Escolar Documentos
Profissional Documentos
Cultura Documentos
A Thesis
to be submitted by
S. Gopalakrishna
Doctor of Philosophy
system. Ratings as high as 500 MVA, 765 kV exist in operation today. The reliability of such
equipment can only be ensured by extensive tests in the factory during manufacturing. Some
tests such as power frequency tests are routine tests. The short circuit (S.C) test is an example
of a type test which is done for a particular design. Here, the mechanical integrity of the
windings is tested by a short circuit conducted as specified in standards [1]. The failure
during test is detected by a change in reactance before and after the test. A newer optional
method is the use of frequency response analysis before and after the S.C. test [2], [3]. In this
work alternative methods of testing and analysis are proposed which can significantly
2. Motivation
test is by impedance comparison before and after the test. This is an offline method. Further,
untanking of the winding is done in order to confirm changes visually. Thus there is need to
develop more sensitive methods of detection which reduce the dependence on visual
inspection and also provides some analysis tools that help analyse failure modes.
The short circuit tests are done only in specialized short circuit testing laboratories such as
KEMA, CPRI, etc. These tests are expensive, involve considerable logistics and are time
consuming. The trend towards larger rating of transformers implies that facilities to test
might not exist anywhere at all. For example the largest rating of transformer that can be
tested at KEMA is 8.4 MVA, 245 kV for three phase and 5.0 MVA, 450 kV for single phase.
There is hence an increasing use of design calculations in lieu of actual tests. The method is
detailed in a 2006 amendment to IEC 60076 [1]. This might not be sufficient. In this context,
model based testing can serve as an alternative to the actual test. A scaled model of the actual
1
3. Objective and Scope
The overall objectives and scope of the work can be summarized as follows
excitation.
transformer.
The first improvement suggested over the conventional method is an analysis of impedance
changes on a cycle to cycle basis during the S.C. test. As the test is done at 50 Hz, several
relevant variables can be measured in conjunction with impedance changes. Among them are
the changes in real power and reactive power drawn by the test object. Equivalently, if a
comparison method with an identical test object is possible, ‘errors’ with and without
deformation are measurable. Here, it is ensured that the test object to which the device under
test (DUT 1) is being compared does not undergo deformation. Consider as an example
a 11 kV/ 3 110 kV/ 3 , 100 VA, class 1.0 voltage transformer (VT) shown in Fig. 1(a)
whose cross sectional view is shown in Fig. 1(b). The secondary is fixed rigidly over the core
and primary is free to move over the secondary from position ‘A’ to position ‘B’. A solenoid
is used to move the primary from position ‘A’ to position ‘B’. A standard VT of same voltage
and current rating is used as the reference. A cycle to cycle deviation between the secondary
currents from the DUT 1 and the reference is expressed as ratio error (RE) and phase error
(PE). The plot of the RE and PE as the solenoid moves primary from position ‘A’ to position
‘B’ is shown in Fig. 2. Cycles 4 to 8 shows the period of movement (approximately 0.1s).
2
VT
a N Primary
Core
Core Secondary
n
A Solenoid
actuator
Solenoid actuator
Reference position, A
Position of primary
winding at maximum travel, B
Fig. 1(a) A 11 kV VT with solenoid arrangement (DUT 1) (b) Cross sectional view of DUT-1
0.6 15
Normalized PE (min)
% Normalized RE
0.4 10
0.2 5
0 0
0 5 10 0 5 10
Cycle Cycle
Fig. 2 Cycle to cycle measurement of ratio error (RE) and phase error (PE) in the VT of Fig.1 when
compared with a standard VT
Similarly the method is also extended to cater to deformation in windings with electrical
excitation.
4.2 Improved Methods for Short Circuit Testing of Power Transformer with High
Frequency Excitation
It has been known that frequency response analysis before and after the test is an alternative
method for assessing deformation [1]. We propose a method that uses concurrent high
deformation detection. The high frequency signal is an optimized signal arrived after
3
Simulation study: In order to study the sensitivity of methods to identify deformation it
would be appropriate to consider a benchmark winding which has been validated for other
studies. We choose the coil proposed in [4] as it has a uniform layer winding, its resonant
frequencies are known and several practical measurements have been performed on it. The
winding was modeled by lumped parameters using 10 section model. It is proposed to use
calculated and compared with impedance changes at 50 Hz. Different cases of the
assumed to occur in one section, without any mechanical effects on the other sections. Fig.
3(a) shows the schematic of the coil without deformation. Fig. 4(a) shows impedance (Z)
changes at 50 Hz due to 4% (of winding length) axial displacement in 1st section as shown in
Fig 3(b). Similarly Fig. 4(b) shows impedance changes at different resonant frequencies,
where nth resonant frequency is denoted as RF-n. A comparison shows that the sensitivity of
high frequency is higher than that of the 50 Hz frequency. Similar results are obtained for
radial inward and radial outward displacements of 5th section as shown in Fig. 3(c) and 3(d)
Experimental demonstration: A power transformer model (DUT 2) with four disc windings
and air core is chosen to demonstrate the method of concurrent excitation. The top two discs
of the model are connected in series and shorted. The bottom two are in series and excited
with high frequency and 50 Hz frequency signals. The circuit shown in Fig. 5 shows the
introduced in parallel to the winding under test such that the circuit resonates at a frequency
fh. The frequency fh is chosen based on maximum sampling rate and data storage capacity of
the measuring equipment. Cs is used to decouple the 50 Hz source from the high frequency
source fh. A continual evaluation of deformation during the test is possible as the 50 Hz
current is measured across Rl and high frequency current is measured across Rh. The shorted
4
coil is moved by mechanical means and current waveforms are recorded using digital
oscilloscope during coil displacement. For the same deformation, the current changes at 50
Hz are only 0.58 %, while that of high frequency is around 4.4 %. Fig. 6 shows cycle to cycle
basis percentage changes in the rms currents for both 50 Hz and high frequency signal. The
sensitivity of deformation detection at high frequency excitation is about 8 times and the
Various high frequency signals such as chirp, impulse, oscillating impulse, etc. are used for
simultaneous excitation along with power frequency for detecting winding deformation [5],
[6]. Here we show that optimization of high frequency signal improves the sensitivity in
obtained in a fixed measurement time for a specified maximum peak value of excitation [7].
Core
First section
Fifth
Winding section
80
0 RF-1
70
RF-2
60 RF-3
-0.2 RF-4
50 RF-5
Z-Changes (%)
RF-6
Z-Changes (%)
40
-0.4 RF-7
30 RF-8
-0.6 20
10
-0.8 0
-10
-1 -20
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Winding displacement(%) Winding displacement(%)
Fig. 4 Variation in impedance with axial displacement expresses as % of winding length of the single layer
winding shown in Fig. 3 (a) 50 Hz frequency (b) nth resonant frequency (RF-n)
5
7
50 Hz
6 High frequency
currents
5 ih
il Rl ih Rh
Irms
4
rms
winding in
in Change
3
ih+il
2
fh
% Change%
fl Cp il
1
Winding
0
-1
0 0.1 0.2 0.3 0.4 0.5
CS Time (s)
Fig. 5 Circuit for S.C. test on DUT 2 at power Fig. 6 Comparison of cycle to cycle variations of rms
frequency excitation (fl) with concurrent excitation (fh) currents in DUT 2 with and without deformation at fl
and fh
The device under test is the voltage transformer (VT) shown in Fig. 1 (DUT 1), whose
showed that 64 kHz is more sensitive and hence all further excitations are done centered at
this frequency. Three types of signals namely chirp, multisine and optimized multisine are
used. The multisine is generated by the sum of 11 instantaneous voltages of equal amplitude
in the range of 60 kHz to 70 kHz at an interval of 1 kHz. Chirp signal frequency varies from
60 kHz to 70 kHz with duration of 1 ms. All the three signals have 20 V peak to peak
voltage. These signals are superimposed on 50 Hz sine wave of amplitude 2 V and applied to
the VT. Currents IA and IB are measured at the two different positions of the primary,
position-A and position-B respectively with each of the excitation. The sensitivity of current
(I B − I A )
Sensitivity (%) = × 100 (1)
IA
The comparison of the sensitivity from equation (1) in frequency domain indicates that the
6
4.4 Model Based Testing for Establishing Design Margin of Power Transformer
The improved methods of detection can also serve to analyse tests on model structures. We
propose model based testing, which involves developing a scaled model of the power
transformer to be tested. The physical relation between the model and complete transformer
winding can be established using finite element based field computations. The model is tested
in two ways. The first one involves electrical testing. The second one is testing with
hydraulic actuators which can simulate the forces under short circuit conditions.
We consider a 5 MVA, 22 kV/11 kV three phase transformer proposed in [8] for study. The
axial forces are calculated considering that the height of HV winding is 1% smaller than LV
winding. The total axial force on HV winding of the transformer is simulated in a FEM based
software. For the sake of simplicity, the windings are assumed as single conductors, as shown
in Fig 7(a). A zoomed view of high voltage (HV) and low voltage (LV) windings is shown in
Fig. 7(b). The axial force developed in a winding is the summation of forces at all
FAxial = ∫ fV dV (2)
V
Fig. 8(a) shows the variation of fV along the length of the HV winding, at the instant of
peak short circuit current. fV is almost constant along the width of the winding as winding
performed and fV is integrated over the 2D winding surface, which gives force per unit
length f L . Total force is obtained by multiplying f L with mean circumference of the winding.
7
The short circuit current in HV winding is 1303 A. The top 91.5% of the winding experiences
a downward axial force (57700 N/m) and the bottom 8.5% experiences upward force (9778
N/m). The total axial force is in downward direction (47922 N/m). A Model is developed
with winding length 10% of the actual transformer, without altering the width of the core and
windings. A current of 1550 A causes a total axial force of 48260 N/m in the downward
direction in the model. The distribution of axial forces in the Model is shown in Fig. 8(b).
The model developed above produce similar ampere turn distribution and axial force as of the
actual transformer. As the windings are of different radii, the developed force has two
components, one in axial and other in radial direction as shown in Fig. 9 (a). As we are
interested in the axial force, the radial component has no useful purpose. We propose a model
as shown in Fig. 9(b) where all the force is only in axial direction, thus improving the force
A hypothetical 11 kV/ 415 V, 63 kVA transformer is considered to evaluate the axial forces
using FEM. The short circuit current is calculated to be 270 A. The net force developed in the
end sections is 632 N. A model with 5% of transformer length is simulated using FEM and
axial force developed in the model is found to same as that of actual transformer at a voltage
excitation of 11.4 V.
In order to demonstrate the principle, a 5% length model is constructed. The winding is made
of aluminum and has two sections, with two discs per section and 10 turns per disc. An
arrangement for excitation was made similar to the requirements for short circuit test of a
transformer. A 400 V, 20 kVA variac was connected to a 400 V/10 V, 40 kVA transformer
8
Core
LV
HV
(a) (b)
Fig. 7 (a) Winding arrangement in 5 MVA transformer developed in FEM software (b)
Zoomed portion of the HV and LV windings
LV HV
50817 N/m
57700 N/m
91.5 %
82.5 %
Length
0.09h
9778 N/m
2557 N/m
8.5 %
17.5 %
Fig. 8 Changes in the axial force in HV (outer) winding as a function of length at peak value of short-
circuited current in (a) simulated Transformer (b) simulated model
Faxial Faxial
LV
Fradial
LV
CORE CORE
HV
HV
Fig. 9 (a) Model with two component forces (b) Improved model
9
Table 1. Comparison of transformer and the model (FEM simulation)
Parameter Transformer Model
Winding Height 1420 mm 142 mm
No. of turns 804 80
SC Current 1303 A 1550 A
Total Axial Force 47922 N/m 48260 N/m
which could source up to 8000 A with a short across its terminals. A timer was used to
restrict the time of energisation to 0.5 s. At full excitation, the current through the model is
778 A. This causes a net axial force that needs to be restrained by the end plate. It was easy to
demonstrate that the upper clamp plate could move if not adequately restrained.
It is known that FEM can be used to obtain exact force profile as a function of time during an
S.C. test. With recent advancements in the field of hydraulic actuators with a good frequency
response, an identical force can be applied to a winding with an actuator rather than through
means be considered such that the force profile obtained from the mechanical device matches
Fig. 10 Arrangement for proving end thus the design margin of end plate is established.
clamping structure with a hydraulic actuator
10
4.5 Force Calculations for Other Components with Short Time High Currents
The currents in the secondary of a power transformer are limited by the short circuit
impedance Thus there is an upper limit to the ratio of the short circuit current to the rated
current which depends on the transformer itself and not on system parameters. On the other
hand for a device such as a current transformer the primary short circuit current depends on
network parameters and can be sometimes 100-500 times the normal full load current. The
forces can hence be very high. The model based test method is extended to current
5. Conclusions
Improved methods for assessment of winding deformation at power frequency during short
circuit test in power transformers are proposed. Simulation and experimental results shows
that the high frequency excitations are more sensitive than power frequency for detection of
axial as well as radial winding displacements. A method is proposed for concurrent high
frequency excitation during short circuit test. A method using scaled model is proposed to
establish design for the power transformer during short circuit. The designed scaled model is
tested by either ways (1) using electrical excitation (2) using hydraulic actuators. Model
6. References
[1] IEC 60076-5: 2006, Power transformers- Ability to withstand short circuit.
[2] E. Rahimpour, J. Christian, K Feser and H. Mohseni, “Transfer function method to diagnose axial
displacement and radial deformation of transformer winding”, IEEE Trans. on Power Delivery, vol. 18,
No.2, pp 493-505, April 2003.
[3] S. A. Ryder, “Diagnosing transformer faults using frequency response analysis”, IEEE Electrical Insulation
Magazine, vol.19, no.2, pp.16-22, March/April 2003.
[4] R. C. Degeneff, “A general method for determining resonance in transformer winding”, IEEE Trans. on
Power Apparatus and Systems, vol. pas-96, no. 2 pp. 423-430, March/April 1977.
[5] S. Santhi, S. Jayalalitha, V. Jayashankar, V. Jagadeesh Kumar, “Detection of winding deformations during
short time current tests”, IMTC-2005, Canada, May 2005.
[6] S. Santhi and V. Jayalalitha, V. Jayashankar, “Continual assessment of winding deformation during a short
circuit test”, IEEJ Trans. Power and Energy, Vol. 126, No.7 pp. 712-713. June 2006.
11
[7] Rik Pintelon and Johan Schoukens, “System Identification”, IEEE Press, pp. 115-138, 2001.
[8] M. Waters, “The short-circuit strength of power transformers”, MacDonald, London, 1966.
7. List of Publications
Journal
Conference
1. Introduction
1.1 Short circuit test in power transformers
1.2 Motivations
1.3 Objectives
1.4 Organization of the Thesis
12
2.3 Deformation due to electrical excitation
2.4 Deformation detection from real & reactive powers
2.5 Discussion
2.5 Conclusions
6. Conclusions
6.1 Summary
6.2 Scope for Future Work
13