Framit3 Ieee Paper

7/27/2019 Framit3 Ieee Paper

1/29

Copyright Starlogic 2004 IEEE FRA SpecificationStarlogic IFRA Submission version 1.0

October 2003

Frequency Response Analysis using theImpulse Frequency Response Analysis

(IFRA) Method

Starlogic Instrument Development

South Africa

COPYRIGHT OF THIS MANUAL IS RESERVED. NO PUBLICATION OR DISSEMINATIONOF ITS CONTENTS IS ALLOWED WITHOUT WRITTEN PERMISSION FROM AUTHOR.


2/29


October 2003

Winding Frequency Response Analysis using the Impulse Frequency

Response Analysis (IFRA) MethodRichard BreytenbachStarlogic

October 2003COPYRIGHT OF THIS PAPER IS RESERVED. NO PUBLICATION OR DISSEMINATION OF ITS CONTENTSIS ALLOWED WITHOUT WRITTEN PERMISSION FROM AUTHOR.

1 Scope ........................................................ ....................................................... .......... 22 Introduction........................................................................... ..................................... 2

3 Causes of Transformer Deformation ...................................................... ................... 24 Transformer Winding Modelling................................ ............................................... 35 Frequency Response Analysis (FRA)..................................................... ................... 4

5.1 IFRA Test Method ................ .................. ................. .................. .................. ................. .........6 5.2 SFRA Test Method .................. .................. ................. .................. .................. ................. ......7

6 FRAMIT Instrument................................................... ............................................... 77 FRA Test Process ....................................................... ............................................. 11

7.1 Performing the FRA test. .................. ................. .................. .................. .................. ............11 7.2 Generating the FRA test results. ................ ................. .................. .................. ................. ....12

8 FRA Test Equipment .................................................. ............................................. 148.1 IFRA.....................................................................................................................................14 8.2 Test data ................. .................. .................. ................. .................. .................. ................. ....14

9 Application of FRA test technology ....................................................... ................. 1510 Interpretation of FRA Test Results................................................ .......................... 16

10.1 FRA example of shorted windings / partial discharge ................. .................. ................. ....16 10.2 FRA example of shorted & burnt windings ............... .................. .................. ................. ....19 10.3 FRA example of localized DISTORTED WINDINGS................... .................. ................. .20 10.4 FRA example of general DISTORTED WINDINGs................ .................. ................. .......21 10.5 Laboratory Experiments: Arizona State University.............................................................21 10.6 Laboratory Experiments: ESKOM TRI South Africas National Electricity

Supply Utility. ................. .................. ................. .................. .................. .................. ............23 10.7 Laboratory Experiments: Institute of Energy Technology from the Aalborg

University in Denmark.........................................................................................................23 10.8 Substation Experiments: CLECO Central Louisiana Electric Company..........................23 10.9 Substation Experiments: ESKOM TRI South Africas National Electricity

Supply Utility .................. .................. ................. .................. .................. .................. ............24 10.10 Substation Experiments: NGC National Grid Company plc UK...................................25 10.11 Substation Experiments: EPRI Electric Power Research Institute USA ................. .......26

11 Summary of Experiments and Field testing..................................................... ........ 2612 Conclusion ................................................ ....................................................... ........ 2713 References................................................................... ............................................. 28


3/29


October 2003

Page 2 of 24

1 SCOPE

This paper covers the background to Frequency Response Analysis (FRA) testing, and details therequirements and specifications needed to perform an FRA test and generate a practical FRAguideline.

At present, FRA testing can be done using 3 different testing methods, namely;

IFRA Impulse Frequency Response Analysis,SFRA Sweep Frequency Response Analysis,PRBS Pseudo Random Binary Sequence ( still a very new technique )

It should be noted that although these 3 methods generate the same FRA test results, all 3methods use very different measurement techniques to calculate the FRA test result. This meansthat if a general FRA guideline and specification is to be generated, it should be such that itencompasses all recognized techniques and remains open ended to allow for future FRA researchand development.

This paper deals specifically with the Impulse Frequency Response Analysis ( IFRA ) method.

2 INTRODUCTION

Power transformers constitute the core of electrical power transmission and distributionnetworks. They are also the most expensive equipment within the substation. Their performancewill determine, to a large extent, the quality of power supply. It is therefore necessary tocontinuously monitor and assess the condition of transformers in order to ensure reliability andavailability of power supply.

Frequency Response Analysis (FRA) is becoming an increasingly popular technique used toexternally monitor and assess the condition and mechanical integrity of transformer windings.This FRA technique calculates and computes frequency-dependent variables of the transformerswindings, i.e. inductance and capacitance. It is these distributed winding parameters that willchange when the windings are; short-circuited, open-circuited, deformed, or loose. The FRA testis performed by injecting a low voltage impulse waveform into one end of a transformerswinding and measuring the voltage appearing at the other end of the same winding. The ratio of the transfer function of the input voltage to the output voltage is then plotted on a frequencydomain graph. This paper demonstrates that FRA can be used and applied as a reliabletransformer condition-monitoring tool to help assess winding condition without the need to openthe transformer for inspection. The FRA technique can help maintenance personnel identifysuspect transformers, enabling them to take those transformers out of service before failure.

3 CAUSES OF TRANSFORMER DEFORMATION

Power transformers are usually very reliable, but when faults occur, the transformer can beaffected catastrophically. Transformers fail in service each year. Most of these failures are caused


4/29


October 2003

Page 3 of 24

by transformer winding faults and through faults generated by lightning and switching surges. Asa transformer experiences a fault, it may suffer mechanical shock that gradually displaces anddistorts the windings. In the process of winding movement, the insulation between the turns can

be abraded, causing a short circuit and damage to the windings. Mechanical vibrations, initiated by short circuit forces, may cause the windings to loose their clamping pressure, eventually

leading to collapse of the windings. The other cause of winding movement may be extensivevibration during transformer transportation. As the windings experience vibration, they mayslacken and subsequently become unable to withstand mechanical forces exerted during faults.Ageing also contributes to winding looseness. In addition, harmonics generated under normaloperating conditions may cause winding and core vibration.

Short circuit faults are potentially very destructive because if the clamping pressure is notcapable of restraining the forces involved, substantial permanent winding deformation or evencollapse can occur almost instantaneously, often accompanied by shorted turns. A common causeof failure is a close-up phase to earth fault resulting from a lightning strike.

It is expected that a transformer will experience and survive a number of short circuits during itsservice life, but sooner or later one such event will cause slight winding movement, and theability of the transformer to survive short circuits in future will then be severely reduced. As thetransformer ages, its components deteriorate and the likelihood of a failure increases.

4 TRANSFORMER WINDING MODELLING

For a clearer understanding of what actually happens when a transformers winding structureundergoes deformation, we need to analyze the equivalent circuit model of a transformerswindings as shown in the figure below. Specifically, we are examining the effects of the small

parasitic capacitance and inductance of the windings of the transformer.

Transformer Winding ModelingTransformer Winding Modelingshowing parasitic inductance, capacitance and resistanceshowing parasitic inductance, capacitance and resistance

CrossCross --section of a transformersection of a transformer

CC

CCCC

R R

LL

LL

Tank WallTank Wall

WindingWindingR R

CCCC

CC

WindingWinding

CoreCore

Cross-section of a transformer showing equivalent circuit model


5/29


October 2003

Page 4 of 24

Malewski, Gockenbach, Maier, Fellmann & Claudi 1 explain that at the first approximation atransformer winding can be represented by a ladder network with series inductance andcapacitance as well as the parallel capacitance to ground. The transfer function of such a network calculated by FRA shows a number of poles at the resonance frequency of the local L and C

circuits. A breakdown between the turns or coils of the transformer under test corresponds to ashort circuit of one or more of these local LC networks. This will result in shifting the resonant pole to another frequency or the creation of a new pole. A partial discharge condition with thetransformer will not significantly affect the resonant frequency of the affected pole, but the poleheight will be reduced. Consequently, a change of the resonant pole frequency indicates a

breakdown in the winding insulation, whereas a reduction in the pole height reveals partialdischarge activity.

Tlhatlhetji 2 argues that the detection of a transformer winding condition that may lead to a faultand subsequent outage, and being able to take corrective action prior to failure (e.g. reclampingor insulation repair), could save a transformer rewind, estimated to cost in the region of $1-million for a large power transformer. The failure of a strategic transformer on the power supplynetwork can be even more costly, considering consequential factors such as systemdestabilization, load shedding, outages and even complete system shutdown.

These factors have led to the search for alternative methods of transformer winding conditionmonitoring and assessment such as Frequency Response Analysis (FRA). FRA is a sensitivetechnique that can be used non-intrusively to accurately predict the condition of transformer windings.

5 FREQUENCY RESPONSE ANALYSIS (FRA)

FRA technology was developed as a diagnostic tool for application to power transformers. Theresults of research showed that the FRA method of testing was reliable and repeatable for thedetection of winding deformation.

There are currently two popular techniques for performing Frequency Response Analysis tests.The first is the voltage impulse method ( IFRA ), and the second is the sweep frequency method(SFRA ). The Department of Technology Research and Investigation (TRI) from South Africasnational electricity supply utility ESKOM undertook a research project to compare the IFRAfrequency response analysis technique against the swept frequency method.

Comparison tests were performed by Tlhatlhetji 3 on a 20 MVA, 66/11 kV transformer. With bothinstrument test leads connected correctly to the transformer, the two sets of test results showedvery close correlation. It was also evident from the results that the resonances lie at the samefrequencies. Further comparison tests were performed on a small 16 kVA, 11kV/400V, single-

phase transformer at the University of Stellenbosch in South Africa. These test results also

1 Malewski, Gockenbach, Maier, Fellmann & Claudi (1992, p1)2 Tlhatlhetji (1999, p4)3 Tlhatlhetji (1999, p29)


6/29


October 2003

Page 5 of 24

showed very close correlation. The same report also detailed comparative tests using IFRA,Pseudo-Random Binary Sequence, and frequency sweep. It was conclusively proven thatfrequency response measurement using these three techniques were almost identical.

Below are some comparison FRA tests between IFRA and SFRA test techniques, performed on

another transformer;IFRA Test Result: Transformer SN: T2300

SFRA Test Results: Transformer SN: T2300


7/29


October 2003

Page 6 of 24

IFRA Test Results: Transformer SN: T2300

SFRA Test Results: Transformer SN: T2300

The following observations and comparisons can be made between the 2 different FRA testingtechniques.

5.1 IFRA TEST METHOD

Description:The test equipment consists of a portable self-contained instrument. The 50 coaxial test leadsare connected across a specific phase winding of the transformer. The IFRA equipment injects a

voltage impulse of about 450 volts into this phase winding. A high-speed digital data capture unitsamples the injected voltage signal, as well as the voltage signal appearing at the other end of thewinding. These two recorded time domain signals are processed and the FRA trace is displayed.Because of the nature of the IFRA Impulse technique, the test technique is very quick, takingonly a few minutes to generate the FRA traces. This makes this technique ideal for fieldapplication where there may be volatile conditions such as bad weather, intermittent highfrequency noise pollution, and little time to perform the test due to traveling time.


8/29


October 2003

Page 7 of 24

5.2 SFRA TEST METHOD

Description:The test equipment consists of a portable self-contained instrument. The 50 coaxial test leadsare connected across a specific phase winding of the transformer. The SFRA equipment injects a

sinusoidal waveform of constant magnitude of 10 Vrms into the phase winding. The equipmentmeasures the injected voltage signal, as well as the voltage signal appearing at the other end of the winding. Because of the nature of the SFRA technique, the test technique is very slow, takingabout 2 1/2 hours to generate the FRA traces. This makes this technique more suited to laboratoryapplication where there is a more controlled environment.

6 FRAMIT INSTRUMENT

South Africas national supply utility, ESKOM, though its research into FRA initiated thedevelopment of a specialized FRA measurement tool. Starlogic was commissioned by ESKOMto develop a complete, self-contained instrument that performs Frequency Response Analysis of

power transformers. The Starlogic instrument called FRAMIT is the result of accumulatedknowledge from the research and application of this test technology over the past 10 years.FRAMIT has been internationally marketed with great success for over the past 7 years. Theinstrument is housed in a rugged carry case in which the measurement electronics, connectingleads and manual are fitted to create a complete, portable instrument as shown in figure below.The software runs under Microsoft Windows 95/98/2000/ME/XP & NT and therefore inherits allof features of this operating system.

FRAMIT Instrument


9/29

UNITED STATESAEP CANTON, OHAEP CENTRAL POWER & LIGHT CORPUS CHRISTI, TXATLANTIC COAST TESTING RALEIGH, NCBALTIMORE GAS & ELECTRIC BALTIMORE, MDCENTER POINT ENERGY HOUSTON, TX (6)CITY PUBLIC SERVICE SAN ANTONIO, TX (2)DELTA STAR LYNCHBURG, VADENTON MUNICIPAL ELECTRIC DENTON, TXEL PASO ELECTRIC EL PASO, TXENTERGY CORPORATION NEW ORLEANS, LA (2)GARLAND POWER & LIGHT GARLAND, TXKUHLMAN ELECTRIC CORP CRYSTAL SPRINGS, MSLOUISIANA GENERATING,LLC NEW ROADS, LALOUISVILLE GAS & ELECTRIC LOUISVILLE, KYMIRANT MID ATLANTIC - NEWBURG, MDNIPSCO MERRILLVILLE, INDIANAOKLAHOMA GAS & ELECTRIC OKLAHOMA CITY, OKORLANDO UTILITIES COMMISSION ORLANDO, FLPEPCO WASHINGTON, D.C.PG&E SAN FRANCISCO, CA (2)PORTLAND GENERAL ELECTRIC PORTLAND, ORPSI CINERGY PLAINFIELD, IN (3)SALT RIVER PROJECT - PHOENIX, ARIZONASOUTHERN CALIFORNIA EDISON ROSEMEAD, CA (2)TXU ELECTRIC ONCOR - FORT WORTH, TX (12)WESTERN FARMERS ELECTRIC COOPERATIVE ANADARKO, OK (2)WAUKESHA ELECTRIC SYSTEMS WAUKESHA, WI (3)

SOUTH AFRICA: ESKOMSOUTH AFRICAN ELECTRICITY SUPPLY AUTHORITY (12)TECHNOLOGY SERVICES INTERNATIONALSPOORNET

CHINA: DEXIN (8)

MEXICO: CFE - COMISION FEDERAL DE ELECTRICIDAD (2)

JAPAN: MITSUBISHI TRANSFORMER JAPAN

AUSTRIA: VA TECH ELIN - WEIZ, AUSTRIAVA TECH EBG LINZ, AUSTRIA

GERMANY: SGB STARTSTROM GERATEBAU GMBBH

FRAMIT INTERNATIONAL USERS LISTDate: April 2004

2004 Starlogic cc


10/29

Operating ConditionsOperating main: 85 - 260 Vac (60 - 50Hz)Max Voltage Withstand: 275 VoltsImpulse Voltage Withstand: 1 kV, 1.2/50 sProtection: Electrostatic Discharge ImmuneOperating temperature: 14F to 140F (-10C to 60C)Max relative humidity: 10% - 90%

Weight: 28.67 lbs (13 Kg)Enclosure: Weatherproof Pelican carrycaseEnsclosure Dimensions: 14.25 x 15.75 x 6.69

(362mm x 400mm x 170mm)Measurement System Calibration: Self Calibration

Amplitude Range: 0 db to 90 dbMeasurement frequency band: 150 Hz - 1 MHz, 10% of active frequencyMeasurement accuracy: 0.1% of full scaleMaximum sampling rate: 10 MHzResolution: 12 bits

Analog input ranges: (+-V) 0.1, 0.2, 0.5, 1.2, 5, 10, 20

Capture memory: 2 x 12 bits, 128K deepMeasurement differentiation: 0.01dBCommon-mode rejection ability: >50dBDisturbance rejection between channels: >120dBMeasurement range: +/- 600VDCOutput Impulse: 500V, 40 SCharge Time: 2 secondsOutput Impedance: 50 OhmsImpulse repeatability: 50ppmImpulse stability: 100ppm

TECHNICAL SPECIFICATION

2003 Starlogic cc

FRAMIT3


11/29


12/29


October 2003

Page 11 of 24

7 FRA TEST PROCESS

7.1 PERFORMING THE FRA TEST.

IFRA Test MethodTest preparationPeforming a FRA test must be done in a safe and controlled manner. The transformer under testmust be removed and isolated from the live system and securely grounded. The local safetyregulations and guidelines must be adhered to. The transformer must be tested in as basic a state

as possible. This means that any external circuitry such as bushing connections etc etc must beremoved where physical practical.

Transformer under testAny size transformer can be tested using IFRA. The transformer can be tested in any state, i.e.with or without oil, with or without bushings, and on any tap setting. It is however important torealize that any change to any of these factors will change the equivalent circuitry of thetransformer under test, and will therefore affect the FRA test results in some way. Thetransformer must be tested with the tap setting set to that tap which exposes all the windings tothe test process. Any unused bushings must be left disconnected (floating).

Test setThe FRA equipment should be grounded and the appropriate calibration performed beforecommencing with a test.

Test leadsThe test leads must consist of a 3-cable system of the same length. I.e. one for injecting the inputsignal into the transformer winding, one to receive back this input signal from the winding, and


13/29


October 2003

Page 12 of 24

one to receive back the signal from the other side of the winding. The ground test lead should beapplied firmly to the metal surface of the transformer under test. This will ensure good groundingof the test signal. The test leads must be visibly inspected to ensure that there is continuity andshielding down the length of lead.

Test procedureThe test-leads are connected across a specific phase winding of the transformer, as shown above.The FRA Equipment injects a voltage impulse of about 450 volts into this phase winding. Ahigh-speed digital data capture unit samples the injected voltage signal, as well as the voltagesignal appearing at the other end of the phase winding. These two recorded time domain signalsare processed and the FRA trace is displayed.

7.2 GENERATING THE FRA TEST RESULTS.

The FRA equipment takes these two time domain signals and calculates the frequency spectrum.It does this by applying a Fast Fourier Transform Algorithm (FFT) to the time domain signals.This yields the frequency spectrum. The frequency response is then determined by dividing thefrequency spectrum of the output signal by the frequency spectrum of the input signal. This isshown in the figure below.

FRA MathematicsFRA Mathematics

300

INPUTINPUT

40

0

Time (uS)

Volts

200

x(t)

X(f)X(f)

FFTFFT

OUTPUTOUTPUT

y(t)y(t) Y(f)Y(f)FFTFFT

300

150

Volts

Time (uS)0 200

FREQUENCY RESPONCEFREQUENCY RESPONSEFINGERPRINT

H(f) = Y(f) / X(f)H(f) = Y(f) / X(f)

0.6

0.3

0 200

Frequency (Hz)Frequency (Hz)

MagMagH(f)

X(f)X(f)

IFRA Mathematics using Fast Fourier Transformer (FFTs)

This test is repeated for each phase winding of the transformer (6 tests for a 3-phase transformer).This allows each winding of the transformer to be independently inspected and evaluated. Thefrequency response of all the tested phases together is called the fingerprint of the transformer.This fingerprint is unique to every transformer, and remains unchanged for as long as thetransformers winding structure remains unchanged.

The FRA theory is as follows: The IFRA input data capture channel is connected directly to theIFRA on-board impulse generator. The input data capture channel will therefore record the signaldelivered by the impulse generator. This voltage signal delivered by the impulse generator isinjected into the selected phase winding of the transformer. The injected voltage creates a currentto flow in the windings. The flow of current sets up a magnetic field. Once the impulse generator


14/29


October 2003

Page 13 of 24

has discharged, this magnetic field starts to decay. This induces an EMF voltage in the samedirection to attempt keep the current flowing (Lenz's Law). The only path for the current to nowflow is into the auxiliary capacitances of the windings themselves. This current charges thesecapacitances up until the potential on the capacitances is equal to the EMF voltage of thecollapsing field. When no more current can flow then the capacitances begin to discharge back

into the windings, setting resonance. The resonant frequency is 1/2.pi.f.c. Resonance decays asthe energy is dissipated into the resistance of the windings. In essence this is what is measured bythe IFRA and is defined as the "fingerprint" of the winding. Any physical change to the windingstructure will cause the winding's inductive and capacitive properties to change, and these can bedetected by Frequency Response Analysis. If there are any external windings near the winding

being tested they will form part of the resonant circuit. Any built-in CT's etc. etc. will remain as part of the transformer setup allowing future IFRA tests to remain consistent.

Typical IFRA test results showing Input and Output voltages and calculatedFrequency Response


15/29


16/29


October 2003

Page 15 of 24

Sort database according to transformer high side/low side voltage, Sort database according to transformer location, Sort database according to transformer winding organization,

9 APPLICATION OF FRA TEST TECHNOLOGYThe international use of FRA test technology is growing steadily. Many companies in the USAand internationally are successfully implementing IFRA testing. These companies haveauthorized that IFRA tests be performed; on all their new transformers before they are purchased from the manufacturer, after they are transported to site, once they have been commissioned, with future routine tests being performed on a regular basis.

This decision is enhancing their transformer preventative maintenance programs.

Before a transformer is loaded onto a truck for delivery to the customer, a FRA test is performedat the manufacturers premises. Once the transformer has been delivered, it is tested again. If thetwo fingerprints match well, it means that there has been no movement of the winding structureduring transport and loading. If a transformer suffers a high through fault condition, it can betested to determine the extent of the distortion of the winding structure. The test results willindicate what further action (if any) needs to be taken. This can save a lot of time, as well astransformers. Transformers can also be tested periodically, to determine the cumulative effects of high through current faults.

As has been discussed earlier, IFRA and SFRA produce the same test results. However, due tothe nature of the IFRA Impulse technique, the test technique is very quick, taking only a fewminutes to generate the FRA traces. This makes this technique ideal for field application wherethere may be volatile conditions such as bad weather, intermittent high frequency noise pollution,and little time to perform the test due to traveling time. Because of the nature of the SFRAtechnique, the test technique is very slow, taking about 2 1/2 hours to generate the FRA traces.This makes this technique more suited to laboratory application where there is a more controlledenvironment.


17/29


October 2003

Page 16 of 24

10 INTERPRETATION OF FRA TEST RESULTS

Over the years significant research has been conducted by companies such as Eskom, Starlogic,TRI, EPRI, Doble and NGC into the correlation between the mechanical winding deformation of

transformers and FRA test results. This research will be examined in greater depth.

10.1 FRA EXAMPLE OF SHORTED WINDINGS / PARTIAL DISCHARGE

The transformer was transported to the repair facility in, Ohio and de-tanked. It was observed thatwater had gotten into the transformer through a failed gasket on the pressure relief device. Thisis located right above the phase C coils. These coils were covered with a sludge that resultedfrom the water mixing with the oil. After cleaning off of sludge, a second IFRA test was

performed (with the transformer outside of the tank).

Sludge covering windings

Sludge at the bottom of the tank


18/29


October 2003

Page 17 of 24

Transformer: SN: 3081107HIGH SIDE - BEFORE CLEANING OFF SLUDGE ON H2 WINDINGS (Yellow Trace)

HIGH SIDE - AFTER CLEANING OFF SLUDGE ON H2 WINDINGS (Yellow Trace)


19/29


October 2003

Page 18 of 24

Transformer: SN: 3081107LOW SIDE - BEFORE CLEANING OFF SLUDGE ON X3 WINDINGS (Blue Trace)

LOW SIDE - AFTER CLEANING OFF SLUDGE ON X3 WINDINGS (Blue Trace)


20/29


21/29


October 2003

Page 20 of 24

10.3 FRA example of localized DISTORTED WINDINGS

The oil sample for DGA analysis was sampled on date 8/30/2002 was taken as the results of agas detector relay alarm. Being as it was Friday before a three day weekend, we de-energized thetransformer. The following Tuesday the sample date 9/03/2002 was taken. There was anincreased gas level, even though the transformer was only energized for an additional three (3)hours!

The Doble tests showed the % power factor of the low voltage windings doubled from .39% to.75%. The transformer turns ratio and the bridge tests showed no changes. The core groundstested good (> 100 Mohm) and an internal inspection did NOT reveal any problems. The other data available was the occurrence of 39 through faults in the previous month from the connectedfeeders. Tranformer loading was well below nameplate.

The transformer was torn down at the factory and visible inspected. There was windingdistortion (petal failure partial axial collapse) of the X3 winding structure. Please see attached

picture.

Transformer: SN: MLL 9315-3LOW SIDE - LOCALIZED WINDING DISTORTION TO H3 WINDING STRUCTURE (Blue Trace)


22/29


October 2003

Page 21 of 24

10.4 FRA example of general DISTORTED WINDINGs

1975 Year old transformer with low clamping pressure with general winding distortion.Transformer: SN: 10552

HIGH SIDE

LOW SIDE - GENERAL WINDING DISTORTION TO X3 WINDING STRUCTURE (Blue Trace)

10.5 LABORATORY EXPERIMENTS: ARIZONA STATE UNIVERSITY.

The department of Electrical Engineering from the Arizona State University initiated a research project to investigate and evaluate the IFRA performance in transformer fault detection. A smalllaboratory 1kVA 800/250V transformer was repeatedly tested using a variety of different testingequipment. The small transformer was subjected to simulated electrical mode faults and wasroutinely tested after each different type of fault. The modes of failure that were simulatedincluded within the laboratory were:

Turn-to-turn short circuit


23/29


October 2003

Page 22 of 24

Layer-layer shorts Coil-coil shorts Open-circuited windings Partial discharges etc.

SENSITIVITY AND REPEATABILITY ANALYSIS.The research report conducted by the Department of Electrical Engineering from the ArizonaState University by Dr. Karady, Dr. Reta & Amarg 4 aimed to quantify the relative change inresonant frequency location (Df) and relative change in amplitude (DA), with respect to thevalues for normal operation, as a result of the simulated modes of failure. In table 2, frequency

peak location (kHz) refers to the frequency at which the maximum value of the transfer functionis observed. Peak magnitude is the magnitude of the transfer function at the peak frequency and

peak range is the percentage change in peak magnitude for different experiments but same faulttype. Relative changes will help distinguish the various types of failure and also provide anindication for test repeatability. Dr. Karady, Dr. Reta & Amarg 5 documented the results of thesensitivity analysis in table 2. The relative changes are computed as

f and 100*A nin

ni f f D A A A D == (1) & (2)Where A n and f n are respectively, the magnitude and resonant frequency location for fingerprint,

Ai and f i are the magnitude and resonant frequency location for all other simulated conditions.

Fault Type Frequency PeakLocation, f (kHz)

Peak Magnitude,A (units) x 10 exp 3

PeakRange (%)

DA (%) Df (kHz)

Normal 525 52.0 7.7 0.0 0.0

Coil Lifted 525 92.0 0.0 76.9 0.0

2 coils shorted 520 21.0 16.0 -59.6 -5.0

5 coils shorted 540 25.0 16.0 -51.9 15.0

10 coils shorted 540 26.4 4.5 -49.2 15.0

20 coils shorted 540 24.0 0.0 -53.8 15.0

2 primary coilsShorted

600 57.0 3.5 9.6 75.0

Sensitivity Analysis for FRAMIT Instrument

Department of Electrical Engineering

from Arizona State University

DA=A i Ar * 100 and Df = f i - fr Ar

Sensitivity Analysis for IFRA equipment

It was concluded that the FRA method of transformer testing using the IFRA equipment isreliable and repeatable for the detection of simulated transformer winding movement.

4 Dr. Karady, Dr. Reta & Amarg (1999, pg 12)5 Dr. Karady, Dr. Reta & Amarg (1999, pg 13)


24/29


October 2003

Page 23 of 24

10.6 LABORATORY EXPERIMENTS: ESKOM TRI SOUTH AFRICASNATIONAL ELECTRICITY SUPPLY UTILITY.

The Department of Technology Research and Investigation (TRI) from South Africas nationalelectricity supply utility ESKOM conducted laboratory experiments using IFRA on a small25kVA; 11kv/400v; three-phase pole-mounted transformer. Tests were performed with thetransformer un-tanked. At lower frequencies the test results show that the transformer ischaracterized by the magnetizing inductance and non-linear behavior of the iron core. Peak resonances were seen to occur in a few places.

Tlhatlhetji 6 explains that the major differences in the responses occur at the lower frequencies, upto 160kHz. Multiple resonances occur for the first short-circuit (yellow trace) between 0 and 40kHz. It can be seen that the responses under short-circuit conditions, not only differ from thereference fingerprint, but also from each other. Frequency responses of short circuit faults at theminimum and nominal voltage taps differ both in magnitude and phase up to 56 kHz. However,from 160 kHz upward all three-frequency responses are very similar. It appears then that thedistributed capacitance of the winding may be dominating at higher frequencies more that

inductance. It can be concluded, that short circuit faults are easily detected in the lower frequencyrange.

10.7 LABORATORY EXPERIMENTS: INSTITUTE OF ENERGY TECHNOLOGYFROM THE AALBORG UNIVERSITY IN DENMARK

The Institute of Energy Technology from the Aalborg University in Demark carried out FRAtests on a two small 30VA; 20kv/110V; single phase transformer. The transformer wasmanipulated in a variety of different ways to simulate fault conditions. FRA tests were performedon the transformer after each set simulations. The object was to establish the correlation betweendifferent simulated fault conditions and actual changes in the frequency response. Differentconditions were applied to the insulation oil/grease, the core, and the winding/coil insulations.

These included a series of simulated accelerated ageing tests. Many different failure conditionswere successfully detected using FRA testing.

10.8 SUBSTATION EXPERIMENTS: CLECO CENTRAL LOUISIANA ELECTRICCOMPANY

In early May of 1999, the Central Louisiana Electric Company, CLECO, were having one of their generator step-up transformers re-blocked. Mr. Sandy Crochet of CLECO requested that IFRAtests be performed on their transformer prior to it being re-blocked. A re-test was also scheduledabout 2 weeks later, after the re-blocking work was complete. As was expected, this particular transformer showed significant FRA result changes when it was re-tested after its re-block.

The FRA differences between the reference test and the test that was performed after thetransformer was re-blocked. It was seen that the response of the re-bock test was completelydifferent from that of the reference test. The deviation is clearly notable above 153 kHz. Thisshows clearly that every winding structure has a unique FRA fingerprint. A re-blocking of a

6 Tlhatlhetji (1998, p11)


25/29


October 2003

Page 24 of 24

transformer will generate a new set of fingerprints since the newly tensioned winding structurewill have its own unique ladder network of series inductance and capacitance as well as parallelcapacitance to ground. This new FRA test that was performed after the re-block will now becomethe reference test against which future routine tests will be compared.

10.9 SUBSTATION EXPERIMENTS: ESKOM TRI SOUTH AFRICAS NATIONALELECTRICITY SUPPLY UTILITY

The Department of Technology Research and Investigation (TRI) from South Africas nationalelectricity supply utility ESKOM performed IFRA tests on many of their large power transformers.

One of the transformer had faulted in service. Tlhatlhetji 7 documents that the fault was caused bya faulty tap-changer on the white phase. He goes on to say that IFRA tests were conducted withthe transformer still inside the tank. There was no oil in the transformer and the bushings wereremoved. No reference FRA tests had been conducted on the transformer. The yellow phase wascompletely different from that of the red and blue phases. The deviation was clearly notable as

from 160 kHz. Whereas the peak resonance occurs at 210 kHz for both the red and blue phases,the yellow phase is flat, which is similar to an open-circuit response. Although there is noreference fingerprint for this transformer, the response shown below for the red and blue phasescompared very favorable with a similar sister transformer. This gives conclusive evidence thatthe yellow phase is faulty. The tap-changer was inspected and found to be defective of theyellow-phase.

Frequency response of a 220 MVA, 420/16.5 kV, generator step-uptransformer that faulted on the yellow tap-changer

7 Tlhatlhetji (1998, p15)


26/29


October 2003

Page 25 of 24

10.10 SUBSTATION EXPERIMENTS: NGC NATIONAL GRID COMPANY PLC UK

The Nation Grid Company plc from the United Kingdom have performed FRA tests on manylarge power transformers. An overview of their general FRA findings is included below:

When examining FRA test results, Lapworth & McGrail8

commented that relatively simpleresponses are obtained at the lowest frequencies, irrespective of winding design. In this frequencyrange the inductance of the winding is influenced by the core. The different iron flux pathsinvolved when outer and center phases of a three-phase transformer are excited explain why it isnormal for the center phase to have a different response. Changes to the core limb remnantmagnetizations can also affect these low frequency responses. An extreme case of core limbreluctance occurs when a shorted turn exists in one of the windings on that limb, and in such acase a very characteristic change in the low frequency response is obtained.

At higher frequencies, Lapworth & McGrail 9 commented that eddy currents effectively screen themagnetic circuit so that the winding inductances are determined by local leakage fluxes. Beingless dependent on magnetic characteristics, the winding responses of the 3 phases become muchmore similar and therefore more sensitive to winding movement. There is no typical form for thehigh frequency responses, which are characteristic of the winding design used and vary greatlyfrom design to design. Interpretation of FRA test results is based upon a subjective comparisonof equivalent responses. Measured responses are analyzed for:

Changes in the response of a windingDifferences between the responses of the three phases of the same transformer,Differences between the responses of transformers of the same design.

The appearance of new features or major frequency shifts is cause for concern. When interpretingany differences observed between phases, Lapworth & McGrail 10 mention that it should be bornein mind that for many windings there may be minor design differences concerning the dispositionof internal connections between windings, bushings and tap-changers which could introducesmall differences between phases in the frequency responses. Therefore, if minor differences

between phases are observed and no reference results are available, it is not possible to make anunambiguous diagnosis of minor winding movement.

Extensive testing by Lapworth & McGrail 11 of NGC on a number of their transformers hasresulted in the successful diagnostic of the following independent transformer conditions:

Collapse of end insulation structuresHoop buckling failure of inner windingsBroken winding clamping structure / displacement of winding assembliesTap-changer failure

8 Lapworth & McGrail (1999, p3)9 Lapworth & McGrail (1999, p3)10 Lapworth & McGrail (1999, p3)11 Lapworth & McGrail (1999, p4-5)


27/29


October 2003

Page 26 of 24

NGC have concluded that the confidence in the interpretation of FRA results is continuing togrow as experience of detecting problems increases.

10.11 SUBSTATION EXPERIMENTS: EPRI ELECTRIC POWER RESEARCHINSTITUTE USA

The Electric Power Research Institute in the USA has performed FRA tests on many large power transformers. An overview of their general FRA findings is included below:

From 1997 to 1999 EPRI tested 7 large power transformers of identical design. The rating of thetransformers is 345 kV, 300 MVA, and 3 phase. They were built over 25 years ago with minor variations in internal design. Vandermaar, Wang, Stefanski & Ward 12 remarked that FRA testswere performed on these seven transformers with the signatures compared to one of thetransformers that had been reclamped relatively recently. Recommendations were made on whichtransformers to open. The transformer, which was ranked first for reclamping has now beenopened and reclamped. When it was opened it was found to be loose.

11 SUMMARY OF EXPERIMENTS AND FIELD TESTING

International FRA research has proven that it is possible to detect a variety of different internalconditions. It can detect a change in the condition and quality of the transformer oil/grease, if thisinvolves a change in the dielectric constant . It is possible to detect a change in the condition of the core, if this involves a change in the total reluctance or total iron losses of the core. It is

possible to detect a change in the condition of the winding insulation and winding structureintegrity, if this involves changes in the lumped/total winding capacitances of the transformer or short-circuit between turns.

The international FRA research has shown that it is possible to detect the following internalmechanical deformation of transformers: Multiple core earths Loss of clamping pressure Shorted Turns (two) & Shorted Turns (Multiple) Magnetized Core Raised Winding & Loose Turns Core Movement & Collapse of end insulation structures Hoop buckling failure of inner windings

Broken winding clamping structure / displacement of winding assemblies Tap-changer failureAt lower frequencies the FRA test results (fingerprints) show that the transformer ischaracterized by the magnetizing inductance and non-linear behavior of the iron core. Thedistributed capacitance of the winding may be dominating at higher frequencies more that

12 Vandermaar, Wang, Stefanski & Ward (1999, p4)


28/29


October 2003

Page 27 of 24

inductance. It can be concluded, that short circuit faults are easily detected in the lower frequencyrange.

At higher frequencies, eddy currents effectively screen the magnetic circuit so that the windinginductances are determined by local leakage fluxes. Being less dependent on magnetic

characteristics, the winding fingerprints of the 3 phases become much more similar and thereforemore sensitive to winding movement.

Many years of testing experience with constant feedback of FRA test results from customers hasallowed developers of IFRA to develop and refine new and more sophisticated ways to interpretthe test results. Test data and research obtained over the last 3 years, has shown that thefrequency range of 0 500 kHz contains the essential indicators which identify the variousmodes of transformer failure. Over this range one can determine the whole range of transformer failure modes including short circuits, localized winding distortion, and general windingdistortion and loss of clamping pressure.

12 CONCLUSION

It is clear that the performance of power transformers will determine, to a large extent, the qualityof power supply. It is therefore very important to continuously monitor and assess the conditionof transformers to ensure reliability and availability of power supply. The detection of atransformer winding condition that might lead to a fault and subsequent outage, and being able totake corrective action (e.g. reclamping or insulation repair) prior to failure, can save atransformer rewind, estimated to cost in the region of $1-million for a large power transformer.The failure of a strategic transformer on the power supply network can be even more costly,considering consequential factors such as system destabilization, load shedding, outages andeven complete system shutdown.

Frequency Response Analysis (FRA) is becoming an increasingly popular technique used toexternally monitor and assess the condition and mechanical integrity of transformer windings for short-circuits, open-circuits, deformation, winding insulation breakdown and lose of clamping

pressure. The FRA technique can help maintenance personnel identify suspect transformers,enabling them to take those transformers out of service before failure.

This paper has covered the background to Frequency Response Analysis (FRA) testing, and hasdetailed the requirements and specifications needed to perform an FRA test and generate a

practical FRA guideline. FRA testing can be done using 3 different testing methods, namely;

IFRA Impulse Frequency Response Analysis,SFRA Sweep Frequency Response Analysis,PRBS Puesdo Random Binary Sequence ( still a very new technique )

It is important to note that although these 3 methods generate the same FRA test results, all 3methods use very different measurement techniques to calculate the FRA test result. This meansthat if a general FRA guideline and specification is to be generated, it should be such that it


29/29


October 2003

Page 28 of 24

encompasses all recognized techniques and remains open ended to allow for future FRA researchand development.

IFRA has proven to be an extremely reliable FRA diagnostic tool for detecting transformer winding structure deformation. If a transformer is suspected of internal damage, the IFRA test

results can assist in deciding whether or not to take the transformer out of service for further investigation. A test on a three-phase transformer can be useful in this instance even if no previous tests have been performed on the transformer. Since the winding structure is usuallysymmetrical from one phase to the next, and since it is unlikely that all three phases wereaffected similarly, the FRA test results (fingerprints) for each phase can be compared to oneanother, in order to check for significant differences.

International FRA research has proven that it is possible to detect a variety of different internalconditions. As a preventative maintenance tool, IFRA is the ideal complement to the traditionaltests of ratio, partial discharge and dissolved gas analysis. Combined with on-going internationalresearch into quantifying specific waveform patterns with corresponding winding faults thesystem is proving invaluable in providing base data on currently healthy transformers.

13 REFERENCES

1. Bak-Jensen J. B, Bak-Jensen B. B, Mikkelsen S. D, Detection of faults and ageing phenomena in transformersby transfer functions . IEEE Transactions on Power Delivery, Vol. 10, No. 1, (January 1995).

2. Dr. Karady G, Dr. Reta M and Mr. Amarg F.A, SRP Transformer Fault Detection Project March 19993. Lapworth J.A, McGrail T.J National Grid Company, Transformer winding movement detection by frequency

response analysis (FRA) , Paper No. 8O, presented at the Sixty-Sixth Annual International Conference of DobleClients, Boston, Massachusetts, USA (April 1999)

4. Malewski R, Gockenbach E, Maier R, Fellmann K.H, Claudi A, Five years of monitoring the impulse test of power transformers with digital recorders and the transfer function method September 1992

5. Minhas MSA, Reynders J.P, de Klerk P.J, Failure in power system transformers and appropriate monitoring

techniques October 19996. Noonan T. J - (ESB International), Power Transformer condition assessment and renewal, frequency responseanalysis update , Paper No. 8B, presented at the Sixty-Fourth Annual International Conference of DobleClients, Boston, Massachusetts, USA (April 1997)

7. Tlhatlhetji N.P, Frequency Response Analysis of transformers , Report No. TRR/E/98/EL097 October 19988. Tlhatlhetji N.P, Frequency Response Analysis of transformers , Report No. TRR/E/99/EL102 October 19999. Vandermaar N.P and Wang M Powertech labs inc. Canada, Stefanski P Commonwealth Edison Company

USA, Ward H EPRI USA, Frequency response analysis using the impulse test method as a transformer diagnostic technique , Paper No. 8C, presented at the Sixty-Sixth Annual International Conference of DobleClients, Boston, Massachusetts, USA (April 1999)

Framit3 Ieee Paper

Documents

Transcript of Framit3 Ieee Paper