Вы находитесь на странице: 1из 14

The Transfer Function Method for Detection of Winding Displacements on Power Transformers after Transport, Short Circuit or 30 Years

of Service
K. Feser J. Christian T. Leibfried A. Kachler C. Neumann U. Sundermann M. Loppacher

Reprint

Cigre 2000

E 1-97

HIGH VOLTAGE TEST

THE TRANSFER FUNCTION METHOD FOR DETECTION OF


WINDING DISPLACEMENTS ON POWER TRANSFORMERS AFTER TRANSPORT, SHORT CIRCUIT OR 30 YEARS OF SERVICE

K. FESER*, J. CHRISTIAN UNIVERSITY OF STUTTGART GERMANY T. LEIBFRIED, A. KACHLER SIEMENS AG, NRNBERG GERMANY

C. NEUMANN, U. SUNDERMANN RWE ENERGIE AG, ESSEN GERMANY M. LOPPACHER HAEFELY TEST AG, BASEL SWITZERLAND

Summary: In spite of the excellent service experience with power transformers and their good long term performance diagnostic measurements on power transformers are discussed. The insulation performance is influenced by thermal, dielectric and mechanical aspects. In todays practice the transfer function method is used to detect defects occuring during lightning impulse tests of transformers in the laboratory. Winding faults during the application of the 50 % and 100 % impulse tests can be detected comparing transfer functions. To get reliable results special requirements for performance data of transient recorders, including software and total measuring circuit have to be fulfilled. Results of measurements during testing of power transformers demonstrate the abiltity and sensitivity of the transfer function. The evaluation of the transfer function can, in addition, be used on-site to detect defects, in particular displacements of windings. This kind of defects can occur during transportation of transformers or during a short circuit near the transformer in the power system. Those defects can also occur after a long service life of transformers in power systems, where several short circuits and overvoltages during this time may happen. The excitation for the transfer function in the field could be a high voltage switching event or an applied low voltage impulse. The measurement of the transfer function with low voltage impulses seems to be of special interest, because fingerprints of the transfer function before and after transportation or before and after a short circuit can be taken and compared. The sensitivity of the method applied to on-site diagnoses is discussed. The sensitivity to detect displacements of windings is shown by principal experiments on an old power transformer (200 MVA,
47, D-70569 Stuttgart, Germany

220/110 kV). Practical limits and first results are discussed with the evaluation of transfer functions on a new 125 MVA, 220 kV power transformer before and after transportation as well as before and after a shortcircuit test. A special investigation is presented in the comparison of transfer functions of 24 old power transformers (200 MVA, 220 kV/110 kV) with identical design and rating. Resulting transfer functions of different phases are compared in addition to the comparison of different transformers of the same type. The results demonstrate the sensitivity and applicability of the transfer function as an off-line tool for diagnostic purposes to detect typical defects e.g. winding displacements. Keywords: Power transformer - diagnoses - transfer function frequency response analysis

1 Introduction
Power transformers are the most expensive single elements of HV transmission systems. Therefore, it is an aim of utilities to decrease the transformer life cycle costs and to increase the usable service life. One possibility is to extend the monitoring and diagnoses of power transformers to all possible types of faults. Special monitoring devices for the detection of different types of faults are in use. The ability to withstand short circuits is a basic requirement for a save operation of power transformers. This ability can be affected by thermal and transient mechanical stresses which occur during operation. Especially, power transformers having a winding

* Prof. Dr.-Ing. K. Feser, Institute of Power Transmission and High Voltage Technology, University of Stuttgart, Pfaffenwaldring

displacement caused by a mechanical stress due to a short circuit current should be classified as critical, fig. 1. Generally such pre-damaged power transformers show a serious fault when a next short circuit event occurs. To ensure a sufficient ability to withstand short circuits life management should identify such predamaged power transformers.

the interturn capacitances and inductances. In order to detect pre-damaged power transformers it is necessary to develop new diagnostic methods having a better sensitivity towards changes in the winding geometry. One basic idea is the wide-band examination of the impulse response of a power transformer by frequency response analysis (FRA), because changes in the winding geometry have an effect on the characteristic frequencies. In the following chapters a special FRA method in the time domain is described. Todays experiences with the transfer function method applied in the laboratory or on-site and future applications of the method are discussed.

2 Basic principle of the transfer function (TF) method


2.1 Theoretical background The TF method is based on the two-port network theory. Transformers represent for dielectric and dynamic testing a linear, complex, time-resistant and passive network as long as there are no non-linear MO-arrestors built-in. The theory allows to define one input force and several output gates, fig. 2.
IE

Transformer
UE
1

IA1 IA2 . .

UA1

IAn

UA2

UAn

Fig. 1: Typical winding displacement found in a scrapped 200 MVA generator transformer The requirements for power transformers regarding the ability to withstand short circuits and testing procedures are documented in IEC 60076-5. In the standard the reactance measurement is described as a diagnostic method to demonstrate the integrity of the windings. In accordance with the standard, deviations of the reactance of more than 2% are inadmissible with power transformers having a rated power of 100 MVA and above. Deviations between 1-2% are subject to an agreement between manufacturer and user. Although the detailed specification suggests a sufficient reliability of the test procedure, the reactance measurement is usually not applicable to detect winding displacements at power transformers already in service. The main reason is the accuracy of the reference values. Usually the accuracy of the reference values is in the range of the sensitivity of the reactance measurement. Additional phase selective reference values often do not exist. In this case a comparison with the results of a repetitive measurement is doubtful and a reasonable assessment is not possible. Furthermore, from physical considerations one can derive that displacements as shown in fig. 1 have extremely low effect on the reactance of the winding in concern and also the input capacitance of the winding is normally less affected. Essential measurable deviations in case of displacements have to be found in

complex RLCM-Network

Fig. 2: Transformer considered as linear two-portnetwork Each defined output signal (output voltages UA and currents IA , = 1...n) generates one TF according to: - output voltages: - output currents: UA,(f): IA,(f): UE (f):

TF Au,( f ) = U A,( f ) U E( f ) TF Ai,( f ) = I A,( f ) U E( f )

(1) (2)

FFT of output voltages FFT of output currecnts FFT of input voltage

A TF signifies a complex quotient of the Fourier transformed output and input signals. The sensitivity to defects and changes in transformers' assemblies for each TF is different, [1,2,3,4]. 2.2 Practical measurements on transformers It is possible to determine the TF either in the time or in the frequency domain, [5]. The determination in the frequency domain is performed with a network-analyzer. The frequency of a sine voltage excitation can be controlled over the required bandwidth. The magnitude

and the angle of the complex TF can be evaluated, fig. 3.


5.0 3.0 1.0 -1.0 0 -3.0 -5.0 0.2 0.4 0.6 0.8 1 1.2 1

Transformer
1 1 1

5.0

U0*(1-K(f1))
3.0 1.0 -1.0 0 -3.0 0.2 0.4 0.6 0.8 1/f1 1 1.2

complex RLCM-Network

-5.0

(f1)

sine generator
(variable frequency)

K(f1)
3.0 2.0 1.0

(f1)
transfer function (phase)

2.5 2.0 K(f) 1.5 1.0 0.5 0.0 0.0

transfer function (magnitude)

(f)

0.0 -1.0 -2.0 -3.0

transient voltages the error of TF calculations increases at higher frequencies for the time domain method. Using transient signals means calculating the spectral density at equidistant frequencies. The frequency resolution depends on the record length of the signal. Network-analyzers usually have several operating modes. They use adaptive frequency steps for each frequency range. The quality of spectral resolutions especially for frequencies < 10 kHz is better than those generated by transient signals. Nevertheless, final results of both procedures are quite equal. 2.3 Signal Processing and requirements upon hardand software using the time domain method The signal to noise ratio of a recorded signal represents the main quantity for the accuracy of TF calculations. 2.3.1 Sampling rate and record length The sampling rate describes how often the analogue signal is measured per second. The faster this is done the higher frequencies can be measured by the digital recorder. For a sampling rate fsample the maximum frequency which can be recorded is given by the Nyquist theorem:
f Nyquist = f sample 2

0.2

0.4

0.6

0.8

1.0

1.2 MHz

1.4

0.0

0.2

0.4

0.6

0.8

1.0

MHz 1.4 1.2

Frequenz frequency ff

Frequenz frequency ff

Fig. 3: Measuring the TF in frequency domain In the time domain test objects are excited by low or high impulse voltages. The input and output transients are measured and analyzed. In low voltage measurements the amplitudes are usually 100 V to 2000 V. The shape of the impulse voltage depends on the test device and the test setup. The bandwidth of the exciting signal should be as high as possible. Typical parameters of the impulse shapes are front-times of 200 ns to 1 s and times to half-value of 40 to 200 s, fig. 4. The spectral distribution of the time domain signals are calculated by using a FFT. The quotient of ouput to input signal represents the TF in the frequency domain.
900 800 700 600 500 U (V) 400 300 200 100 0 0.0 20.0 40.0 Zeit (s) 60.0 80.0 100.0

(3)

transformer
1

0.80 0.60 0.40 0.20

I (A) 0.00 -0.20 -0.40 -0.60 -0.80

complex RLCM-Network

0.0

20.0

40.0 Zeit (s)

60.0

80.0

100.0

input signal

response signal

Frequencies larger than the fNyquist are reproduced with strong aliasing components. The resolution f of FFT is determined by the record length N of time domain signal:
f = f sample N

(4)

filtering, sampling, windowing, FFT, quotient

2.5 2.0 1.5 |TF(f)| 1.0 0.5 0.0 0.0

transfer function (magnitude)

3.0 2.0 1.0 0.0 arc(TF(f)) -1.0 -2.0 -3.0

transfer function (phase)

2.3.2 Analogue-to-digital conversion (ADC) - vertical resolution The amplitude resolution a describes the smallest increment a digital recorder can resolve.
a = 2 k +1

0.2

0.4

0.6 0.8 Frequenz frequency f

1.0

1.2 MHz

1.4

0.0

0.2

0.4

0.6

0.8

1.0

MHz 1.4 1.2

Frequenz frequency ff

(5)

Fig. 4: Measuring the TF in time domain In practice measurements in time domain can be performed much faster than in frequency domain. One impulse excitation takes only up to 1 ms. Storage of samples, signal processing and calculation take a few seconds. Depending on the test setups and the number of channels inside the digital measuring system several TF's can be measured at the same time. The duration of measurements in frequency domain usually is about a few minutes. Because most common network-analyzers are constructed in a two channel version only one response signal can be analyzed and only one TF can be calculated for each test in the frequency domain. The amplitudes of harmonic excitations at measurements in the frequency domain are only a few volts. Distortion by ambient noise can decrease the accuracy of the measurement. Because of the limited bandwidth of

k: number of bits of ADC The part of noise caused by ADC consequently is a criteria for the accuracy of calculated spectral density. Fig. 5 illustrates the effect of a 9 bit- and 12 bitquantization to the noise level.
-40 -60 -80 -100 -120 -140 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 frequency [MHz]
19.1 dB -123.7 dB 12-bit digitizing record 9-bit digitizing record

|U(f)| [dB]

-104.6 dB

Fig. 5: Spectral density of low voltage impulses The procedure of oversampling or resolution enhancement relies on averaging multiple measurements of the same quantity. Oversampling can enhance the resolution of signals in time domain at the cost of a reduced bandwidth. 2.3.3 Transformation to frequency domain - transfer functions The comparabilty of TF results is affected by the test setup, the measuring equipment, analogue and digital signal processing and algorithms of calculation. Sometimes, the software-based part of signal processing is disregarded. Before a calculation of the spectral density is performed software-based signal processing is used. For this, several possibilities exist: kind of windowing function (Hamming, Hanning, exponential, ...) parameters of windowing functions zero-padding using FFT or CZT (Chirp-Z-Transformation, [12]) The amplitude spectrum of a standard lightning impulse excitation decreases strongly at higher frequencies. The experiments.
High-voltage
Input Divider Pre Amp. A/D Conversion Memory

frequency range with a signal to noise ratio > 40 dB is only up to 400 kHz. The definition of tolerance bands take under consideration the noise level of the recorded signals, [7]. Two TF's are declared 'equal' if there is an overlapping of both TF tolerance bands in the analyzed frequency range. The Coherence Function is another tool, which provides an indication of the areas where the ingress of noise made the signal processing unreliable. Assuming a linear behaviour of the examined winding and an ideal, noisefree measuring system, the coherence shall be equal to unity over all the analyzed frequency range, [8]. 2.4 Hardware and software solutions Measuring systems require a sufficient bandwidth and vertical resolution. A high performance measuring hardware and up-to-date software technologies are necessary. Nowadays laboratories' test equipment are integrated in communication network systems. Consequently, standard hardware components of personal computers (CPU, mainboards, ...) and standard software applications (Windows, Linux, Office, ... ) have to be integrated. Fig. 6 shows the hardware concept of the digital measuring device, used for the measurements and evaluation of the

Low-voltage

Digitizer

Trigger Logic

Timing Control

2 1

3 4

Chanel Selector

HV in

Digitizer CH 1

HV in

Digitizer CH 2

b u f f e r e d A T B U S

CPU (Pentium)

Monitor

Keyboard

HV in

Digitizer CH 3

COM 2

HV in

Digitizer CH 4

LPT 1

IEEE 488 Zip Drive 100 MByte Floppy Drive


RS 232 C COM 1 (Remote)

Optical (LWL) Converter HOST

IEEE 488 (Remote) SCSI-2 Power Supplies / Line filters MO-Drive CD-Drive etc. = on request = Option

HDD

Fig. 6: Modern high performance digitizing measuring device: HIAS 743 (12 Bit, 120 MHz, 128 kS)

3 Application of the transfer function analysis in dielectric testing (lightning and switching impulse test) and dynamic testing in the laboratory
3.1 General For several years the TF method is routinely applied during the impulse test of transformers in the HV laboratory. The real time records and the frequency domain records of TF's represent a very sensitive method of control for part winding breakdowns in dielectric perspectives or for mechanical defiances such as partwinding deformations, winding shifts, tilting or buckling of windings in dynamic perspectives. Furthermore, the TF method is the first tool which allows to differentiate between test circuit deficites and test object incipient faults. As postulated for a passive network, the TF must stay constant and independent of the applied waveshape. If the TF stays essentially constant at reduced and full test level and one or more other real time records show deviations, there is a clear indication for a test circuit problem, fig. 7. 3.2 Lightning impulse (LI) testing Today, LI tests are routine tests on large power transformers. They may performed as full wave tests 1 x full wave at a reduced test level from 50% to 75% (RFW) 3 x full wave at 100% test level (FW) or as a combined full and chopped wave test 1 x RFW 1 x FW 1 x RCW (chopped wave at reduced test level) 2 x CW (chopped wave at full test level) 2 x FW For transformers with built-in MO varistor protection elements, the test sequence must be adjusted. In this case, only TF's of records obtained at identical test voltage can be compared. Meanwhile the measuring and software tools have been largely improved and some of the earlier problems with the TF method have been solved [3]. Fig. 7a shows clear indications for a failure (decrease of the neutral current) but a fairly constant TF except a pole flattening which is an indication for partial discharges without changing the capacitive coupling of the system. However, a detailled investigation showed a ladder was sparking to a different ground and therefore the neutral current was reduced. Fig. 7b shows a complete match after removal of the ladder of the neutral current and the TF.
b.)

a.)

Fig. 7: TF analysis during the lightning impulse test of a 300 MVA, 230 kV power transformer: a.) with failure indication caused by a flashover in the test circuit b.) without failure indication after correcting the test setup Fig. 8 shows the records of the lightning impulse voltage, the impulse current (iHV) from the neutral terminal to ground, the current from the medium voltage (MV) winding to ground (iMV) and the current from the low voltage (LV) winding to ground (iLV) obtained during the impulse test on the high voltage (HV) phase 1U of a 300 MVA (400 kV/115,5 kV/ 31,5 kV) power transformer. A breakdown occurred after about 45 s. Since the breakdown occurred in the regulating winding, the change of the current from the neutral terminal to ground is much higher than the change of the other currents. The breakdown is clearly visible in the time domain waveforms. Fig. 9 shows the TF's of the above mentioned test calculated as the quotient of the spectral densities of impulse voltage and currents. There is a visible failure indication in all three TF's. The oscillation in the TF is typical in case of an internal breakdown and was observed in several other cases of internal breakdowns.

500 kV 0 -500

-1000 -1500

4 Application of the transfer function for diagnostic purposes to detect winding displacements
4.1 Basic consideration Using the TF for the detection of changes within the transformer is a relative method of diagnoses, fig. 10. Actual results of measurements are usually compared with fingerprints of former times (time-based evaluation). If there is no significant deviation of the compared functions, there should be no change inside the transformer during this time. Unfortunately, fingerprints of old transformers are rarely available. Thus, other informations have to be taken for a comparison. TF's of identical constructed transformers should be comparable (type-based evaluation). If available, it can be taken as reference. If no fingerprint and no identical test object exists, the symmetric properties of the core-and-coil-assembly can be used to compare several measurements (construction-based evaluation). Sometimes, mechanical damages affect only one of the three phases. A separate examination of all phases will detect the differences of one phase towards the others, [10]. The three evaluation procedures can also be taken for fault identification and localisation.
3.5 3.0

a.)

500 A 0

iHV

-500 -1000

b.)
iMV

200 A 0 -200 -400 -600

c.)
iLV

200 A 0 -100 -200

20

40 t

60

80

120

d.)

Fig. 8: Records obtained during the lightning impulse test of a 300 MVA power transformer at 100 % of the test level: a.) lightning impulse voltage b.) neutral current to ground c.) current from the MV system to ground d.) current from the LV system to ground As a result winding faults can be seen in the time domain measurements as well as in the TF. The evaluation with the TF during tests in the laboratory can improve the confidence in the test results and provides additional information in case of an internal breakdown.
12 1 k 8

Database
|TF1(f)|
11 10 12 1 2 3 4 7 6 5 9 8

2.5 2.0 1.5 1.0 0.5 0.0 0.00

time-based (fingerprint) constructionbased (3 phases) type-based

actual Measurement Transfer Function


0.20 0.40 0.60 0.80

Comparison

Result of the Comparison


without fault internal breakdown

I TF1 = HV U

no Changes

significant Deviations

Measuring Accuracy Measuring Error

TF1

4 0 3 1 k 2

a.)
TF2

TF2 =

IMV U

Evaluation of the Measurement

1 0

b.)
1.2 1 k 0.8

Fig. 10: Chronology of on-site diagnostics for power transformers using the TF method
TF3 = ILV U

TF3

0.4 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 MHz 0.8

c.)

Fig. 9: Comparison of TF's from records with and without internal breakdown for the time signals of fig. 8: a.) TF neutral current to ground(HV) b.) TF current MV to ground

After comparing the results and considering the expected accuracy of the measurement, the test engineer or in future a software application has to decide, if the results deviate significantly. If a significant deviation is detected, the measurement has to be evaluated in detail. The results will be compared with a database, which includes an amount of experimental experiences and rules of correlations of TF characteristics and changes in core-and-coil assemblies. The expectations of the user are some conclusions and recommendations. In case of a detected damage inside the test object, other examinations should follow. Measurements of partial discharges, thermal supervisions, gas-in-oil, short-circuit impedances and acoustic analyzes of vibrations can be such further methods, [9].

On-site measurements should be as efficient as possible. Performing on-site measurements represents a time out of service for the transformer and consequently additional costs. Test and measuring equipment to perform TF measurements must be mobile, adaptive to different test objects and easy to handle. Measurements in time domain on one transformer with a mobile equipment, including records at different phases and tap positions, take a few hours, fig. 11. The duration of being out of service is one day.

a.) b.) c.) Fig. 12: Experimental simulation of radial deformations at a tapped winding: a.) hydraulic jacks (300 kN) b.) thrust bolt c.) radial deformation in tapped winding (depth 5 cm) The absolute effect is comparable to the uncertainty of the measurement. The influence on the TF of the transferred voltage is much more obvious, fig. 13b. In a range of 400 kHz up to 800 kHz the differences are very significant. Trends of increasing and decreasing absolute values in this frequency range is detectable. There is also a movement of the resonant frequency near 600 kHz.
20.0 deformation degree 1 (k )
-1

Fig. 11: On-site TF tests at a 200 MVA power transformer using a mobile test equipment Impulse generator (Umax = 4 kV, T1 500 ns, T2 100 s) and transient recorder (HIAS 743) are integrated in a van which should be located as close as possible to the transformer. For connecting test and measuring equipment to the bushings of the transformer coaxial cables and special adapters are used. The experience has shown that arrangements are critical and have to be performed with special care. 4.2 Investigations to the sensitivity of the transfer function to mechanical displacements To get an information about the sensitivity of the TF method a lot of measurements have to be performed. With these measurements deviations of TF's can be correlated to mechanical displacements, [9]. Because damaged windings are rarely available special experiments were performed. Fig. 12 and fig. 13 show one experimental simulation of a radial deformation at a tapped winding of a 200 MVA generator transformer. The experiment simulates damages in windings caused by short-circuits in substations, fig. 1. The upper layer of a tapped winding has been pressed in radial direction using a thrust bolt and hydraulic jacks. The effect of deformation is not very obvious in the TF of the neutral current, fig. 13a.

deformation degree 2 deformation degree 3

12.0

deformation degree 4

|TF1(f)|
8.0

4.0

a. a.)

0.0 0.0 0.9 0.8 0.7 0.6 deformation degree 1 deformation degree 2 deformation degree 3 deformation degree 4 extent of deformation 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 MHz 0.8

|TF3(f)|

b.) b. frequency f Fig. 13: Experimental simulation of radial deformations at a tapped winding: a.) TF: current of the grounded neutral b.) TF: transferred voltage (impulsed block) Short-circuits nearby a transformer may also cause an axial shift of winding cylinders. This effect can only be simulated with a separate block of windings. In an experiment an axial shift of two windings have been investigated by lifting the high voltage cylinder towards the low voltage one with a crane. Fig. 14 shows the determined TF's of the grounded current and the transferred signal of the low voltage winding.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

MHz

0.8

16.0 (k) 14.0 12.0 10.0 |TF1(f)| 8.0 6.0 4.0


-1

0.25
axial shift 0 cm axial shift: 5 cm axial shift: 15 cm axial shift: 25 cm axial shift: 35 cm axial shift: 50 cm

0.20

at manufacturers laboratory at customers substation

0.15 |TF 2 (f)| 0.10

0.05
2.0 0.0 0.0 5.0 axial shift: 0 cm 4.0 axial shift: 5 cm axial shift: 15 cm 3.0 axial shift: 25 cm axial shift: 35 cm axial shift: 50 cm 0.2 0.4 0.6 0.8 1.0 MHz 1.2 1.4

0.00 0.0 0.1 0.2 0.3 0.4 0.5 0.6 frequency f 0.7 0.8 MHz 0.9 1.0

Fig. 15: TF measurements for monitoring a transport of a 125 MVA-power transformer (pulsed winding 1U, transferred voltage 2U) The TF's show significant deviations because of a different setup. At the substation the test and measuring equipment could not be installed in the same way as in the factory. The result of this comparison shows the effect of a non-reproductive test arrangement. It demonstrates the influence of the test setup on the TF and the need of a special care to perform such measurements. Other investigations determined a sufficient reproducibility of currents' TF's up to 400 kHz. The deviations for frequencies more than 400 kHz were caused by different grounding setups, [13]. 4.3.2 Control of short-circuit tests or short circuits in power systems Short circuits near transfomers usually cause currents of high amplitudes. This leads to extreme mechanical stress of core-and-coil assembly, fig. 1. The mechanical forces do not always cause a failure. Sometimes, there are only some pre-damages, which are not recognized and further service is still possible. In order to prove the short-circuit withstand capability, special tests can be performed. Any damages inside the transformer caused by the tests should be indicated by appropriate diagnostics. According to IEC standard a short-circuit test is accepted with a deviation lower than 2 % of the measured short-circuit impedances before and after the tests. However, practical experiments did not show a sufficient sensitivity of this method concerning mechanical changes in core-and-coilassemblies. The TF method seems to represent a method of a better sensitivity. To get some practical experience a short-circuit test of a 125 MVA power transformer was evaluated with TF's before and after the tests. The TF of the neutral current and the transferred voltage of the pulse tested winding block were recorded, fig. 16. There is almost a perfect match of the TF's of the current, fig.16a. On the TF of the transferred voltage is a slight deviation at higher frequencies resulting from the measuring circuit layout and the low signal to noise ratio. This experiment demonstrates the sensitivity of the TF method in order to improve the current uncertainties in assessment of successful dynamic tests, [10]. A visual

|TF2(f)|
2.0 shifting

1.0

0.0 0.0 0.2 0.4 0.6 0.8 1.0 MHz 1.2 1.4

frequency f

Fig. 14: Experimental simulation of axial shifts of two winding cylinders: TF's of the a.) current of the grounded end b.) transferred voltage of the low voltage winding The TF of the transferred voltage signal illustrates the effect of an axial shift. The main resonant frequency decreases, while the damping rate up to 700 kHz increases. A displacement of 5 cm axial height can indicate a significant shift in the TF and represents therefore the sensitivity of the method. The results of the neutral current show no significant changes. 4.3 Examples of diagnoses with the transfer function method on new and old power transformers 4.3.1 Control of transportation After the test of transformers in manufacturers' laboratories there is a partial dismantling for transport (drain off the oil, dismantling of bushings, ...). Dismantling, transport and final installation represents mechanical stress for the core-and-coil assembly. To detect possible mechanical displacements TF's can be measured before and after the transport. To achieve a high comparability of both results, there must be identical conditions for setups. The equipment, the setup of signal wires and the grounding concept should be as similar as possible. Fig. 15 illustrates the results of TF measurements for a 125 MVA power transformer. The TF has been analyzed before dismantling in the factory and after installation in the substation.

inspection of the core-and-coil assembly after the tests has shown no mechanical damages.
12.0 (k)
-1

7.0 (k )
-1

Transformer 1 Transformer 2

5.0

before short-circuit test after short-circuit test

4.0

|TF1(f)|
3.0 2.0

8.0

|TF1(f)|
6.0

1.0

4.0
0.0

2.0

a. a.)
0.25 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHz 1.0 0.20 before short-circuit test after short-circuit test

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

MHz

1.0

frequency f

a. a.)

0.0

Transformer 1 Transformer 2

0.30 0.25 0.20

0.15

|TF2(f)|
0.10

|TF2(f)|
0.15 0.05 0.10 0.00 0.05 0.00 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHz 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHz 1.0

b. b.)
frequency f

frequency f

b. b.)

Fig. 16: Evaluation of a short-circuit test of a 125-MVA transformer using the TF method (performed at KEMA high power laboratory, pulsed block 1V, nominal tap): a.) TF of the current of the grounded neutral b.) TF of the transferred voltage 2V 4.3.3 Inspection of 30 years old transformers The actual condition of a transformer 30 years in service is much more interesting than those of new ones. Unfortunately, fingerprint measurements of TF from former times are rarely available. In substations, transformers were sometimes installed as a set, so that there are two or more identical constructed transformers at one location. If there is no fingerprint of an earlier measurement, results of such 'twins' can be compared. Fig. 17 presents a comparison of two new 125 MVA power transformers in a substation to demonstrate the type-based comparison. The TF characteristics of both transformers are equivalent. The TF's of the transferred voltage do not differ significantly. There are small differences between the results of the neutral current, fig. 17a. This is an effect of a small signal-to-noise ratio. If transformers of the same type are installed at different locations, a typebased comparison is also possible. To check the practical use of the developed test and measuring setup the TF's of several 200 MVA-transformers of one manufacturer are examined.

Fig. 17: TF's of two identical constructed 125 MVAtransformers: pulsed block 1V, nominal tap: a.) TF of the current of the grounded neutral b.) TF of the transferred voltage 2V The service life of the 24 tested transformers are from 24 to 32 years. Fig. 18 presents a comparison of the TF's of two transformers in different substations. Both transformers were constructed in the same year.
14.0 12.0 (k)-1 10.0 8.0 |TF1(f)| 6.0 4.0 2.0 0.0 UA Limburg Trafo 22 UA Brauweiler Trafo 22

a.)

0.0 0.7 0.6 0.5 0.4

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

MHz 0.9

1.0

UA Limburg Trafo 22 UA Brauweiler Trafo 22

|TF2(f)| 0.3 0.2 0.1 0.0 0.0 0.1 0.2 0.3

b.)

0.4 0.5 0.6 frequency f

0.7

0.8

MHz 0.9

1.0

Fig. 18: TF's of two identical constructed 200 MVA power transformers located at different substations (pulse 1V, tap 1): a.) TF of the current of the grounded neutral b.) TF of the transferred voltage

In spite of different locations for the measurement the results of both TF's are identical. This experiment indicates the possibility to achieve comparable TF results at different places if there is an effort to have a reproduceable setup and well developed measuring equipment and it proves the equality of the mechanical construction of windings. To recognize more about the variations of TF results, fig. 19 shows the comparison between 8 transformers of the same design but manufactured at different years.
12.0 10.0 (k)-1 8.0 6.0 |TF1(f)| 4.0 2.0 0.0
Limburg 1 Osterath 2 Utfort Pfalzdorf 2 Opladen 2 Ibbenbren 1 Lstringen 1 Lbbecke

14.0 (k)-1 12.0 10.0 8.0 |TF1(f)| 6.0 4.0 2.0 pulsed phase 1U pused phase 1V pulsed phase 1W

a.)

0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHz 0.9 1.0 frequency f

0.8 0.7

pulsed phase 1U 0.6 pulsed phase 1V 0.5 0.4 0.3 |TF2(f)| 0.2 0.1 pulsed phase 1W

a.)

0.0 0.7 0.6 0.5 0.4

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

MHz 0.9

1.0

0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHz 0.9 1.0

|TF2(f)| 0.3 0.2 0.1 0.0 0.0 0.1 0.2 0.3

Limburg 1 Osterath 2 Utfort Pfalzdorf 2 Opladen 2 Ibbenbren 1 Lstringen 1 Lbbecke

b.)

frequency f

Fig. 20: Using symmetric properties of a 200 MVApower transformer: pulsed blocks 1U, 1V, 1W (UA Brauweiler, tap 10+) a.) TF of the current of the grounded neutral b.) TF of the transferred voltage 2V The results for the 200 MVA transformer show equivalent resonant frequencies of the neutral current. Above 550 kHz the accurracy is decreased due to the noise level. Some deviations in the TF's of the transferred voltage are detectable. Other experiments indicate the dependence of the TF upon 3-phase symmetrie characteristics, vector group and specifications of construction, [10].

b.)

0.4 0.5 0.6 frequency f

0.7

0.8

MHz 0.9

1.0

Fig. 19: TF diagnostics at a set of 200 MVA transformers (impulse 1U, tap 1): a.) TF of the current of the grounded neutral b.) TF of the transferred voltage The resonant frequencies of the neutral current are nearly identical. There is a variation concerning attenuation of poles. The deviations in the range 420 470 kHz is an effect of a low signal-to-noise ratio. The spectral distributions of the transferred voltage are identical. Most of the 24 examined transformers have a similar mechanical state of core-and-coil assembly. Sometimes there is neither a 'twin transformer' (identical construction) nor a fingerprint of an earlier measurement available. The evaluation of the TF method correlate upon the symmetric properties of the phases. Fig. 20 illustrates an example of an inspection of a 200 MVA power transformer using the symmetric properties. For the measurement of an impulse response the transformer is usually exited on one single phase. According to the test setup, several TF's can be calculated. Then the results of separate measurements with different excited phases can be compared for the detection of irregularities.

5 Discussion of the results


The TF is a diagnostic tool to detect winding faults during standard impulse tests in laboratories. There is an evident difference in the characteristics of the TF if the fault is inside the transformer or in the test setup. The TF method is very sensitive to winding faults. The evaluation of TF's is a much more sensitive method than evaluating time domain signals. To get reproduceable TF measurements it is necessary to use standard test setups with qualified measuring equipment. If all requirements are fulfilled TF can be used for the evaluation of transport damages. Utilities are interested to detect mechanical and electrical pre-damages on transformers in service. The effect of radial deformations in transformer windings are visible in the TF's of transferred voltages. Radial deformations are of effect to electric field components, so the influence towards transferred voltages is much bigger than towards neutral currents. Axial shifts of windings effect obvious changes in TF characteristics of

transferred voltages. According to the applied tests the sensitivity of the TF method can recognize a shift of 2 % of the axial height of the winding. The TF method is a relative method. Three types of comparison for measurements can be realized, fig. 10. Time-based comparisons are used to make diagnostics at certain moments. The reason for this can either be special tests or a yearly performed inspection. Using standard equipment and test setups allows to get comparable results, even at different locations, fig. 19. Comparing identical designed transformers represents another method to get reference data. In spite of a certain inaccuracy of TF measurements the main characteristics must be equal (resonant frequencies, damping rate at res. frequency). The results of 24 same-typed transformers illustrate the comparability of such investigations when the same concept of test setup is used. If there is an effort, identical results can be achieved at different locations, fig. 20. The third kind of comparison uses the 3-phase construction of the core-and-coil assembly. The results of the separate tested blocks are compared. There may be differences in the TF caused by the typical features of mechanical constructions or the type of vector group. Consequently this method can be applied to locate and to identify the fault. The evaluation of the TF for large power transformers can be performed in a certain frequency range. The lower frequency is limited by the tail-time of the applied impulse. The higher frequency is limited by the signal to noise ratio. In practice, the used frequency range is between 10 kHz and 1 MHz. Below 10 kHz the FRAmethod with network-analyzers should be applied. But mechanical displacements can be recognized at frequencies of some hundred kHz and therefore the evaluation of the TF with measurements in the time domain is recommended.

breaker a record lenth of at least 200 kSamples (e. g. sampling rate 10 MHz ) is necessary. Because of the statistical distribution of amplitudes, the vertical resolution should be at least 10 bit. Fig. 21 shows one zoomed record of a signal at a 200 MVA 220/110 kV transformer in a power station. The transient is caused by the switching of phase 1W of the breaker poles next to the transformer. The TF's of two records are analyzed in fig. 22. The comparison indicates no significant difference.
200 150 A 100 50
250

kV 200
150

u1U(t)
100

i1N(t)
0 -50 -100

50 0 -50 0.00

a.)
u1W (t)

-150 0.00
50

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90 ms

1.00

b.)
u1V(t)

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

ms 0.90

1.00

Zeit

250

kV 0
-50 -100 -150 -200 -250

kV 200
150 100 50 0

-300 -350 0.00


ms 0.90

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

1.00

-50 0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

ms 0.90

1.00

c.)

Zeit

d.)

Zeit

Fig. 21: Transient signals for TF on-line diagnostics at a 200 MVA transformer in a power station (switching 1W): a.) neutral current (HV) b.) transient voltage 1U c.) transient voltage 1W d.) transient voltage 1V This on-line monitoring system is in continious operation since 1995.
14.0 (k)-1 12.0 10.0 8.0 |TF (f)| 6.0 4.0 2.0 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHz 0.9 1.0
transient record 02.09.96 transient record 27.09.95

6 On-line measurements of transfer functions


TF measurements in time domain can be practiced with high or low voltage impulses. On-site measurements are performed offline. For this usually low voltage excitations are used. Performing TF analysis with high voltage signals is typical for standard impulse tests at manufacturers' laboratories. It can also be performed onsite in a substation. High voltage transients at the bushings of transformers are overvoltages e.g. switching transients originated by a breaker operation, [11]. The quality of a TF calculation depends mainly on the spectral quality of the transient signal. Recording such transients during the regular service represents a possbility to practice on-line-TF-monitoring. For this a high performance digitizing system, a trigger device, high voltage sensors and current sensors have to be installed at the transformer. Because of the required bandwidth sampling rates should not be lower than 10 MHz. To record all single transients caused by a 3-phase

frequency f

Fig. 22: On-line TF diagnostics at a 200 MVA transformer in a power station (TF of the neutral current, excitation at 1W) There are some additional requirements for a comparison of TF's calculated by on-line recorded transients: thermal conditions of the transformer must be similar identical position of tap-changer comparable circuit of the power system next to the transformer sufficient vertical resolution Due to these requirements not all measured transient signals can be used for TF analysis. In addition, economic considerations clearly demonstrate the advantage of the low voltage method for the evaluation of the TF with time domain signals for diagnoses.

7 Conclusions
The TF method seems to represent a quite sensitive method to detect faults in test setups and transformers. In high voltage laboratories this method is able to evaluate impulse tests. Detectable faults usually are dielectric faults. The aim of evaluating TF's during impulse tests in manufacturer's laboratories should be to generate a data base for the identification and localization of faults. Previous investigations indicate a sufficient reproducibility of low voltage on-site measurements even at different locations. This is an essential requirement to do on-site diagnoses for power transformers. The presented experiments to estimate the sensitivity to mechanical defects and pre-damages are first results. To confirm the presented results and trends further experiments and evaluations have to be performed. Only a suitable set of experimental investigations allow to create rules about correlations between TF characteristics and mechanical faults. In addition to these experiments mathematical models must be developed to prove the general validity of the experimental results. The results indicate limits and possibilities in diagnoses of mechanical displacements with the evaluation of TF's (low voltage method). Especially for TF analysis the amplitude resolution of the analogue to digital converter is the most important quantity. An amplitude resolution of 12 real bits was improving the sensitivity as well as the upper frequency limit. The sensitivity of the TF method with time domain measurements of voltages and currents was sufficient in a frequency range of 10 kHz f 1 MHz. Below 10 kHz the tail of the impulse was too short and above 1MHz the signal to noise ratio became too small. Further researches are nescessary for the application of the TF method for diagnoses on transformers.

Transfer Function Method. International Conference on Large High Voltage Electric Systems (CIGRE), 1992, Bd. I [5] Christian, J.; Feser, K.; Sundermann, U.: Beurteilung des Zustandes von Transformatoren mit der bertragungsfunktion. ETG-Fachbericht 79, Internationale ETG-Tage '99, Mnchen, 1999 [6] Claudi, A.; Loppacher, M.: New Methods for Improving the Reliability of Non-Destructive High-Voltage Impulse Testing. 5th International Conference on Transformers, Mumbai, 1998 [7] Leibfried, T.; Feser, K.: A new Method for Evaluating Transfer Functions of Power Transformers. 10th International Symposium on High Voltage Engineering, Vol. 4, Montral, 1997 [8] Malewski, R.; Claudi, A.; Josephy, Ch.; Jud, St.: Checking electromagnetic compatibility of a HV impulse measuring circuit with coherence functions. ERA Technology Conference H.V. Measurements and Calibration, Arnhem, 1995 [9] Christian, J.; Feser, K.; Sundermann, U.: Diagnostics of Power Transformers by using the Transfer Function Method. 11th International Symposium on High Voltage Engineering, London, 1999 [10] Christian, J.; Feser, K.; Leibfried, T.; Jeaggi, F.: Die bertragungsfunktion als Methode zur Beurteilung der Stosskurzschlussprfung und VorOrt-Diagnose. Elektrizittswirtschaft, Jg. 98 (1999), Heft 7, p. 40-44 [11] Feser, K.; Feuchter, B.; Lauersdorf, M.; Leibfried, T.: Gerneral Trends in Condition Monitoring of Electrical Insulation. International Symposium on Electric Power Engineering, Stockholm Power Tech Conference, SPT IS 05-1, Stockholm, 1995 [12] Pfeiffer, W.; Scheuerer, F.: Vergleichende Bewertung der Chirp-Z-Transformation und der Fast-Fourier-Transformation. Elektrie, 10 (1993), p. 370-374 [13] Leibfried, T.; Feser, J.; Miksa, T.: On-lineberwachung eines 350-MVA-Transformators durch die bertragungsfunktion. Elektrizittswirtschaft, Jg. 96 (1997), Heft 10, p. 477-482

8 Literature
[1] IEC 722: Guide to the lightning impulse and switching impulse testing of power transformers and reactors. International Electrotechnical Commission, 1982 [2] Malewski, R.; Poulin, B.: Impulse Testing of Power Transformers using the Transfer Function Method. IEEE Transactions on Power Delivery, Vol. 3, No. 2, 1988 [3] Maier, R.; Kachler, A. : Neues Verfahren zur Beurteilung der Stospannngsprfung von Transformatoren" Elektrizittswirtschaft Jg. 89 (1990), Heft 18, p. 984-989 [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

Haefely Test AG High Voltage Test Division


CH-4028 Basel/Switzerland Phone +41.61.373 41 11 Fax +41.61.373 49 12 www.haefely.com e-mail: sales@haefely.com

HIGH VOLTAGE TEST

Вам также может понравиться