Page 1 of 11

Medium Voltage Switching Transient Induced

Potential Transformer Failures; Prediction,
Measurement and Practical Solutions

Daniel McDermit David D. Shipp, PE Thomas J. Dionise, PE Visuth Lorch

Sr. Project Manager Eaton Electrical Group Eaton Electrical Group Eaton Electrical Group
Turner Construction Power Systems Power Systems Power Systems
Company Engineering Engineering Engineering
Chicago, IL Warrendale, PA Warrendale, PA Warrendale, PA
Fellow, IEEE Senior Member, IEEE

Abstract— During commissioning of a large data center, while including 12 generators, UPSs and PDUs. This equipment
switching medium voltage circuit breakers without any supports four CoLo rooms of data center white space. The
appreciable load, several potential transformers failed mechanical and electrical infrastructure also supports the
catastrophically. A detailed investigation, including computer equivalent of seven CoLo rooms of critical load in
simulation was performed. Ferroresonance produced by approximately 50 to 60 containers (each container is capable of
switching transients associated with opening and closing the supporting 350 kW of critical load).
vacuum breakers, was determined to be the cause. The analysis
also determined that the close-coupled power transformers were
also in jeopardy. Field inspections involving grounding
B. Commissioning History
improvements coupled with solution simulations were made. On February 18, 2009 the construction team,
High speed switching transient measurements were performed to commissioning team, testing engineers and contractors were
verify the analysis and the surge protective devices solution performing manual open transition testing from one of the data
(arresters and snubbers). This paper walks the reader through centers three main utility feeds to Generator power. The
problem recognition, simulation, field measurements and solution Appendix gives an excerpt from the Sequence of Operation
implementation. Special focus will be made on the field testing that was being performed when the first Unit Sub
measurements verification. Station (USS) had an event where a Potential Transformer (PT)
failed. The Appendix describes each step in detail.
Keywords- switching transients, vacuum breakers, potential
transformers, ferroresonance, EMTP simulations, surge arresters, The technician performing the aforementioned sequence of
RC snubbers. operations, specifically Step #32, noticed that there was an
issue with the generator kW reading on the digital feeder
I. INTRODUCTION protection relay which caused him to stop and investigate with
the NM1 (Normal Main – Primary Unit Supply) closed, the
A. Facility Description GM (Generator Main) open and the water-cooled load banks at
600 kW. During his investigation in the Unit Substation
The ultimate build out of this facility strategically located in
Room, smoke began to appear from the bus PT truck, Phase A.
the midwest is approximately 400,000 sq. ft. of colocation
The PT was easily de-energized by opening NM1 in manual
(CoLo) data center white space and containers that will be
mode. A quick investigation indicated that there were no
supported by 250,000 sq. ft. of Central Utility Building (CUB)
primary or secondary fuses blown on the PT.
and approximately 25,000 sq. ft. of office and support space.
The data center will ultimately support approximately 59.4 A similar incident occurred two days later while the
MW of critical load. technician was setting-up to perform the same testing
sequences on another USS. The technician had the generator
The first completed phase of the facility, Phase I, is
tags cleared and started the associated USS generator to verify
approximately 200,000 sq. ft. of data center space, which
parameters. In manual mode the technician open-transitioned
consisting of 95,000 sq. ft. of standard CoLo type white space
the source back to Utility A. Within a matter of minutes the
and 96,000 sq. ft. of container based data facility, as well as
technician heard something “pop” and then noticed smoke
125,000 sq. ft. of utility/mechanical plant space. The
appearing through the vents in the NMI (Utility A) section.
utility/mechanical plant consists of 1) 12 chillers and cooling
Upon investigation of the USS, it was noted that there was a PT
towers and associated equipment, 2) thermal storage tanks, 3)
failure in NM1.
air handlers, and 4) critical electrical support equipment

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

 Page 2 of 11

C. Recognition of a Problem
The project team experienced a series of PT failures that
occurred while the commissioning team and equipment
vendors were performing open transition sequence testing
while utilizing the feed from UTS-A and the feed from UTS-C.
The stress on the project team became heightened and everyone
from the end user down to the technicians working on the
equipment were driven to discover the reason for the PT
failures. The question arose, had anyone on the project
experienced such an event or had knowledge of similar events?
A few of the team members recalled similar events while
working on their last data center project which reminded them
of the PT failures experienced here. In summary, on the
previous data center project, the authors performed an
engineering analysis at their facility to simulate the problem
observed at the present jobsite.
In the previous report, the authors stated “We are very
much aware of how vacuum breaker induced switching
transients can cause transformer failure. We have been doing
many computer simulated switching transient studies recently
to quantify the problem and to verify the solution. A key
element to the failure mechanism is short cables between the
vacuum breaker and the transformer”. This immediately sent
up a red flag to the team electrical manager and others on the
team, since all of the PT failures occurred around the time of
open transition (switching) operations. Additionally, the
Figure 1. Damage to PT but Fuses Remained Intact
distance between the NM1 breakers and the primary side of the
transformers is relatively short (contained within the substation Fortunately, the PT failure mode at this facility was not
itself). We discovered that this facility was potentially catastrophic. Instead, after vacuum breaker switching, the
experiencing a phenomenon called "Transient Voltage following was observed on two separate occasions: 1) the fuse
Restrike". In transient voltage restrike, the combination of did not blow, but smoke came out of the PT compartment, and
variables that can cause transformer failures usually involve a 2) the fuses blew before the ionized gasses could take out the
vacuum circuit breaker interrupting inductive loads that are entire cell.
supplied by cast resin power class transformers.
E. End User Response
D. Failures of PTs During Vacuum Breaker Switching
Although the PT failures did not result in massive failure,
Fig. 1 shows the PTs which failed during vacuum breaker the end user was concerned the PT failure could have escalated
switching. The photo shows the PT sustained the damage. The into severe damage of the switchgear. The end users response
fuses did not blow and remained intact. Typically, a PT may was to investigate the event, develop a test procedure to
have on the order of 8000 turns per winding. Close investigate the root cause of the failed PTs, and to ultimately
examination of the PT shows the damage consisted of a series recommend a solidly engineered correction to the sequence of
arc. Such damage is indicative of ferroresonance. When the operation or a re-engineering of the electrical gear itself.
breaker opens, a DC charge is trapped on the stray capacitance
of the cable, which is imposed on the primary winding of the
PT. The DC trapped charge saturates the iron of the PT, which II. SWITCHING TRANSIENT THEORY
fatigues the winding insulation. The frequency of the PT
ferroresonance observed at this facility was about 20 Hz. In A. Decision to do a Study
this special case of ferroresonance called PT saturation, the PT Fig. 2 summarizes the PT failures at Substations USS1B
may draw only 0.1A which is not enough to blow the fuse on and USS8B on the line-side and load side of the 1200A
the PT primary. Consequently, the series arc could last for vacuum circuit breakers. At USS1B, there were both line-side
hours or up to weeks until the insulation breaks down, at which and load-side PT failures. At USS8B, there was one line-side
point ionized gases are produced, and a complete fault occurs PT failure and three load-side PT failures. The figure calls
blowing the fuse. In the worst case, the ionized gasses attention to the issues: 1) a large number of PT failures at two
contained in this confined space develop an arc phase-to-phase- different substations, 2) the failures involved both line-side and
to-phase, i.e. a three-phase fault. This three-phase fault would load-side failures, and 3) the failures occurred with regularity
cause significant damage to the switchgear. Such a failure during switching of the 13.2 kV feeders when transitioning
would result in significant downtime, reducing the reliability of from utility to generator. For these reasons, it was decided to
the power delivered to the mission critical loads. create a model of the power system and conduct a switching

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

Page 3 of 11

Figure 2. Summary of PT Failures at USS1B and USS8B

transients study to determine the root cause of the PT failures. lowered. It is possible at some higher voltage this lower
The decision to conduct a switching transients study permitted saturated value of XL may equal XC, forming a series resonant
simulation of a variety of switching events, determination of circuit called ferroresonance. As in the normal series resonant
sensitivity of parameters contributing to the failures and circuit, the source voltage (VS) does not change much, but the
demonstration of the effectiveness of proposed solutions. voltage for nonlinear inductance (VL) and voltage for stray
Further field testing was placed on-hold until the study was capacitance (VC) increase and oppose each other. Since ZL is
completed and the system response understood. nonlinear, the voltages VL and VC become distorted or
irregular.
B. Background on Ferroresonance Some type of system disturbance is needed to “jolt” the
The primary focus of this study was ferroresonance. transformer XL into a lower saturated value equal to the system
ANSI/IEEE Std 100-1984 defines ferroresonance as “a capacitance XC. This “jolt” allows XL = XC and ferroresonance
phenomenon usually characterized by overvoltages and starts. Ferroresonance can continue for a long time (minutes,
irregular wave shapes and associated with the excitation of one hours or even days) since little resistance (R) is in the circuit to
or more saturable inductors through a capacitance in series with damp the oscillations.
the inductor.” The key elements are saturable inductors in
series with capacitance. The nonlinear inductance (XL) is In the case of this medium voltage distribution system, the
usually associated with the core of a transformer. The nonlinear inductance is either the 14,400/120 V PT or 13.2/0.48
transformer core will saturate with flux as voltage increases. kV, 3000/4500 kVA power transformers. The capacitance is
The transformer has a saturation curve, which gives flux as a dominated by the stray capacitance of the cables. The “jolt”
function of voltage. XL has high value for nonsaturation, and needed to initiate the ferroresonance is the opening of the
XL has low value when the core saturates. The saturation curve vacuum circuit breaker which feeds downstream PTs and the
has a “knee” where the change takes place. Transformers are power transformer.
designed to operate near the “knee”. In the ferroresonant
circuit, the capacitance (XC) can be the capacitance of cable, C. Modeling Ferroresonance
overhead line, or stray capacitance of transformer windings or In modeling this medium voltage distribution system for
bushings. such ferroresonance analysis, it was important to accurately
Under normal operation, XC is smaller than XL. However, represent the opening of the primary vacuum breakers, stray
if some switching event causes the voltage to increase, then the capacitance of the cable, and nonlinear inductance of the
transformer core may be pushed into saturation and XL transformer being switched, i.e. transformer saturation. The

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

 Page 4 of 11

authors modeled these critical circuit components in the breaker, cable, PT, and Transformer T-8B. The cable was
Electromagnetic Transients Program (EMTP). represented by a Pi model consisting of the series impedance
and half of the cable charging at each end. (In some cases,
The authors’ study approach was to first model the steady- multiple Pi models are used to represent the cable.) The
state conditions with the transformer (PT or power transformer) vacuum breaker was represented by a switch with different
energized. In this way, it was possible to show the normal models for opening (current chop of 5A), re-strike (excessive
excitation current drawn by the transformer to magnetize the magnitude of TRV), re-ignition (excessive frequency of TRV)
nonlinear inductance. Next, the authors simulated the actual and closing (pre-strike). The three phase transformer model
switching conditions which produced ferroresonance during consisted of the leakage impedance, magnetizing branch,
transition from utility source to generator source. These winding capacitances from high-to-ground and low-to-ground.
opening conditions produced ferroresonance, as evidenced by The PT model included saturation effects. Actual switching
erratic voltage and current waveforms shown later in the paper. scenarios that resulted in PT failures were simulated, and the
Finally, the authors added mitigation in the form of RC results of these simulations are described below.
snubbers to provide damping and mitigate the PT
ferroresonance. a) Open utility-side 13.2 kV feeder to USS8B followed by
The actual switching conditions consisted of opening the closing generator-side feeder (open transition to generator)
13.2 kV utility feeder to USS8B, and closing the 13.2 kV Case N6 simulates the actual case during the testing even
generator feeder breaker. The worst case for ferroresonance though the transition time in the simulation is much shorter
occurs with an unloaded transformer. Either a PT or power than the actual time. Case N7 is the same as Case N6 but with
transformer may experience ferroresonance. On this basis, a snubber installed. Fig. 3 compares the study results for Cases
Eaton examined the electrical distribution system and selected N6 and N7 for the primary voltage at the 3,000 kVA dry type
the worst case conditions to check for PT and power transformer T-8B. The application of the snubber circuit (Case
transformer ferroresonance. Snubbers sized for this worst case N7), greatly reduced the transient overvoltage (TOV)
will protect the PT and power transformers during less severe magnitude at the 13.2 kV bus and the oscillation frequency.
switching operations. The oscillation frequency of roughly 22,000 Hz (Case N6) can
be reduced to 2,000 Hz (Case N7) and the resistor in the
D. Computer Simulations of Actual Conditions snubber circuit will damp the oscillation within 3 milliseconds.
Switching transients simulations were conducted in the A closer examination of the transient overvoltage is given
electromagnetic transients program (EMTP) to investigate the in Fig. 4. Fig. 4 compares the study results for Cases N6 and
possible failure of the PT due to transient overvoltages during N7 for the primary voltage at the 3,000 kVA dry type
the circuit switching of the vacuum circuit breakers. The transformer T-8B zoomed from 0 to 30 milliseconds. The
circuit model developed in EMTP consisted of the source, figure illustrates the opening of the utility side circuit breaker.

Case N6 (no snubber installed) Case N6 (no snubber installed)

Case N7 (snubber installed)

Case N7 (snubber installed)

Figure 3. TOV at USS8B During Transition from Utility to Generator with Figure 4. TOV at USS8B During Transition from Utility to Generator with
and without Snubber Protection and without Snubber Protection (Zoom of Utility Side Breaker Opening)

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

Page 5 of 11

The application of the snubber circuit (Case N7) reduced Fig. 7 illustrates the study results for Case N8 for the
the transient overvoltage from 170 kV peak with an oscillation primary voltage at the 3,000 kVA dry type transformer T-8B.
frequency of 1,594 Hz to 26.6 kV peak with an oscillation The figure was zoomed from 5 milliseconds to 10 milliseconds
frequency of 215 Hz. The study also shows a high DC offset and the resonance condition is clearly illustrated in the figure.
for Case N6 due to the energy transfer between the stray The simulated ferroresonance condition in Fig. 7 is a near
capacitor and inductance in the Transformer T-8B and PT. The match for the ferroresonance condition captured with the high-
magnitude of the transient overvoltage may be smaller due to speed power quality meter in Fig. 9.
the operation of the surge arrester, however the oscillation
frequency will remain the same. Fig. 8 compares the study results for Cases N8 and N9 for
the bus voltage at USS-1B. Since this location is only 10 feet
Similarly, Fig. 5 compares the study results for Cases N6 from Bus USS-8B, the results will be similar to the primary
and N7 for the primary voltage at the 3,000 kVA dry type voltage at the 3,000 kVA dry type transformer T-8B. The
transformer T-8B zoom from 90 to 120 milliseconds. The application of the snubber circuit (Case N9), reduced the
figure illustrates closing of the generator side circuit breaker. oscillation frequency from 22,700 Hz (Case N8) to 1,485 Hz
The application of the snubber circuit (Case N7) reduced the (Case N9) and the resistor in the snubber circuit will damp the
oscillation frequency from 20,000 Hz (Case N6) to 2,000 Hz oscillation within 3 milliseconds. Again, the oscillatory
(Case N7) and the period of transient can be was reduced from condition may cause the PT failure over a long period.
oscillatory down to 3 milliseconds. The oscillatory condition
may cause the PT failure over a long period. 3) Comparison of PT connections - Open utility-side 13.2
kV feeder to USS8B
2) Close generator-side 13.2 kV feeder to USS8B The PTs that failed were connected wye-grounded wye-
(enhanced model of open transition to generator) grounded (Yg-Yg). The analysis was expanded to consider the
Case N8 simulates closing the generator-side feeder breaker benefit, if any, of the open delta PT connection. In Case T3,
with an enhanced model of open-transition. Case N9 is the there is a nominal load of 100 kW with 0.97 lagging power
same as Case N8 but with the snubber installed. Fig. 6 factor on Transformer T-8B. The 13.2 kV circuit breaker is
compares the study results for Cases N8 and N9 for the primary opened on the utility side feeding transformer T-8B. First, the
voltage at the 3,000 kVA dry type transformer T-8B. The Yg - Yg PT connection was considered with no snubber. Fig.
application of the snubber circuit (Case N9), reduced the 9a shows the Transformer T-8B primary voltage. After breaker
oscillation frequency from 22,700 Hz (Case N8) to 1,486 Hz opening, the transient overvoltage was as high as 103.61 kV
(Case N9) and the resistor in the snubber circuit will damp the with an oscillation frequency of 1,378 Hz. Such a TOV will
oscillation within 3 milliseconds. Again, the oscillatory result in re-ignition of the breaker and a higher TOV.
condition may cause the PT failure over a long period.

Case N8 (no snubber installed)

Case N6 (no snubber installed)

Case N9 (snubber installed)

Case N7 (snubber installed)

Figure 5. TOV at USS8B During Transition from Utility to Generator with Figure 6. TOV at USS8B During Energization of Transformer T-8B from
and without Snubber Protection (Zoom of Generator Feeder Breaker Closing) the Generator with and without Snubber Protection

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

 Page 6 of 11

Next, in Case T4 the Yg - Yg PT connection with a snubber cables was about 30 ohms while the generator cables was about
was considered. Fig. 9b shows the transformer T-8B primary 20 ohms. Since the snubber must take care of both conditions,
voltage. After breaker opening, the TOV was reduced from 25 ohms was selected as a compromise. The authors have
103.61 kV (Case T3) to 17.413 kV (Case T4). The oscillation found that a perfect match is not required, but rather a few
frequency was reduced from 1,378 Hz (Case T3) to 214 Hz ohms difference is effective. By matching the resistor to the
(Case T4). However, the PT Ferroresonance is not eliminated cable surge impedance, the high frequency waveform
due to the PT connection of Yg-Yg in the temporarily impinging on the circuit will not double due to reflection. The
ungrounded circuit during open transition. resistor cancels the reflection as well as provides damping,
especially for the DC offset voltage, that causes ferroresonance.
Finally, in Case T5 the open-delta PT connection with a
snubber was considered. Fig. 9c shows transformer T-8B Next, the snubber values were entered into the computer
primary voltage. After breaker opening, the TOV was reduced model in EMTP. (EMTP is recognized world-wide as the
from 103.61 kV (Case T3) to 17.416 kV (Case T5). The premier tool for transient analysis of power systems.) In
oscillation frequency was reduced from 1,378 Hz (Case T3) to ETMP, the switching transients were simulated to ensure that
214 Hz (Case T5). With the PT connected open delta, the PT the selected snubber components actually accomplished their
Ferroresonance is eliminated. Adding a resistor to the PT equal objective. The discharge current through the resistor was
to 50% of the burden will assist in damping the ferroresonance. evaluated for three criteria: 1) the Joule (or equivalent watt
rating) must be well within thermal ratings, 2) the switching
E. Solutions involving snubbers, grounding and transient voltages must be well below BIL and 3) the dV/dt
ferroresonance elimination must be well below transformer limits. All three criteria must
be satisfied. For 15 kV transformers, if the ring wave
1) Snubbers frequency can be reduced to between 250 Hz and 1000 Hz,
Once the need for snubbers was determined, the surge then there will be no problem. This final simulation also
capacitorwas selected. The surge capacitorwas selected based proves that the natural frequency is not too low to cause other
on system voltage. Years of experience has shown the industry problems. Once the surge capacitor was added, its capacitance
that certain values work well for given applications. For dominated over the other capacitances in the entire circuit, and
example, 0.25 micro-Farad surge capacitorworks well for most reduced the oscillation frequency to an acceptable value.
15 kV class applications. This value does a good job of
slowing down fast transients to be within the dV/dt rating of 15 The specifications for the RC snubber circuit are given in
kV class transformers, switchgear, motors, generators, etc. The Fig. 10. The resistor average power rating is 750 W at 40 °C,
capacitor industry makes these surge caps with standard and the peak energy rating is 16,000 Joules. This snubber was
available values. Alternatively, one could carefully evaluate applied at other locations in the system. Although cable surge
the ring waves and fine tuned the natural frequency to meet a impedance varies somewhat from location to location, this
special value, but such an fine tuned approach generally
requires a custom build, i.e. special capacitance values with Case N8 (no snubber installed)
long delivery times. Such custom values are not required, and
the long lead times did not meet the schedule of the project.
Once the need for snubbers was determined, the first attempt
was to use standard available surge caps for that voltage level.
First, the resistor was sized to match the surge impedance
of the source cable feeding the circuit. In this case, two (2)
cables were address. The surge impedance of the utility source

Case N8 (no snubber installed)

Case N9 (snubber installed)

Figure 7. TOV at USS8B During Energization of Transformer T-8B from Figure 8. TOV at USS1B During Energization of Transformer T-8B from
the Generator with and without Snubber Protection (Zoom of ferroresonance) the Generator with and without Snubber Protection

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

Page 7 of 11

snubber circuit provides adequate mitigation of the voltage As a result, ground mats would be exposed to transferred
transients produced by the vacuum breaker at this location. earth potentials during switching and lightning, imposing a
Selection of one set of RC snubber values simplifies the design potential difference on such components as surge arresters,
and installation, i.e., only one type of resistor and capacitor cable shields, and Yg-Yg PTs. Transferred earth potentials
must be used. The 13.2 kV RC snubber circuit is required at refers to the phenomena of the earth potential of one location
every 13.2 /0.48 kV transformer. appearing at another location where there is a contrasting earth
potential. The transfer of potentials may occur over conductors
2) Grounding that have been purposely placed between the two locations.
The wiring methods and grounding means for the 13.2 kV For example, consider the situation of Fig. 11. A ground fault
system were inspected. The evaluation included the method of occurs at the remote substation between the primary
grounding, effectiveness of all connections, presence of ground transformer busing and its respective ground mat. The return
loops, and compliance with the NEC, FIPS 94, and other path for the ground fault current is through the earth back to the
applicable standards. The evaluation determined both the main substation ground mat and its respective system ground.
utility substation and the generation plant had well designed
ground mats that provided effective grounding for their T-8B
respective electrical equipment. However, it was noted that the
ground mats were not tied together. Conduit was relied upon to 3000
make this connection, but the use of PVC in several locations KVA
broke the continuity. R
25OHM SA
15kV Rating
C 12.7kV MCOV
Case T3 (no snubber installed) 0.25uF

Figure 10. Snubber Specifications and Surge Arrester Arrangement for the
Transformer Protection

Case T4 (snubber installed)

Case T5 (snubber installed)

Figure 9. Open utility-side 13.2 kV feeder to USS8B with a) Yg - Yg PT, b) Figure 11. Transferred Earth Potentials a) Typical substation, b) fault current
Yg - Yg PT with snubber, and c) open-delta PT with snubber path, and c) fall of potential distribution.

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

 Page 8 of 11

This earth return path will have a finite amount of high level of dV/dt signal. It was noted that the duration and
impedance Zg = Rg + j Xg where Rg is the earth resistance and peak voltage of the transient had increased from previous test
Xg is the earth reactance between the two substation ground sequences at higher load levels. The peak voltage measured
mats. The potential difference between the mats is given by: was 49.6 kV, and the duration was 876 µsecs.
∆Eg = Ig Zg (1)
B. Test Equipment Selection
Any electrical conductors, which may extend from within The system under test consisted of a 4.5 MVA dry
one substation to within the other, may be subject to this transformer connected delta-wye with PTs connected Yg-Yg.
potential difference ∆Eg. When ∆Eg is high, due to excessively The medium voltage power transfers were performed by
high ground fault currents or excessive earth impedance Zg, vacuum breakers. Three (3) 13.2 kV sources power the
that serious voltage stresses may be imposed upon various substation: 1) utility source UTS-A to the NM1 breaker, 2)
system components never designed to withstand them. For this UTS-C to NM2, and 3) GEN11B through the generator output
facility with independent ground mats, high frequency breaker (GM1) to the substation breaker GM. Test probes were
impulses on vacuum circuit breaker closing may cause installed on each phase of the line side of each substation
momentary voltage differences between the ground mats. This breaker, and one set at the line side of the transformer. Each
imposes a potential difference on the PTs well above the rated test set consisted of a laptop, a power quality meter, three
line-to-ground voltage, forcing the PT into saturation. voltage divider (100/1 ratio) probes, and connecting wires as
shown in Fig. 13 and described further in Table I.
III. MEASUREMENTS
C. Personnel Safety Issues at 13.2 kV
A. Before Solutions
To perform this test, the specialized test equipment was
Prior to the installation of the snubbers, several transients installed at USS-11B. The installation of these devices
were captured on one of the Unit Sub Stations. One of these is significantly changed the hazard in the area. Specialized pre-
shown in Fig. 12. It was noted that the lower load levels test briefings were implemented to verify all participants were
resulted in higher over-potential and higher dV/dt signals being aware of their roles, as well as the hazards of the test
developed from the breaker operations. Before installing procedure. Additionally, steps were implemented to isolate the
snubbers, the captured screen shot of Fig. 12 was obtained USS-11B from USS-7B. This simplified the test and the model.
when utility feeder NM-1 was opened at 300 kW showing a Table II gives the most important prerequisites and precautions.

TABLE I. EQUIPMENT FOR SWITCHING TRANSIENTS MEASUREMENTS

TABLE II. TEST PREREQUISITES AND PRECAUTIONS

Figure 12. Switching Transient Capture at PT

Figure 13. Voltage Dividers and PQ Meter for High-Speed Transient Capture

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

Page 9 of 11

D. Test Procedures to Assure Testing will not Fail the PT or parties involved were satisfied the snubber was effective in
3 MVA transformer (sneak up on problem) mitigating both the dV/dt and TOV problem. The client was
A testing scheme was developed to create individual confident that system reliability would not be compromised.
breaker closing and opening electrical transients at four discrete
load levels on USS-11B: 1200, 600, 300 and 100 kW. These IV. SOLUTIONS
events were evaluated by all parties involved, and concurrence Given that the facility was in the final stages of
was given to proceed to the next test step. This test was commissioning, a solution had to be both economical and
designed to approach the potentially more damaging conditions timely. Using this criteria, three technically sound solutions
in a deliberate manner. Prior to conducting the actual tests, were proposed and evaluated. Ultimately two of three
each test condition was simulated in EMTP. Table III proposed solutions were implemented as explained below.
summarizes the results of the simulations of the test conditions.
As can be seen in the results, as the load level on the A. Change to Different PT Connections (rejected)
transformer was reduced, the severity of the switching transient
overvoltage, ring frequency and ferroresonance condition The PTs that failed were connected Yg-Yg. In fact, all of
increased. This advance knowledge of the system response the PTs on the 13.2 kV circuits were of the same configuration.
enabled careful tracking of the field test results. Should the The analysis has shown the benefit of the open-delta PT over
field results depart dramatically from the predicted response, the Yg-Yg PT, even when the snubber is applied in both
the test could be stopped. In this way, it was possible to configurations. With the open delta PT, the ferroresonance was
“sneak-up” on the problem. completely eliminated. However, with the Yg-Yg PT
ferroresonance of about 20 Hz persists. The authors discussed
E. Confirmed Problem Existed replacing the existing the Yg-Yg PTs with open delta PTs.
However, it became apparent this would be cost prohibitive for
Testing began with no snubbers and the 1200 kW load and several reasons: 1) the relays would have to be changed, 2) the
proceeded down to the 300 kW load. The test sequences prior metering would have to be changed, 3) the commissioning
to the snubber installation were terminated at this point at the would have to be repeated because the commissioning was in
concurrence of the design engineers, facility operators and the the final stages, and 4), the project was nearly completed and
authors because the symptoms of TOVs and dV/dt were such a change-out would add significant delay. For these
worsening with each lower load increment. Fig. 12 shows the reasons, the possibility of changing to different PT connections
worst of transient overvoltage captured at the PT. All parties was rejected.
were satisfied the root cause of the PT failure was determined
without actually failing a PT. Next, the snubbers were installed, B. Tie Ground Mats Together (implemented)
and the captured screen shot of Fig. 14 was obtained when
utility feeder NM-1 was opened at 300 kW load. The snubbers As explained, the different ground mats, i.e. utility
eliminated the dV/dt transient response, and no TOVs occurred substation ground grid and generation plant ground grid were
at the PTs as shown in the voltage waveforms of Fig. 14. All not tied together resulting in transferred earth potentials. Such
transferred earth potentials would occur during switching and
TABLE III. SUMMARY OF TEST CONDITION SIMULATIONS
lightning, imposing a potential difference on such components
as surge arresters, cable shields and Yg-Yg PTs. One solution
was to increase the insulation strength of the affected
equipment. Another option was to tie the ground mats
together. The former option would require replacement of
equipment which was undesirable for many of the same
reasons mentioned in regard to changing the PT connections.
The latter option required minimal investment in an equipment

Figure 14. Switching Transient Mitigated by RC Snubber

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

 Page 10 of 11

grounding conductor (EGC) and the labor to install the EGC. waveforms at the PT and transformer primary produced during
For these reasons, the authors recommended tying the ground switching of the primary vacuum circuit breaker. The voltage
mats together to minimize the transferred earth potential. An dividers were made of capacitive and resistive components
EGC was installed outside the conduit runs between the utility with a bandwidth of 10 MHz. The power quality meter was
substation and the generation plant, and the EGC was bonded capable of transient voltage waveshape sampling, 8000 Vpeak
to the corresponding ground mats. full scale, 200 nsec sample resolution (5 MHz sampling). This
test equipment ensured accurate capture of the high frequency
C. Add Snubbers as Study Dictated (implemented) transients. The measurements verified the waveforms did not
The study showed that the application of the snubber circuit exhibit excessive high frequency transients (magnitude, rate-of-
greatly reduced the transient overvoltage magnitude and rise and frequency). The measurements also verified the PT
oscillation frequency at the 13.2 kV PT and Transformer T-8B ferroresonance condition was damped-out.
primary winding. The resistor in the snubber circuit damped the
oscillation and reduced the DC offset of the transient E. Typical Snubber Installation
overvoltage to within acceptable levels. Further, the study Fig. 15 shows the side view of the custom snubber circuit.
showed the snubber reduced the ferroresonance of the Yg-Yg The custom design was required to fit the snubber in the
PT by providing a source of damping that would otherwise not bottom of the transition section, between the No. 3 breaker
be present. The snubber, in combination with the existing section and the transformer. The snubber was field installed in
surge arrester, provided the maximum surge protection for the the bottom of the transition section. The surge capacitor serves
PT and the Transformer T-8B. The snubber components were as the base. The resistor is bonded to the surge capacitor
of relatively low cost and readily available. The snubbers did bushing and the bus bar, where the cables to the transformer are
not require any modifications to the PT circuit, relaying or secured. The flanges on each end of the resistor provide a solid
meters. The performance of the snubbers was proved by support for the fragile resistors. The capacitor was grounded to
measurements and did not require repeating the commissioning the ground bus through a flat-braided highly-stranded strap.
tests. For these reasons, it was decided to install the snubbers. The snubber was treated like a lightning arrester, i.e., no fuse
was installed.
D. Prove Solution with After-the-Fact Measurements
Following installation of the snubbers, power quality V. CONCLUSIONS
measurements were taken to ensure the proper operation of the This paper provided a detailed investigation of several PTs
snubbers. A high speed power quality meter and capacitive that failed catastrophically during switching medium voltage
voltage dividers were used to measure the transient overvoltage circuit breakers to transition from utility to generator.
Switching transient simulations and field measurements
determined the root cause of PT failure was ferroresonance as
well as switching transient overvoltages associated with
opening and closing the primary vacuum breakers. The
analysis determined that the close coupled power transformers
were also in jeopardy. The study showed that the application
of the snubber circuit greatly reduced the transient overvoltage
magnitude and oscillation frequency at the PT and power
transformer primary winding. The resistor in the snubber
circuit damped the oscillation and reduced the DC offset of the
transient overvoltage to within acceptable levels. Further, the
study showed the snubber reduced the ferroresonance of the
Yg-Yg PT by providing a source of damping that would
otherwise not be present. Also, grounding improvements were
made by tying together the ground mats in the utility substation
and generator plant, to minimize the transferred earth potentials
imposed on the PT primary winding. High speed switching
transient measurements verified the analysis and proved the
surge protective devices solution (arresters and snubbers) were
effective.

VI. ACKNOWLEDGMENTS
The authors wish to thank Turner Facilities Management
Solutions as well as Eaton Power Systems Engineering for
their collaboration in the presentation of this work: Dan Wilder,
Tim McConnell, and Rashad Hammoudeh of Turner as well as
Bill Vilcheck of Eaton. Additionally, we thank Scott Seaton.
Figure 15. RC Snubber (Typical Installation for this Facility)

978-1-4673-0651-5/12/$31.00 © 2012 IEEE

Page 11 of 11

REFERENCES APPENDIX
Table IV gives an excerpt from the Sequence of Operation
[1] IEEE Standard Dictionary of Electrical and Electronics Terms, testing that was being performed when the first Unit Sub
ANSI/IEEE Std 100-1984. Station had an event where a Potential Transformer (PT) had a
[2] Hopkinson, R.H., “Ferroresonant Overvoltage’s Due to Open failure. The PT failure occurred during the performance of
Conductors,” General Electric, 1967, pp. 3 - 6.
Step #32 described in the table below.
[3] Westinghouse Distribution Transformer Guide, Westinghouse Electric
Corp., Distribution Transformer Division, Athens, GA, June 1979,
TABLE IV. SEQUENCE OF OPERATION TESTING
revised April 1986, Chapter 4 Ferroresonance, pp. 36 - 40.
[4] IEEE Guide for Application of Transformers, ANSI/IEEE C57.105-
1978, Chapter 7 Ferroresonance, pp. 22 – 28.
[5] Distribution Technical Guide, Ontario Hydro, Ontario, Canada, May
1999, original issue May 1978, pp. 72.1-1 – 72.1-10.
[6] Greenwood, A., “ Electrical Transients in Power Systems”, Wiley &
Sons, 1971, pp. 91-93.
[7] Kojovic, L., Bonner, A., “Ferroresonance - Culprit and Scapegoat”,
Cooper Power Systems, The Line, December 1998.
[8] Shipp, Dionise, Lorch and MacFarlane, “Transformer Failure Due to
Circuit Breaker Induced Switching Transients”, IEEE Transactions on
Industry Applications, April/May 2011.
[9] ANSI/IEEE, A Guide to Describe the Occurrence and Mitigation of
Switching Transients Induced By Transformer And Switching Device
Interaction, C57.142-Draft.
[10] D. Shipp, R. Hoerauf, "Characteristics and Applications of Various Arc
Interrupting Methods," IEEE Transactions Industry Applications, vol 27,
pp 849-861, Sep/Oct 1991.
[11] ANSI/IEEE, Standard for AC High-Voltage Generator Circuit Breakers
on a Symmetrical Current Basis, C37.013-1997.
[12] ANSI/IEEE, Application Guide for Transient Recovery Voltage for AC
High-Voltage Circuit Breakers, C37.011-2005.
[13] D. Durocher, “Considerations in Unit Substation Design to Optimize
Reliability and Electrical Workplace Safety”, ESW2010-3, 2010 IEEE
IAS Electrical Safety Workshop, Memphis.
[14] D. Shipp, N. Nichols “Designing to Avoid Hazardous Transferred Earth
Potentials,” IEEE Transactions Industry Applications, July/August 1982.
[15] Shipp, Dionise, Lorch and MacFarlane, “Vacuum Circuit Breaker
Switching Transients During Switching of an LMF Transformer”, IEEE
Transactions on Industry Applications, January/February 2012.

978-1-4673-0651-5/12/$31.00 © 2012 IEEE