training and good understand-
ing of parameters which can affect the
results help produce valid data which can
be used for decision support.

Understanding CONTEXT
Assuming that we have exerted Figure 4 Data as supplied by a Doble Client
good controls and made a valid set of
SFRA measurements, the next question
is: What does it mean?
If all the results look good, and we are fortunate enough to
have reference (baseline) results and they match, then we have
strong evidence that nothing has changed in the transformer. If,
however, there are some variations compared to baseline, or we
have no reference results, what are we to do?
First we should note that SFRA is not the only tool in the
bag. Powerful though it is, we should relate SFRA results to
other results. Figure 3 shows how SFRA open and short circuit
measurements relates to other key electrical tests: DC winding

Figure 5 Three in service single phase units

service after routine maintenance as a result of SFRA tests. And

Figure 3 SFRA as part of an Integrated Test Suite even then, further testing was recommended to ensure the deci-
sion was a good one!
If the transformer trips out of service then we have far
resistance, power factor/capacitance, Leakage impedance (also greater grounds for being critical. But we must ask which phases
called leakage reactance) and Exciting currents: were at fault, was it an internal fault or a through fault? What
When faced with data which may be anomalous, we need do the DGAs (should we have them) look like?
to look at the whole picture. That includes the motivational
question: why am I here? Drawing CONCLUSIONS
If the test results were generated as part of routine mainte- Understanding the context allows us to draw appropriate
nance, then it is almost certain that the transformer is going conclusions. For illustration we have some simple examples.
back in service. I can think of only a couple of instances over
the last ten years where a transformer has not been put back in Continued on page 18

10 Electricity Today
ET - Jan Feb Issue 1 2/19/07 3:45 PM Page 11
ET - Jan Feb Issue 1 2/19/07 3:05 PM Page 18

TRANSFORMER SFRA 5
Continued from Page 10

CASE 1:
SHORTED TURN
One of the easiest situ-
ations to identify with
SFRA is a shorted turn in a
winding. The effect is to
make what is an open cir-
cuit measurement look like
a short circuit measurement.
The data in Figure 4
was supplied by a Doble
client who was able to rec-
ognize good data and
understand that the phase-
to-phase variation he was
seeing was not normal and
was not a result of poor set
up or site procedure.
Even without reference
results, it is possible to say
that one winding result is
acting as if there is a short
circuit on that phase; a
more subtle aproach is to Figure 6 Three single phase units post repair
try and investigate the pos-
sible level of damage to the
windings. However, given
the nature of the problem, it is unlikely to be
a field repair, though such things have been
possible in some cases.
In this case the shorted phase was diag-
nosed, the transformer removed from service
and repair effected.

CASE 2: SISTER UNIT REFERENCE

Results from four single-phase units
were used to identify a problem with one unit.
Three Westinghouse 7-million series units
were in service; the fourth was a spare. Figure
5 shows the results for the three in service
units; clearly one phase is different: but why?
The results from the spare unit were
similar to those of the ‘good’ in-service
units; variations at higher frequencies were
noted and attributed to different grounding
regimes on the spare transformer. Detailed
analysis of the results for the suspect unit
S
Figure 7 HV-LV for Suspect Transformer
lead to a theory that the underlying capaci-
tances and inductances were unchanged. But
there seemed to be an impedance variation
and a loose connection was possible within the transformer. CASE 3: RESULTS FROM A GENERAL ASSESSMENT
An internal inspection revealed loose connections for the In this case a transmission auto-transformer of 750 MVA
bushing draw leads; these were tightened and the transformer and 400/275 kV was identified as suspect as a result of routine
retested. The results in Figure 6 show that the results have DGA. SFRA was performed as part of a routine inspection. No
‘come together’ again as expected. previous results were available, but reference results from a sis-
The low frequency variation is due to remnant core mag- ter unit of the same vintage (1965) were on file.
netization and is a well understood phenomenon while the Figure 7 shows a section of the HV-LV results for the sus-
higher frequency variations are acceptable – we do expect pect transformer; clearly one phase shows significant variation
some degree of variability in results! Continued on page 20

18 Electricity Today
ET - Jan Feb Issue 1 2/19/07 3:05 PM Page 19

5 to 10 arc flash explosions occur in electric equipment every day in the US*

Are YOU protected?

Everyday, workers in the electrical maintenance and utility industry work in environments that may
expose them to hazards that could cause severe or fatal burn injuries. In the event of a momentary
electric arc, everyday non-flame resistant work clothes can ignite and will continue to burn. Untreated
fabrics continue to burn until the fabric is totally consumed and non-flame resistant synthetic fabrics will
burn with melting and dripping causing severe contact burns to skin. The use of flame resistant clothing
will provide a level of thermal protection, and after the source of ignition has been removed flame
resistant garments will self-extinguish.
AGO INDUSTRIES INC. manufactures ARC/FR APPAREL to help protect workers in the electrical
maintenance and utility industry from the hazards of electric arc flash exposures.

Shirts • Pants • Coveralls • Coats • Jackets • Faceshields • Hoods

Comply with NFPA 7OE • Garments to ASTM F1506 Standards

AGO
Featuring
PO Box 7132 PH: 519-452-3780
London, Ontario FX: 519-452-3053
N5Y 4J9 EM: [email protected]
CANADA

INDUSTRIES INC.

www.arcfr.net * CapSchell Inc.

ET - Jan Feb Issue 1 2/19/07 3:05 PM Page 20

TRANSFORMER SFRA
Continued from Page 18

in response compared to the other two phases.

But we often see variation in the 20 kHz
region for the center phase. Is it normal for this
unit?
The same frequency range for a sister unit
showed no such variation (Figure 8). This was
substantial evidence of significant winding
movement and/or deformation – the frequency
range indicates that main winding capaci-
tances would be expected to be ‘out’.
However, it was known that the trans-
former had been in service before the tests were
carried out, albeit gassing somewhat. Further
testing showed that the main winding power
factors were acceptable, but the capacitances
had changed by more than 5% from the known
good unit.
A decision was taken based on the
strength of the SFRA and capacitance not to
return the transformer to service but to scrap it
and make ready a spare unit.
During the tear down of the ‘failed’ unit
significant winding deformation was found on
the suspect phase, the center phase, as predict-
ed.
At Doble we receive SFRA results every
day from around the world – we have SFRA
practitioners in over 40 countries at present.
The majority of results are acceptable; occa-
sionally they show some poor field practice
leading us to suggest retraining or a refresher
course for the field guys. Occasionally we get
an example of results which make it into the
Doble Conference – they are not just results to
be added to the thousands in the Doble data
base, but examples where the whole user base
can learn – both the experts at HQ and the guys
in the field.
If you have SFRA results and are interest-
ed in what they really mean – we’d love to
hear from you!

Figure 9 Winding showing significant buckling

CONCLUSIONS
SFRA is an easy measurement to
make, and with proper training and good sup-
port anyone can get good value from their
SFRA investment. The three Cs of SFRA:
Control-Context-Conclusion show the
thought process that needs to be applied to
SFRA.

Figure 8 HV-LV for Known Good Transformer

20 Electricity Today