On – Line Transformer Diagnostics
For decades, transformer asset managers have sought ways to assess the general condition of electrical power apparatus and identify specific problems. Periodic off-line
diagnostic tests still play an important role in industry. However, “continuous” or “online”
monitoring can overcome some of the fundamental limitations of off-line tests while
increasing performance and reliability of the monitored equipment.
As technology advances, additional continuous or on-line monitoring techniques are
being developed and proven the in the field. This paper will present some newer online
diagnostics methods and provide examples of monitoring and diagnostics for
bushings and windings on transformers. These same techniques can be applied to high
voltage current transformers, CCVTs and other equipment.
Bushings
Bushings are subjected to high dielectric and thermal stresses, and bushing failures are
one of the leading causes of forced outages and transformer failures. Some studies
have shown they can account for up to 40% of transformer failures. Data also shows
that 52% of bushing failures are violent, in that there is an ensuing fire and collateral
damage.
The two most common failure mechanisms are moisture contamination and partial
discharge. Recent notices have been issued by bushing manufactures concerning the
effects of corrosive sulfur in the bushing oil causing PD and bushing failure. Moisture
can enter the bushing through deteriorated gasket material, cracks and loose terminals.
Moisture will cause an increase in the dielectric losses and consequently an increase in
power factor. As the deterioration in the bushing insulation continues there will be a
breakdown in the capacitive layers and tracking will become apparent which produces
partial discharges.
The method of detecting bushing insulation deterioration is well understood and
traditionally off-line tests have been performed. The struggle facing the asset manager
is that many of the failure mechanisms on bushings can occur very quickly, while some
are temperature and voltage dependent. Offline testing at 10 kV and at ambient
temperatures coupled with long intervals between tests can provided less than attractive
results. On-line monitoring of the bushings provides data during all weather conditions,
loads, and at rated voltage, with the same sensitivity as an off-line measurement.
It is evident that the online measurement of power factor and capacitance is a very
useful and reliable diagnostic indicator. A very sensitive method for obtaining these
parameters on-line is the sum current method. The basis of this on line monitoring
method is to compare insulation characteristics of a three-phase bushing system. By
vectorilly adding the currents from the test tap, one can determine the condition of the
bushings. If the bushings have the same specifications and the system voltages are
perfectly balanced, the sum current will equal zero. Of course in actual installations this
is not the case.
Figure 1 shows a basic block diagram of a bushing monitoring system that uses the
sum of currents method. During commissioning the null-meter is balanced to zero. The
purpose of the balancing circuit is to take into account the differences in system
voltages and bushing characteristics. As a defect develops the complex conductivity of
the bushing insulation changes and the current and its phase angle in one of the phases
also changes. Therefore, the null-meter will no longer be null. The amplitude of the
change reflects the severity of a problem and the phase angle indicates which phase is
experiencing the change.
Figure 1
Block Diagram of the Sum of Currents Method
The change can be approximately represented by the formula under the assumption of
a single defective phase:
Figure 2 shows a bushing monitor IED and a sensor installed in a bushing test tap. The system continuously monitors the power frequency current through the insulation of a 3-phase set of bushings as well as top oil temperature and load current.
Figure 2
Bushing Sensors
The test tap on bushing are normally grounded. If there were left open circuited, phase
to ground voltage could build up at the tap and cause a bushing failure. Also, the
instrumentation must be protected from system surges such as lighting strikes and
transients. Bushing sensors should meet the following protection requirements.
- Open circuit: Bushing insulation monitors usually have relatively low input
impedance that keeps voltage at a tap at a safe level. If a tap is left open
circuited, high voltage may build up on a tap according to capacitances C1 and
C2/C3 operating as a capacitive divider. It is common to have capacitances C1
and C2 of a bushing of comparable values, therefore, if tap is open circuited the
voltage up to half of normal phase to ground operating voltage may be present
on the tap pin providing both danger to operating personal and the bushing
insulation. In such a case, the insulation of C2 is over stressed by a voltage that
creates sparking inside the bushing, contaminating the bushing oil and finally
contaminating the main C1 insulation with full phase to ground insulation
breakdown. Therefore, a bushing sensor must keep “infinitely” safe rms voltage
on the tap even if disconnected from an instrument or signal cable is
unintentionally cut. There are several known solutions such as: installing a
resistor or a capacitor in parallel to C2 bushing insulation or use another voltage
limiting device inside a sensor body.
- Switching and lightning strike surge protection: During normal operation, a
bushing may see short duration surges created by a normal switching on a
substation or a lightning strike in some proximity of the bushing. Such a surge
may approach double operating voltage of a system and have duration for a
portion of a microsecond to several tens/hundreds of microseconds. In an open
circuited bushing tap, up to half of the surge may be present on the tap stressing
C2 insulation over design limits. Surge protection of a sensor must keep the
surge voltage at the tap at a safe level of several hundreds of volts. Some limiting
devices inside bushing sensors do not provide protection against a surge while
other do. As an example, a resistive load that is normally in the range of several
hundreds of ohms to several kilo-ohms which forms C-R circuit with C1
capacitance. This circuit provides voltage limitation at power frequency but allows
fast surge being almost fully applied to the C2 insulation.
- Fail safe protection. This type of additional protection ensures shorting the
bushing test tap to ground in case of all other protection layers fail.
The simplified protection circuit block-diagram is shown below in the figure 3.
Figure 3 – Simplified protection circuit of a bushing sensor
Open circuit protection consists of four voltage limiting elements that keep rms voltage at the level of 18-20V. Current equalizing circuitry is added to ensure identical operation of each circuit and override individual difference of limiters. This protection is doubled to
the requirements of the worst case scenario. Therefore only two limiters are sufficient
for normal operation. Surge protection is designed to withstand up to fifty full BIL wave hits. Circuits are also double design requirements. The fail safe circuit will short to the ground bushing tap, if all other protection circuitry fails.
Why Continuous Monitoring of Bushings
Some failure mechanisms of bushing occur very quickly. Others have voltage and
temperature dependencies that are very difficult and time consuming to simulate during
offline tests. If one is performing offline tests on bushings every three years, they are
lucky if they find a problem. Bushing problems have been known to occur with a very
short period of time measured in days to weeks.
Figure 4 shows the response of a known bad bushing over a 200 hour period with both
voltage and temperature.. Trace 1 shows the response of the bushing at 25 °C and at
10 kV. This represents normal offline testing conditions. As one can see, the power
factor readings are quite stable. Trace 2 shows the response of the same bushing at 70
°C and 10kV. This represents the temperature dependency factor. As can be seen, the
power factor is higher. Trace 3 represents the response at 25 °C and 70 kV. This
response represents the voltage dependency of the defect. Trace 4 is the response at
70°C and 70kV.
Figure 4
Response of Bushing Power factor as a Function of Temperature and Voltage
System Voltage Variations
System voltage behavior is one of the main contributors to method accuracy as a whole.
This issue becomes very critical when the precision of 0.1% is required. A variation in
system voltage (magnitude or relative phase shift between phases) creates an
unbalance and may be interpreted as a capacitance or power factor change. Magnitude
variation may be interpreted as capacitance change and phase shift variation – as
power factor.
For changes in all phases the formulas should be changed to vector summations. The
bushing monitor will react on asymmetric changes in system voltage only. All
symmetric voltage changes will compensate each other (the same increase of all
voltage magnitudes, for example, will not disturb a balance).
Therefore, accuracy of the method depends upon the statistics of the asymmetric
voltage variations in the particular location and statistical data processing procedure.
The effects of system voltage variations can be limited by using statistical processes
and correlating these fluctuations with load current.
Diagnostics
The technology as originally introduced and implemented was focused on producing
timely alarms and then suspected bushing should be further evaluated with additional
off line tests. This part remains unchanged and Sum of Current parameter is a very
reliable indicator of a dangerous trend in bushing insulation system. In addition, modern
microprocessor based instrumentation allows for additional diagnostics performed on
line while a unit is running. On line diagnostics provides additional valuable information
and therefore advantages in maintenance strategy and as a result saves money. The
main goal of on line diagnostics is to locate defective bushing, determine the
predominant failure mode and finally predict timely critical insulation triggering shut
down and bushing replacement. The diagnostics has three parts: time trend,
temperature dependencies and defect identification. Defect identification requires
determining the tangent delta and capacitance of all three bushings.
In the worst case of a single stand-alone unit installed on one three-phase transformer
(or three single-phase transformers) five independent quantities can be obtained: three
current magnitudes from the test tap and two independent phase angles between the
currents. The number of variables is twelve, three of each: tangent deltas,
capacitances, system voltage magnitudes, and phase angle between system voltage
vectors. The situation partially improves by learning the statistical behavior of the
system voltage at the particular location for a period of time and assuming that the
tangent deltas and capacitances are known at the time and equal to their off line values.
Based on the voltage behavior statistics we can then compensate for the change in the
various quantities over time.
Practical Results
Case Study 1
A bushing monitor is installed on three single-phase transformers with all three having
different bushings. The transformer operates in a peaking mode and generally has
either full load or no load. Two clusters are observed in the phasor graph (Figure. 5a)
reflecting different load modes: left – loaded and right – unloaded. Temperature
variations during the observation period are from 150C to 640C. Figure 5b shows the
variation in the sum of current rends over time.
Figure 5
Figure 6 shows the trend of power and capacitance of each bushing. Phase A power
factor is showing a slight increase which correlates with the DGA from that bushing
(Figure 7) showing slight overheating.
Figure 6
Bushing Power
Factor and Capacitance Trend
Figure 7
DGA on Oil from Phase A bushing
Case Study 2
A 650 MVA GSU unit catastrophically failed in August, 2005. In June 2005 a bushing
monitor was installed on a 650 MVA GSU which subsequently failed catastrophically in
August 2005. Unfortunately the monitor was alarming, but it was not connected to the
Plant DCS system.
Figure 8
Data from 650 MVA transformer that failed catastrophically
Figure 8 shows data from this transformer. Magnitude Alarms, trend alarms and high
temperature dependencies all indicated near term trouble.
Winding Distortion
Transformers are designed to withstand high levels of mechanical stress caused by
through faults. The stray magnetic flux will cause large forces within the winding. If
there is reduced clamping pressure or the clamping pressure is not designed to
withstand these forces, permanent winding deformation or winding collapse can occur.
Often this damage as this is not known and the transformer remains in service.
Subsequent faults most likely will cause transformer failure. Many companies have
started to use frequency response analysis measurements to determine condition of the
winding.
With the addition of three phase currents, the bushing monitor module shown in Figure
2a has the ability to monitor winding distortion on-line by monitoring and calculating the
stray reactance. The stray reactance is what determines the impedance of the
transformer.
The bushing sensors provide both voltage and phase angle information. Direct
measurement of the load current on the transformer using an auxiliary current
transformer on the secondary CT winding will provide additional needed information as
to the phase shift between the voltages and currents. This allows one to calculate the
impedance of the system and compare to factory and/or nameplate values. Figure 9
shows a block diagram of the system. A change in impedance of 3 – 5% will indicate
severe deformation.
Figure 9
Block Diagram of Bushing and Winding Health Monitor
Partial Discharge
Failures of any of the dielectrics inside of a transformer may be preceded by partial
discharge (PD). An increase in PD activity or an increase in the rate of the increase
should be cause of concern. Since partial discharge can deteriorate into complete
breakdown, it is desirable to monitor this parameter on-line. PD sources most commonly
encountered are related to moisture in the insulation, tracking on paper and barriers,
cavities in solid insulation, metallic particles, and gas bubbles generated due to some
fault condition.
Partial discharges in oil will produce hydrogen which is dissolved in the oil. However,
the dissolved hydrogen may or may not be detected, depending on the location of the
PD source and the time necessary for the oil to carry or transport the dissolved
hydrogen to the location of the sensor.
Most transformers are tested of PD activity during normal factory acceptance tests. Typical levels of PD activity are shown in Table 1.
Table 1
Common PD levels
Both electrical and acoustic PD detection each have advantages and disadvantages
and are complimentary rather than exclusive.
A partial discharge exhibits, besides other phenomena, a fast transient electrical pulse
and an acoustic "bang". Depending of the location of the PD and the coupling path
between the event and the detector, the electric or acoustic signal can be used to detect
the PD. Both methods have different detection ways and sensitivities for unwanted
signals (noise). The acoustic PD detection is most useful for events within the line-ofsight
of the acoustic transducers. This limits the detection range, but also the amount of
noise.
The electric PD detection covers a wider area, including e.g. bushing and tap changer.
External noise will also be detected and is difficult to remove. The correlation between
instrument reading and actual discharge magnitude is better than with the acoustic
method. Several international standards exist that define the instrument response,
which is the readout in pico-Coulomb or micro-Volt, allowing a better comparison
between manufacturer and in-field measurements.
Table 2 provides a comparison of both methods.
Table 2
Comparison of Electrical and Acoustical PD detection on Transformers
Electrical Method
The electrical signals from PD are in the form of a unipolar pulse with a rise time that
can be as short as nanoseconds. The pulse rise time at the origin is dependent upon
the type of discharge. Breakdown of an oil gap is a very fast process while a surface
discharge may have up to ten times longer duration. PD pulses have a wide frequency
content at the origin. The high frequencies are attenuated when the signal propagates
through the equipment and the network and pulse shape is also modified due to multiple
reflections and exciting resonant frequencies of elementary circuits. The detected signal
frequency is dependent on the original signal, pulse propagation path to the sensing
point and the measurement method.
Electrical PD detection methods are often hindered by electrical interference signals
from surrounding equipment and the network. Most common and most difficult noise
sources are aerial corona discharges and discharges to electrostatic shields that are not
properly connected to either the HV bus or ground. Any on-line PD sensing method
must have methods to minimize the influence of such signals.
The most common method for PD detection is to decouple the High Frequency partial
discharge signals using sensors that are capacitivly coupled to the HV bus (coupling
capacitor). Most HV apparatus have a natural “capacitor” built into the HV bushings or
CTs have a convenient point for connection of the PD instrument. Bushing test tap or
CT shield leads are frequently used for partial discharge measurements along with
power frequency insulation tests.
The most popular method to interpret PD signals is to study their occurrence and
amplitude as a function of the power phase position; this is called phase-resolved PD
analysis (PRPDA). This method can provide valuable insight into the type of PD
problem present.
The best method of noise rejection for in field measurements employs the use of
multiple sensors. Use of a single sensor model in the field is unlikely to produce
satisfactory results. If several sensors of different types or at different locations are
employed, the possibilities to reduce external influences are greatly enhanced.
Generally, the multi-sensor approach can be split into two processes: separate
detection of external signals and energy flow measurements.
Energy-flow measurements use both an inductive and a capacitive sensor to measure
current and voltage in the PD pulse. By the tuning of the signals from the two sensors,
they may be reliably multiplied and the polarity of the resulting energy pulse determines
whether the signal originated inside the apparatus or outside.
A modern PD instrument should employ both processes of
the multi-sensor approach allowing the comparison of PD
pulse magnitude from different sensors and pulses polarity
for energy flow measurements.
Acoustic Methods
Acoustic emissions (AE) are transient elastic waves in the
range of ultrasound, usually between 20 kHz and 1 MHz,
generated by the rapid release of energy from a source.
Partial discharges are pulse-like and cause mechanical
stress waves (acoustic waves) to propagate within the
transformer. If the stress waves propagate to the transformer
tank wall, they may be detected with a transducer that is
tuned to the right frequency.
PD sources can be located by measuring the relative time of
arrival of acoustic waves at multiple transducer locations.
In typical applications, the signals from a group of externally mounted
acoustic sensors are collected simultaneously and
analyzed to detect and locate PD. However, as the acoustic tank
signal propagates from the PD source to the sensor, it will generally encounter different
materials. Therefore, acoustic signals can only be detected within a limited distance
from the source. Consequently, the sensitivity for PD inside transformer windings, for
example, may be quite low.
Figure 10 – Acoustical sensors installed on transformer
Though not disturbed by signals from the electric network, external and internal
influences in the form of rain or wind and non-PD vibration sources like loose parts,
cooling fans and oil flow from transformer oil circulating pumps will generate acoustic
signals that interfere with the PD detection. These non-PD acoustic signals may extend
up to the 50 to 100 kHz region. To diminish the effects of this disturbance, acoustic
sensors with sensitivity in the 150 kHz range are usually employed. Such sensors may,
however, have less sensitivity to PD signals than lower frequency sensors.
Figure 11 (a) shows an on-line continuous PD monitor that is available for use on
motors, generators, switchgear, cables and transformers. It uses the same bushing
sensors as the BHWM sown in Figure 2. An additional radio frequency current
transformer mounted on the neutral bushing bus or cable or on the tank ground for
noise cancellation purposes. Figure 11 (b) shows a complete monitoring system
covering all aspects covered in this paper, which includes a power supply, Main CPU
with communications and other I/O, Bushing and Winding Health Monitor (BHWM) and
a 15 channel PD Module (PDM).
Figure 11
Summary
As technology advances, additional continuous or on-line monitoring techniques are
being developed and proven the in the field. This paper presented some newer on-line
diagnostics methods and provided examples and methods of monitoring and
diagnostics for bushings and windings on transformers.
Product Link
Technical Paper
Application Note
Contact our application engineers for further details click here.
