Anda di halaman 1dari 6

1

Experimental Research of Vibration Sweep


Frequency Response Analysis to Detect the
Winding Deformation of Power Transformer
F. H. Wang, J. Xu, Z. J. Jin, and S. S. Gui

timely and correctly to prolong the service life of power


Abstract—In order to further recognize the relationship transformer and ensure the secure and reliable operation of
between the fault of winding deformation of power transformer electric grid.
with the measured tank vibration, the short-circuit test is done in Traditionally, frequency response analysis (FRA) [2],
a 220kV transformer. The vibration sweep frequency response
sweep frequency response analysis (SFRA) [3] and leakage
curves for some pre-setting winding deformation fault are
obtained by virtue of the vibration measurement system. reactance method [4] are widely used to detect the winding
Correlation coefficient is defined and calculated for the whole deformation of power transformer. These techniques are
frequency segment and low frequency segment and high based on the variations of electrical parameters of power
frequency segment separately to evaluate the degree of winding transformer and always used in offline tests, although some
deformation quantitatively. Meanwhile, the amplitude response publications have appeared reporting some online application
curves required by the frequency response analysis are also
of these techniques. Furthermore, they are not sensitive to
plotted and correlation coefficients are calculated. It is showed
that the vibration sweep frequency response analysis is more detect the some incipient failures such as winding loosen,
sensitive to the slightly winding deformation than that of drops or twist of wingding block and slightly deformation of
frequency response analysis. Moreover, the defined correlation the winding etc.
coefficient for a certain frequency segment has a certain relation Due to the fact that the transformer tank vibrations are
to the different fault type of winding deformation. closely related to the variations of internal structure of
transformer. Therefore, methods of monitoring the vibration
Index Terms—Power transformer, winding deformation,
of transformer to detect the transformer winding deformations
vibration sweep frequency response analysis, correlation
coefficient, frequency response analysis. are paid more attention by scholars at home and abroad
recently. Mechefske [5] reports the significant differences
I. INTRODUCTION between the amplitude of harmonic vibrations, the time to
reach the steady-state after a load change and harmonic
I T is known that power transformer is an essential element in
the power system. With the increasingly growth of the
capacity of electric grid and accordingly the augment of short-
content in the high-frequency range in both twin transformers,
one of them with a loose winding clamping. Garcia and
Burgos [6], [7] proposes a model to calculate and estimate the
circuit capacity, the high electro-dynamic forces during short
transformer tank vibration starting from some input variables
circuit is becoming a serious threaten to the mechanical
that can be easily measured on the transformer based on the a
strength and dynamic stability of the transformer winding.
series of experiments to study the factors which influences the
Statistics shows that most of the failures of power transformer
tank vibration. S. C. Xi [8], [9] makes some researches on the
are provoked by the fault of outlet short circuit [1]. The large
core vibration and winding vibration. The relations between
short-circuit currents result in loosen or deformation of the
load current and wingding vibration are studied and wavelet
transformer winding with the consequence of weak
package analysis is introduced to obtain the features of tank
mechanical strength and degraded insulation between turns.
vibration signal of transformer. H. F. Wang [10], [11] presents
Furthermore, a change in the distances among conductors
the dynamic mode of transformer winding. P. A. Xie [12] and
implies a variation in series and shunt capacitances and the
Y. Y. Shao [13] propose the preliminary relations between
voltage distribution in case of lightning or switching over-
winding vibration amplitude, natural frequency and make
voltage being different from the design value to increase the
some theoretical and experimental researches on the
failure risk of transformer. Therefore, it is necessary to detect
transformer winding vibration and clamping pressure based on
and diagnosis the winding deformation of power transformer
the theoretical modeling and experiment test of power
transformer.
F. H. Wang, J. Xu and J. Z. Jin are the Department of Electrical In the paper, the relationship between the winding
Engineering, Shanghai Jiaotong University, Shanghai 200240 China (e-mail: deformation and the vibration sweep frequency response
fhwang7723@sjtu.edu.cn, xujian@sjtu.edu.cn, zjjin@sjtu.edu.cn,).
S. S. Gui is the Shanghai Municipal Electric Power Company, Shanghai
(VSFR) are further investigated by virtue of short-circuit test
200230 China (e-mail: guishunsheng1314@163.com) of transformer through some pre-setting winding deformation

978-1-4244-6547-7/10/$26.00 © 2010 IEEE


2

conditions. Some kinds of winding deformation faults are The excitation signals are inputted to the high voltage side of
preset in the winding of a 220kV power transformer. The three-phase winding of transformer. The signal frequency is
VSFR curve is obtained by virtue of the vibration swept from 115Hz to 310Hz. Eleven accelerometers are
measurement system. In order to evaluate the degree of placed on the surface of the tank with different places and are
winding deformation quantitatively, the correlation coefficient illustrated in Fig.2.
of VSFR curve of different winding deformation is defined
and calculated. The results of frequency response analysis are
also showed in the paper.

II. THEORY OF VIBRATION SWEEP FREQUENCY RESPONSE


METHOD
Vibrations in a transformer are generated by the different
forces appearing in the core and winding during the operation.
Core vibration is caused by magnetostriction and magnetic
forces. Magnetic materials suffer a change in its dimensions Fig. 1. Schematic diagram of vibration signal measurement
of about few parts per million when they are submitted to a
magnetic field. This phenomenon is called magnetostriction.
If neglecting the hysteresis effect and admitting the elongation
proportional to the force, the magnetostriction forces are
propotional to voltage squaredand based on the relationship
between applied voltage and magnetic induction. The
fundamental frequency of magnetostriction forces is 100 Hz
with harmonics being an even multiple of 50 Hz.
Consequently, vibration amplitude and frequency resulted
from core are same to that of magnetostriction. Winding
vibrations are caused by electrodynamic forces resulted from
the interaction of the current in the winding with leakage flux,
which are proportional to the current squared. Because the
current is practically sinusoidal, the main harmonic
component of winding vibration is that of 100 Hz. Therefore,
core vibration can be detected through the no-load test of
transformer and winding vibration can be detected through the
short-circuit test of transformer by virtue of accelerometer Fig.2. Position of eleven accelerometers in the tank
adhered to transformer tank.
If the windings of the power transformer in the short-circuit The experimental steps are as follows.
test are regarded as the vibration source and the whole Step 1. Connected the low-voltage winding in short-circuited
transformer including the core and tank is composed of the conditions through the copper bus-bars;
vibration system, the frequency response curve can be Step 2. Placed the accelerometers in the tank and accessed the
obtained when the sinusoidal excitation with variable excited signal to the three-phase high-voltage winding;
frequency and constant current is input to the vibration Step 3. Select and set the range of sweep frequency in the
system. When the frequency is selected in a certain degree, control signal software in the computer and obtained the
the VSFR curve of transformer winding for some given corresponding VSFR curve;
monitoring points can be plotted. Through the comparison of Step 4. Suspended the core of tested transformer, preset firstly
VSFR curve of different transformer with different degree of the winding fault of transformer manually;
winding deformation, the relationship between winding Step 5. Put the core of tested transformer in the original
deformation and VSFR curve can be obtained. This is the position again and repeated the step 3;
basic principle of VSFR method. Step 6. Suspended the core of tested transformer again and
preset secondly the winding fault of transformer manually;
III. EXPERIMENT DESCRIPTIONS Step 7: Repeated the step 5 and step 6.
According to the previous experimental process, state 1 and
Fig.1 is the schematic diagram of vibration signal state 2 are defined correspondingly to descript the two kinds
measurement. It includes the computer with signal control of simulated winding deformation fault. For the state 1, two
software, signal source, power amplifier, accelerometers and compression nail of A-phase winding are loosened with the
the DH5920 signal acquisition. The test transformer is a distance of two whorls. Two stow-woods on the top of B-
120MVA, 220kV transformer with winding connection types phase winding are moved transversely 5 centimeter or so.
of star grounding. The low voltage side of the test Two stow-woods on the top of C-phase winding are dropped
transformer is in short-circuited conditions in the experiment.
3

completely. State 2 is set based on the state 1 and is described (3) R12- the correlation between the state 1 and state 2.
as follow. Two loosened compression nail of A-phase Meanwhile, the obtained VSFR curve is divided into low
winding are loosened completely. For the B-phase winding, a frequency parts (230Hz-400Hz) and high frequency parts
wood bar is implanted in the stow-wood between the high (400Hz-620Hz) to improve the faith degree of correlation
voltage winding and the low voltage winding to vary the coefficient and to determine the effect of different winding
radial force distribution based on the movement of stow-wood fault to the correlation coefficient of different frequency
of state 1. Two stow-wood on the top of C-phase winding are segment in the VSFR curve.
dropped completely. Moreover, a mallet is used to knock the
three layer coil on the top side of the winding to lead to the
radial deformation of the coil. The deformed arc length is
about 20 centimeter and is fall into 5 millimeter or so.
Consequently, the state 0 refers to the three-phase winding
with no deformation.

IV. RESULTS AND DISCUSSIONS

A. Results of VSFR method (a) Accelerometer-1 (b) Accelerometer-2


Due to the limited length of the paper, the VSFR curve of
the accelerometer 1, 2, 3 and 4 of different phases
corresponding to the three states of winding are plotted in the
Fig. 3, Fig.4 and Fig. 5. In order to display the curve clearly,
vibration amplitudes are plotted in logarithm coordinate. It is
seen that the amplitudes in parts of frequency are slightly
increased in the VSFR curve after the windings are deformed
by man-made. The frequency corresponding to the peak point
is moved distinctly.
(c) Accelerometer-3 (d) Accelerometer-4

Fig. 4. VSFR curve of B-phase winding

(a) Accelerometer-1 (b) Accelerometer-2

(a) Accelerometer-1 (b) Accelerometer-2

(c) Accelerometer-3 (d) Accelerometer-4

Fig. 3. VSFR curve of A-phase winding


(c) Accelerometer-3 (d) Accelerometer-4
Here, correlation coefficient which gives an indication as to
the strength of the linear relationship between two random Fig. 5. VSFR curve of C-phase winding
variables is defined to describe the variations of the VSFR
curve. These indices are Table I shows the correlation coefficient of the three-phase
winding, where L stands for the low frequency segment, H
Cov ( x, y )
Rxy = − lg(1 − ) (1) stands for the high frequency segment and T stands for the
D( x ) D( y ) whole frequency segment. The following conclusions can be
(1) R01- the correlation between the state 0 and state 1. drawn from the table I.
(2) R02- the correlation between the state 0 and state 2.
4

(1) For the correlation coefficient of the A-phase winding in the correlation coefficients in the low frequency part and
the whole frequency segment, R01 is similar to R02 and both of middle frequency part for the state 0, state 1 and state 2 vary
them are less than R12. It means that the winding mechanical slightly and no regular law.
properties vary distinctly when the clamping pressure
decreases compared to the criterion value. However, there is
no apparent influence on the winding property when the
clamping pressure decreases to certain degree.
(2) For the correlation coefficient of the B-phase winding in
the whole frequency segment, R01 is similar to R02 and both of
them are less than R12. That is to say, the effect of the
variation of the radial force of low voltage winding to the
winding property is less than that of the movement of stow-
wood.
(3) The correlation coefficient of the C-phase winding of the
whole frequency segment is less than A-phase and B-phase
distinctly. That is to say, the effect of the drop of the stow-
wood and the deformation in the radial direction to the (a) FRA curve corresponding to the state 0
winding property is larges than that of the loosening of
clamping pressure, the movement of stow-wood and the
variation of radial force.
(4) When the clamping pressure of A-phase winding alters,
the correlation coefficient of low frequency segment is
decreased distinctly. When the stow-wood and radial force of
B-phase winding alters, the correlation coefficient of low
frequency segment is decreased distinctly too. When the
stow-wood of C-phase winding drops and the radial
deformation occurs, the correlation coefficient of low
frequency segment and high frequency segment are all
decreased distinctly.
TABLE I (b) FRA curve corresponding to the state 1
CORRELATION COEFFICIENT OF THREE-PHASE WINDINGS OF POWER
TRANFORMER

Correlation coefficient
Windings of transformer
R01 R02 R12
L 0.409 0.773 0.407
Winding A H 0.658 1.056 0.703
T 0.903 1.246 0.884
L 0.877 1.065 0.874
Winding B H 0.737 0.696 0.613
T 1.009 1.127 0.971
L 0.294 0.201 0.378 (c) FRA curve corresponding to the state 2

Winding C H 0.391 0.362 0.497 Fig. 6. VSFR curve of three-phase winding in the high-voltage side
T 0.657 0.632 0.836
TABLE II
CORRELATION COEFFICIENT OF THREE-PHASE WINDING IN THE HIGH-
VOLTAGE SIDE
B. Results of FRA method
Fig.6 is the FRA curve of three-phase winding of the test State 0 State 1 State 2
transformer in the high-voltage side corresponding to the state
R1-2 R2-3 R1-3 R1-2 R2-3 R1-3 R1-2 R2-3 R1-3
0, state 1 and state 2 separately. Here, the current is input
from the grounding point. The correlation coefficient of the L 2.94 3.14 2.30 2.68 1.91 2.20 1.97 1.82 2.29
curve 1, curve 2 and curve 3 denoted by R1-2, R2-3 and R1-3
H 3.13 2.55 2.24 2.40 2.82 3.00 3.20 2.49 2.24
separately in the low frequency (L) part and middle frequency
(H) part are calculated and showed in Table II. It is seen that
5

Fig.7 is the FRA curve of three-phase winding of the test V. CONCLUSIONS


transformer in the low-voltage side corresponding to the state Through the short-circuit test in a 220kV transformer, the
0, state 1 and state 2 separately. The curve 1 is the B-phase VSFR curves are obtained by virtue of the vibration
response when the current is input from the C-phase. The measurement system by setting some kinds of winding
curve 2 is the A-phase response when the current is input from deformation in advance. The calculated results of correlation
the B-phase. The curve 3 is the C-phase response when the coefficient for the whole frequency segment and the low
current is input from the A-phase. Similarly, the correlation frequency segment and the high frequency segment show that
coefficient of the curve 1, curve 2 and curve 3 in the low the method of vibration frequency analysis is more sensitive
frequency (L) part and middle frequency (H) part are than that of FRA method to the slightly winding deformation
calculated and showed in Table III. It is seen that the of power transformer which is important to prevent of the
correlation coefficients in the low frequency part varies incipient fault from the power transformer. It is shown that
distinctly, which corresponded to the variation of the radial the defined correlation coefficient for a certain frequency
force in the B-phase. Other kinds of winding deformation can segment has a certain relation to the different fault type of
not be found from the variation of correlation coefficients. winding deformation which can be applied to distinguish the
degree of winding deformation based on the VSFR curve of
transformer with no winding deformation. This is the next
work for us.
TABLE III
CORRELATION COEFFICIENT OF THREE-PHASE WINDING IN THE LOW-
VOLTAGE SIDE

State 0 State 1 State 2

R1-2 R2-3 R1-3 R1-2 R2-3 R1-3 R1-2 R2-3 R1-3

L 2.94 3.14 2.30 2.68 1.91 2.20 1.97 1.82 2.29

H 3.13 2.55 2.24 2.40 2.82 3.00 3.20 2.49 2.24


(a) FRA curve corresponding to the state 0

VI. REFERENCES
[1] Y. M. Jiang, "Common position and reason analysis of operating
transformer damage by short circuit," Transformer, vol. 42, pp. 34-38,
Apr. 2005.
[2] J. L. Wu, "Study on frequency response analysis method in diagnosing
transformer winding deformation," M. Eng. D. dissertation, Dept. E. E.,
Wuhan University, 2004
[3] G. M. Kennedy, A. J. McGrail, and J. A. Lapworth, "Using cross-
correlation coefficients to analysis transformer sweep frequency
response analysis (SFRA) traces," in Proc. 2007 IEEE PES Power
Africa conference and Exposition, pp. 16-20
[4] D. K. Xu, "Technology of short-circuit on-line measurement to monitor
the winding deformation of power transformer," Ph. D. dissertation,
(b) FRA curve corresponding to the state 1 Dept. E. E., Xi’an Jiaotong University, 2001
[5] C. K. Mechefske, "Correlating power transformer tank vibration
characteristics to winding looseness," Non Destruct. Test. Cond. Monitor,
Vol.37, pp. 599-604, Aug.1995
[6] B. Garcia, J.C.Burgos, and A.M.Alonso, "Transformers tank vibration
modeling as a method of detecting winding deformations, Part I:
Theoretical foundation," IEEE Trans. Power Delivery, vol. 21, pp. 157-
163, 2006
[7] B. Garcia, J.C.Burgos, and A.M.Alonso, "Transformers tank vibration
modelling as a method of detecting winding deformations, Part II:
Experimental verification," IEEE Trans. Power Delivery, vol. 21, pp.
164-169, 2006
[8] S. C. Xi, "The study of winding and core vibration characteristics and
application in fault monitoring of transformer," Ph. D. dissertation, Dept.
E. E., Xi’an Jiaotong University, 2003
[9] S. C. Xi, S. S. Wang, and Q. Q. Li, "The application of vibration method
(c) FRA curve corresponding to the state 2 in monitoring the condition of transformer winding," High Voltage
Engineering, Vol. 28, pp. 12-14, Apr. 2002
Fig. 7. VSFR curve of three-phase winding in the low-voltage side [10] H. F. Wang, N. Q. Wang, and T. H. Li, "Influence of axial
precompression level on axial vibration in transformer windings," Power
Technology, Vol. 9, pp. 8-11, Sep. 1999
6

[11] H. F. Wang, N. Q. Wang, and T. H. Li, "Axial nonlinear vibration of


large power transformer windings," Power Technology, Vol. 24, pp. 42-
45, Mar. 1999
[12] P. A. Xie, "Study on application of vibration analysis to the condition
monitoring of the winding of power transformer," Ph. D. dissertation,
Dept. E. E., Shanghai Jiaotong University, 2008
[13] Y. Y. Shao, "Theoretical and experimental research on the vibration
characteristics of the large-scale electric transformer winding," M. Eng.
D. dissertation, Dept. Mech. Eng., Wuhan University, 2004

F. H. Wang was born in Henan, China, in 1973.


She received the Ph. D. degree in electrical
engineering from Shanghai Jiaotong University,
Shanghai, China, in 2007.
Currently, she is a lecturer in the Department
of Electrical Engineering, Shanghai Jiaotong
University. Her research interests include
grounding technology, high-voltage technology
and fault diagnosis of power transformer.

Z. J. Jin was born in Shanghai, China, in 1965. He


received the Ph. D. Degree in electrical engineering
from Shanghai Jiaotong University, Shanghai,
China, in 2001.
Currently, he is the Professor and Deputy
Director in the Department of Electrical
Engineering, Shanghai Jiaotong University. His
research interests include high-voltage technology,
grounding technology, fault diagnosis of power
transformer and smart gird.

J. Xu was born in Jiangsu, China, in 1984.


Currently, he is the postgraduate student in
electrical engineering from Shanghai Jiaotong
University, Shanghai, China. His research interest
is fault diagnosis of power transformer..

S. S. Gui was born in Anhui, China, in 1984. He


received the M. Eng. Degree in in electrical
engineering from Shanghai Jiaotong University,
Shanghai, China, in 2007.
Currently, he is the employee in the Shanghai
Municipal Electric Power Company.

Anda mungkin juga menyukai