456

Model of Distribution Transformers Winding for Lightning

Credson de Salles and Manuel Lus Barreira Martinez

AbstractThis paper presents a mathematical model for distribution transformers winding to simulate its lightning impulse behavior. The aim is to investigate the real source of some signals present in the current wave of lightning impulse test. These signals are generally attached to the testing system and not to the testing transformer. This kind of interpretation can consider approved a failed transformer. The procedure takes into account the comparison between lightning impulse tests on actual transformers and their simulation results. The main contribution of this paper regarding previous publications [1]-[2] is concerning the current wave of chopped impulses. The idea is to use the chopped waves to confirm the existence of small failures on tested transformers that present some of noises in the current wave of full impulses. Index Terms Circuit Modeling, Testing, Transformers, Transient Analysis

transformer. The transformer is modeled by its surge impedance (ZSTR) and its self-capacitance (SC). These parameters respond, respectively, for the inductive and the capacitive component of the current waves recorded during the transformer impulse tests. The results from the simulations of the circuit in Fig. 1 are applied to determine ZSTR of the actual testing transformer. These simulations consider the shunt value, impulse generator parameters and several attributed values of ZSTR that represent different transformers. Also, these simulations consider a surge propagation velocity (v) of 160 m/s [3] and a voltage of 1000 V.
F Rd Zsrt IMAX
Shunt

CS

Re

Cb

I. INTRODUCTION HE experience with Brazilian manufacturers shows that the analysis of the shapes of transformer impulse tests must consider the performance of measurement equipments. Some of them are not able to show specific failures as for instance small turn-to-turn short-circuits. In other cases, concerning equipments able to show these failures, the usual scaling adopted during testing do not permit a clear visualization of the failure. In some cases, small noise signals are considered as a problem attached to the measurement system due to induced pulses, and not to the testing apparatus failure. The idea of this paper is to show, aided by ATP simulations, that some measurements procedures permit that small failures do not be noticed. To avoid it the current signal of the chopped impulses must be considered on the analysis adding extra evidences that the high frequency oscillations on the current traces of full waves are real failure evidences.

Fig. 1. Circuit used to determine the Winding Inductance

The output data from the simulations of the circuit of Fig. 1 are the winding length, and the maximum inductive current obtained for the time of voltage wave zero cross. This permits to construct current amplitude surfaces matching the variation of the maximum current, voltage zero-crossing time with the winding surge impedance, and coil length as shown by Fig. 2.

II. SIMULATION OF TESTING CIRCUIT PARAMETERS The testing circuit to simulations takes into account the parameters of the impulse generator, and of the testing
C. de Salles and M. L. B. Martinez are with the Federal University of Itajub, Electrical System and Energy Institute, High Voltage Laboratory Po Box 50 Itajub, Minas Gerasis, Brazil 37500-050 (phone: +55 35 3622 3546; fax: +55 35 3622 5168; e-mail: credson@ lat-efei.org.br). Fig. 2. Maximum Current according to coil length, voltage zero-crossing time and transformer surge impedance

Likewise the inductance determination the results of the simulations of the circuit in Fig. 3 are used to determine the winding self-capacitance (SC) of the transformer.

456
F Rd SC

2 As long as the voltage impulse test is carried out in the HV bushing, only all high voltage capacitances to the ground are taken into account.
H
CHL

IP
Shunt

CS

Re

Cb

Fig. 3. Circuit used to determine the self-capacitance

CH

CL

The simulations take into account several values of SC and the output data are the current associated to the charging selfcapacitances phenomena, in other words, the peak current (IP). This procedure permits, for each connection of the impulse generator, to set plots matching the IP to the SC as shown by Fig. 4.
1400 1300 1200 1100 1000 900

Source

measurement

shielding

Result

H L L

mass mass L

L H mass

CH CL CHL

Fig. 6. Transformer capacitance measurement model

IP [mA]

800 700 600 500 400 300 200 100 0 0 100 200 300 400 500 600 700 800 900 1000

B. Winding Self-capacitance (SC) and Inductance (LHF) These parameters are obtained from the application of two reduced wave in the transformer. The first is with a volt-scale able to show the peak of the capacitive current as shown in Fig. 7. The second uses a time scale able to show the time when the voltage cross the zero value as shown in Fig. 8. This voltage zero crossing time (Tds) shows the maximum current value attached to the traveling waves phenomena [3]. Fig. 7 and Fig. 8 are related to a 45 kVA 125 kV BIL threephase transformer. The current wave in Fig. 7 presents three failure signals.
SC [pF]

Fig. 4. Peak-Current versus Winding Self-Capacitance

In Fig. 1 and Fig. 2, CS is the impulse capacitor, F is the switching gap, Cb is the load capacitor, Re is the discharge resistor and Rd is the damping resistor.

III. MODEL PARAMETERS The approach adopted in this paper considers leakage capacitances, the winding self-capacitance and impulse inductance, connected in a suitable way to represent the transformer during impulse tests as shown by Fig. 5, for a winding divided in 10 parts.
1 SCx10 2 SCx10 10 SCx10

Fig.7. Peak acquiring of the capacitive component of current wave

LHF/10

LHF/10 (CH+CHL) 10 (CH+CHL) 10

LHF/10 (CH+CHL) 10

shunt

(CH+CHL) 10

(CH+CHL) 10

Fig. 5. Transformer winding model used in this work

A. Capacitance Measurement The leakage capacitances can be obtained according to the traditional Schering Bridge approach as shown by Fig. 6 for a three-phase transformer. For a single-phase transformer just the capacitance measurement CHL is carried out to evaluate.

Fig.8. Zero cross time acquiring for pointing the maximum amplitude of inductive component of current wave

The channel 2 in Fig. 7 shows the IP voltage input through a 10 shunt. In other words, 52.8 A under 43.6 kV. This value referred to 1000 V of Fig. 3 simulation corresponds to is 600 mA that, in Fig. 4 means a SC of 280 pF.

456 Fig. 8 shows a Tds of 160 s and a maximum current of the inductive component of 35 mA referred to 1000 V. These data in the Fig. 2 results in a ZSTR of 40 x 103 . Therefore, with the values of ZSTR, CH, CHL, power frequency capacitances and Eq. 1 it is obtained the winding high frequency inductance LHF as Eq.1.

LHF = Z STR (C H + C HL )
2

(1)

Extra Eq.2 and Eq.3 can be also applied to verify these parameters. This also suggests the development of a new set of characteristics able to take into account variations of the propagation velocity.

Fig. 10. Applied Voltage Impulse in an approved transformer

1 SC = vZ str

(2) (3)

To verify the model accuracy Fig. 11 shows the testing and the simulated current attachment for the transformer of Fig. 10. These oscillograms show the total current to ground [6].
10 0 -10

LHF =

Z str v

Actual trace

C. Impulse generator The impulse generator used to test the transformers is a low energy one. So, according to the characteristics of the testing transformer it is necessary to change its capacitors and resistors arrangement to get the standard wave shape of 1,2/50 s. Fig. 2 and Fig. 4 are related to the 3s2p configuration (three stages in series each stage with two parallel capacitors). Also, for scales compatibility between the reduced and full current waves, it can be used shunts of different resistance. D. Model Parameters Composition Applying real parameter values to the model of Fig. 5 it is possible to find results similar to those of Fig. 9, where it is represented a full-wave impulse test. The current wave shows clearly two of its three components: the capacitive and the inductive. The last one, generally, is the most reliable for fault detection [4].
20 *10 3 5

2 shunt current (A)

-20 -30 -40 -50 -60 -70 -20 0 20

Simulated Trace

40

60

80

100

120

140

160

180

Time (s)

Fig. 11. Comparison between Testing and Simulating Results

V. FAILURE SIMULATIONS The ATP failure simulations consider the circuit of Fig. 12 where each short-circuit is modeled by the closing of the switches in each part of the divided winding. Rcc is a resistance value adjusted to limit the amplitude of the high frequency oscillations according to the values observed from actual oscillograms.
RCC 1 RCC 2 RCC 10

-10

-25

SC x10
-40

SC x10 LHF/10 (CH+CHL) 10 (CH+CHL) 10

SCx10 LHF/10 (CH+CHL) 10

-55

LHF/10
0.04 0.08 0.12 0.16 [ms] 0.20

shunt

-70 0.00

(file MD1-SURTO.pl4; x-var t) v: CS c: TNKfactors: 1 1 250 offsets: 0,00E+00 0,00E+00 0,00E+00

(CH+CHL) 10

(CH+CHL) 10

Fig. 9. Composition of winding model simulated in ATP program voltage and current waves

Fig. 12. Winding model to the winding failure simulations

IV. MODEL ACCURACY Fig.10 shows the first full-impulse [5]-[6]. Applied to a 15 kVA, 95 kV BIL, single-phase transformer.

A. Full-wave Failure Simulations Fig. 13 shows the comparison between the actual and a simulated current for the reduced impulse applied to the transformer of Fig. 8.

456

4 The failure signal on the voltage trace was amplified in the current one. It is possible to show with several impulse test results that for an approved transformer there is not distortion at the beginning of the current wave of the chopped impulse according [4]. This simulation, as others, suggests that the use of chopped wave current trace must be considered on evaluation of a transformer. Fig. 16 and Fig. 17 show, respectively, the first and the second chopped-wave [5][7] applied to a 75 kVA, 150 kV BIL three-phase transformer. There is not any failure signal in the voltage wave, and beside the difference in the time to the chopping, there is no distortion in the current wave.

Simulated trace

Actual trace

-20

20

40

60

80

100

120

140

160

180

Time (s)

Fig. 13. Comparison between Testing and Simulation Results

Each simulation considered a development of a short-circuit in 2% of the winding. The actual transformer presents an insulation failure and, for this reason, the oscillogram was acquired with a reduced impulse. The actual trace as in Fig. 8 has three signals of arc extinction associated to three short-circuits at the beginning of the winding. B. Chopped-wave Failure Simulations Fig. 14 and 15 show the comparison between the actual and the simulated chopped impulses of a 10 kVA, 125 BIL singlephase transformer. Respectively, voltage and current traces.
Actual trace

Fig. 16. First chopped-wave from an approved transformer

Simulated trace

-1

Time [s]

Fig. 17. Second chopped-wave from an approved transformer

Fig. 14. Reduced Chopped Impulse Actual and Simulated traces comparison Voltage Traces

VI. DIGITAL TRANSIENT RECORDERS VERSUS DIGITAL OSCILLOSCOPES Some recording apparatus are not able to acquire clearly the inductive component of current wave from the full-waves due to the overflow caused by the capacitive component. In these cases, the most reliable wave for fault detection according [4] is only available with low resolution. On the other hand, the time scale used by some equipment can hide failure signals. Consequently there are some points to be commented about time and vertical scales of test recording equipments.

Actual trace

Simulated trace

-1

4 5 Time [s]

Fig. 15. Reduced Chopped Impulse Actual and Simulated traces comparison Current Traces

A. Time Scales Fig. 18 and Fig. 19 show the oscillogram of a reduced impulse applied to a 15 kVA-150 kV BIL single phase transformer. The first is recorded by a standard digital

456 oscilloscope with 9 bits 2.5Gs/s per channel, and the second shows its simultaneous recording by a prototype 12 bits 50Ms/s per channel digital recorder of own internal development.

5 The short-circuit signal at the beginning of the winding is showed in the voltage and current waves of Fig. 20. Therefore, it is possible to see that the sensitivity of the digital oscilloscope to the current wave is higher than to the voltage. The most interesting point is that the used sweep time in this case is the same of the chopped waves [5]-[7]. The question is not to discuss the time sweep proposed by the standards, but to show that, in some cases of signals attributed to high frequency oscillations, the actual source signal has internal transformer causes and the time scales do not permit a clear visualization of them.
1 0 -1 -2 mV -3 -4 -5 -6 -1000 0 1000 2000 3000 4000 ns 5000 6000 7000 8000 9000 x 10
4

Impulso Reduzido - Sinal de Tenso

Fig. 18. Reduced impulse in a transformer with turn-to-turn short-circuit standard digital oscilloscope recording
0 -2000
1 0 -1 -2 mV -3 -4 -5 -6 -2 0 2 4 6 8 ns 10 12 14 16 18 x 10
4

Impulso Reduzido - Sinal de Corrente

x 10

Impulso Reduzido - Sinal de Tenso

mV

-4000 -6000 -8000 -10000 -1000

1000

2000

3000

4000 ns

5000

6000

7000

8000

9000

Fig. 21: Reduced full impulse with time scale 1s/Div Fig. 18 simultaneous recording

Impulso Reduzido - Sinal de Corrente 500 0 -500 -1000 mV -1500 -2000 -2500 -3000 -2 0 2 4 6 8 ns 10 12 14 16 18 x 10
4

In some recording equipments, the resolution is lost if the time sweep of, for instance, 180 s is closed to 10 s. For analogue oscilloscope (with tubes), this resource is impossible, but working with an input voltage up to 3 kV, the simultaneous record of external noise is virtually impossible. B. Vertical Scales Even when the transient recorder is able to acquire waves able to shown a failure development and detection, sometimes the vertical scales do not permit a clear visualization as, for instance, when the scale is chosen for acquiring the capacitive component of the full-wave current. Fig. 22 shows a first full wave applied to a 75 kVA 150 kV three-phase with double acquiring of current wave to show the equivalence between the probes, and Fig. 23 a second full wave with double acquiring, but this time with the channel 2 vertical scale adjusted to recording the peak of its capacitive component.

Fig. 19. Simultaneous recording of Fig. 18

In this case, the digital oscilloscope could not register the insulation failure in the voltage wave channel 1 as the prototype digital recorder did. This can be attached to the time scaling. The current wave channel 2 shows some failure signals due to corona or arc extinction from some insulation failure at the beginning of the winding. These signals are not so visible in the prototype digital recorder because the probe attenuation. By other side, closing the time scale from 20 s/Div in Fig. 18 and Fig. 19 to 1 s/Div in Fig. 20 and Fig. 21, respectively, the failure visualization becomes clear.

Fig. 20. Reduced full impulse with time scale 1s/Div by standard digital oscilloscope

Fig. 22: Full impulse with channels 2 and 3 vertical scales to acquire the peak of current wave capacitive component

456 Fig. 23 shows in channel 3 that the failure signals are clearly visible in their vertical scales. At the same time, the channel 2, with its vertical scale five times greater that of Fig. 22 caused the concealing of the failure signals, in favour to the peak of capacitive component acquiring. With this vertical scale, the very small signals that appear in channel 2 can be attributed to high frequency oscillations. VII. COMMENTS

Fig. 23: Full-impulse with vertical scale to acquire the peak of current wave capacitive and inductive components channels 2 and 3 respectively

Fig. 24 shows the simulation of a reduced full-impulse applied to a 75 kVA-95 kV BIL three-phase transformer with a short-circuit at 10% from the winding beginning. The signals in the voltage and current wave around 160 s are related to the arc extinction of the failure that occurred around 3,5 s as shown by Fig. 25 with a closed sweep time.
15.0 *10 3 7.5 0.0 -7.5 -15.0 -22.5 -30.0 -37.5 -45.0 0.00

The model adopted in this paper does not take into account the construction characteristics of the transformers, and at certain point, the number of short-circuited turns. The parameters are simply obtained from the voltage impulse tests and from capacitance measurement. As it can be observed even without considering the construction characteristics the model presented a good attachment with the testing behaviour. The recommendations on testing transformers according [5][7] are clear for analogue oscilloscopes, but there is no mention about possibilities offered by the digital oscilloscopes and recorders. The same can be stated regarding some previous IEC equivalent standards. It was possible to observe added to ATP simulations that some signals attributed to high frequency noise can be actual failures evidence as well as shunt flashover. Fig. 14, Fig. 15, Fig. 16 and Fig. 17 show the possibility of considering the current of chopped waves in a transformer evaluation. The Brazilian Standards are not so clear about this point. But there is a mention in [8] about the use of the chopped-wave tests with the ground current method. However, the evaluation of the current of chopped in this work can confirm that some signals in the full-wave current can be arc extinction signals of a turn-to-turn failure at the beginning of winding. Anyway, the evaluation of the chopped current wave or even the use of time scaling factors, common in some scopes and digital recorders can confirm that some signals in the fullwave current can be arc extinction signals of a turn-to-turn failure at the beginning of winding. Next steps for investigation are better defining the transient component modelling and to investigate the influence of the winding design and the number of winding sections (layers).

VIII.
[1]
0.04 0.08 0.12 0.16 [ms] 0.20

REFERENCES

(file MD1C-SURTO.pl4; x-var t) v: CS c: TNKfactors: 1 1 100 offsets: 0,00E+00 0,00E+00 0,00E+00

[2]

Fig. 23: Reduced full-impulse with failure extinction signal

15.0 *10 3 7.5 0.0 -7.5 -15.0 -22.5 -30.0 -37.5 -45.0 2.0

[3] [4] [5] [6] [7] [8] [9]

C. de Salles, R. J. Bachega, and M. L. B. Martinez, Distribution Transformer Impulse Test Comments, in Proc. 2005 Saint Petersburg - Power Tech. C. de Salles, R. J. Bachega, and M. L. B. Martinez, Distribution Transformer Impulse Test Comments, in Proc. 2005 SIPDA International Symposium on Lightning Protection. R. Rndenberg, Electrical Shock Waves in Power Systems, 1968. IEEE Guide for Transformer Impulse Tests, IEEE Standard C57.981993, Dec. 1993. NBR5380 Transformador de Potncia - Mtodo de Ensaio, Maio/1993. NBR 7570 Guia para Ensaios de Impulso Atmosfrico e de Manobra para Transformadores e Reatores Procedimento Novembro/1982 NBR5356 Transformador de Potncia Especificao, Agosto/1993. IEEE Standard Test Code for Liquid-Immersed Distribution, Power, and Regulating Transformers, IEEE Standard C57.12.90-1999, Jun. 1999. R. D. Fuchs, Transmisso de Energia Eltrica, vol. II. Rio de Janeiro, Livros Tcnicos e Cientficos, 1977.

2.5

3.0

3.5

4.0

4.5

5.0

5.5 [us] 6.0

(file MD1C-SURTO.pl4; x-var t) v: CS c: TNKfactors: 1 1 100 offsets: 0,00E+00 0,00E+000,00E+00

Fig. 25: Reduced full-impulse with the failure signal at the beginning of winding