Presentation 07.1.

A New Approach for In-Situ Calibration of Voltage Transformers Part I: Theoretical Background
Michael Freiburg / Prof. Dr. Frank Jenau, Technical University Dortmund, Germany Florian Predl, OMICRON, Australia / Dr. Michael Krger, OMICRON, Austria For this autarkic method, knowledge of the parameters of the equivalent circuit is necessary. The parameters are determined by measurements from the secondary resp. low voltage side of the transformer. This approach uses low frequency measurements to calculate the parameters with respect to the rated frequency as physical restrictions do not allow measurements at rated frequency.

Introduction
Instrument voltage transformers (IVT) are used in electrical grids for metering or protection purposes. The high requirements regarding their precision demand a calibration of the objects before installation. The accuracy classes of IVT are different for protection and metering transformers [1] and are classified depending on the maximum ratio and phase-angle error between the vectors of the primary voltage and the secondary voltage related to the primary side. This calibration is performed in the laboratories of the manufacturers, operators or testing institutes. Once calibrated, the IVT typically operates without re-calibration for its lifetime. In some cases, e.g. reconstruction of the switchyard, the accuracy of the IVT is reconfirmed with a laboratory test or extensive On-Side measurements [2]. The accuracy of the IVT is dependent on the leakage inductance, the winding resistance and the excitation current at power frequency. Core or winding deformation as a result of external influences for example can thus result in a change of the error of the IVT. In addition it may be interesting to obtain the excitation characteristics of the IVT out of field measurements for ferroresonance analysis or simulation programs. Against this background it is very helpful to have the opportunity to get information about the accuracy class and the excitation curve out of field measurements. Thus, this test can be added to standard measurements in respect of on-site testing resp. condition based maintenance [3, 4] or delivery approval, for example. In this paper the theoretical background of a novel model-based approach is proposed being able to obtain the accuracy class of IVT on-site. It has several advantages to other known methods (model based or with use of a reference object [5, 2]) as it does not need a fingerprint of the IVT based on laboratory tests and it is possible to perform the measurements without the need of having high voltage and heavy equipment on-site available.

Equivalent Circuit and Transformer Diagram

An equivalent circuit of the IVT is presented in Fig. 1. The ideal transformer without any losses is connected to the T equivalent circuit (TEC) of a transformer. The dashed notation indicates the reference to the secondary side. The standard Tec is completed with a concentrated primary winding capacitance Cp as this is influential at secondary side measurements. At frequencies around power frequency, the elements Cps and Cs can be neglected. The nonlinear inductance LH represents the voltage-current behavior of the magnetic unit. LH is assumed to be frequency independent. The dependence on frequency is modeled with RFe.
PRIMARY R1 Ls1

Cps

SECONDARY

n1

n2

I1
Iex

Ls2

R2

I2

VZ1
V1HV V1

VZ2 VCore
Cs

Cp

LH

RFe

Vb

Zb

ideal transformer w=n1/n2

Fig. 1 Equivalent circuit of instrument voltage transformers. Cps and Cs are neglected

Methodology
The model-based approach allows calculating the accuracy class in general by inversely applying the Mllinger-Gewecke Diagram [6] comparable to [5].

The voltage drops VZ1 and VZ2 across the primary and secondary impedances result in a voltage Vb that is different from V1 in magnitude and angle. According to Mllinger and Gewecke [6], this error of the IVT can be determined with the diagram in Fig. 2. The vector of the secondary voltage Vb is drawn on the vertical axis. The end of this vector is considered as the zero-point of the scale. The abscissa is divided in minutes (1% - 34.4min) whereas the ordinate is divided in percent of the primary voltage. The error becomes negative when the secondary voltage vector is smaller than the primary. When the transformer is operated in no-

OMICRON 2013 Instrument Transformer Measurement Forum

Presentation 07.1.2

load condition, the excitation current Iex causes a voltage drop across R1 and Ls1. This is indicated by point A in Fig. 2. The distance 0-A is called noload error. When the transformer is loaded with a burden, the additional current I2 causes a voltage drop across both primary and secondary impedances (R1, Ls1, Ls2, R2). The load error (distance A-B) is independent of the primary voltage and with current I2 proportional to the burden. Once drawn, the error for any load can be easily obtained from the diagram by shifting point B in direction of Sn (change of power level) or by circular shifting point B in direction of cos (change in power factor). Point A is not influenced by the load. When the applied voltage is changed, point A shifts due to the change of excitation current Iex and loss angle . The previous explanations where made with the assumption of a transformer ratio equal to the nominal value. Manufacturers usually perform a winding correction to enable a point of operation in between the positive and negative error limits for the intended operation range of the transformer [7, 8]. In the vector diagram (Fig. 2), the winding correction is considered with a shift of the zeropoint with the percentage correction to -u.
-u
cos

winding correction can be considered in two different ways (see separate chapter). The goal is to obtain the transformation ratio of the ideal transformer in Fig. 1. The secondary winding resistance is measured with a standard DC measurement. The sum of the leakage inductance (Ls=Ls1+Ls) and the effective primary winding resistance R1 can be calculated out of an AC short circuit measurement. Parameters LH and RFE representing the magnetic unit are measured with an open circuit test. The excitation current Iex and the corresponding loss angle are measured for the whole magnetization curve and represent LH and RFE.

Capacitive Influence
The method presented uses open circuit measurements from the secondary side of the transformer to obtain the parameters to allow low testing power. Depending on the IVT, the effective winding capacitance Cp affects the measurement, its effect depends on the ratio w of the IVT as Cp = Cp* w [9]. (1)

B
(winding correction)
Sn

jI2(Xs1 +Xs2)

I2(R1 +R2) jIexXs1

0

Even if the transformer has a large primary winding, Cp was found to be in the range of 100 pF to 400 pF in a wide range of IVT [9]. Considering the nonlinear and time dependent characteristics of the core (Fig. 3), the additional capacitive current increases with frequency and changes the excitation current in magnitude and angle. The capacitive current results in a shift of the hysteresis graph in the zero-point of the diagram (see Fig. 3).
0.2

V1

- u

IexR1

u
0.1
Increasing Cp

Vb
I2 Iex I Ir I1

flux [Vs]

-0.1 -0.2 -3

-2

-1

0 1 current [A]

Fig. 2 Vector diagram of the voltage transformer of Fig. 1 according to Mllinger and Gewecke with the ratio error () and the phase-angle error ()

Fig. 3 Schematic representation of the influence of the effective winding capacitance on the nonlinear time-variant behavior of the magnetic unit of IVT at open circuit measurements from the sec. side

Parameter determination
To be able to calculate the accuracy of the IVT, the parameters of the equivalent circuit of Fig. 1 have to be determined. The parameters are measured from the secondary side of the transformer. The For the methodology described in this paper, the RMS value of the excitation current Iex and the corresponding loss angle are important. As shown in Fig. 4, the capacitive current

OMICRON 2013 Instrument Transformer Measurement Forum

Presentation 07.1.3

compensates the excitation current and the loss angle is shifted to smaller values. This effect is most significant in the linear region of the excitation curve with the highest impedance. In addition, the capacitive current at rated frequency measurements requires high power levels for the test.
0 -20 3.0 2.5 2.0

-40 -60 -80

1.0
0.5 0.0 0 0.1 0.2 flux [Vs] 0.3 0.4

-100

loss angle of core at 50Hz exc. current of core at 50Hz

loss angle of VT with cap. infl. exc. current of VT with cap. infl.

Fig. 4 Excitation current and loss angle of a magnetic core of a 24kV IVT: magnetic unit without primary winding (solid line) and with primary winding (dashed line)

This method uses low frequencies (LF) for the measurement to enable a low test power and low voltages for safety reasons. The capacitive current at the LF-measurement is subtracted [9] or neglected (valid for IVT with a small winding capacitance and turns ratio). When the measuring frequency is reduced, the excitation current decreases and the loss angle shifts due to the reduction in power losses. Thus, power loss models have to be applied to the low frequency data to compensate for this. As the measurement is dependent on the constraint Rm+R2<<LH, where Rm is the measuring resistance, the reduction in frequency should be performed in respect of the form factor of the core voltage. When the core voltage gets nonlinear do to the nonlinear voltage drop across the winding and measuring resistance especially at high inductions the power losses increase due to the arising harmonics.

current [A]

1.5

density (intrinsic material structure) whereas the dynamic losses are dependent on frequency and flux density. The classical eddy current losses as one part of the dynamic losses are dependent on the conductivity of the material whereas the additional eddy current losses are dependent on the response of the material structure on an alternating external field. Depending on the material of the magnetic core, the power losses are nonlinear over frequency. Compared to NO silicon iron, GO silicon iron is more nonlinear especially at low frequencies below approximately 10Hz [11]. Additionally, the power losses over frequency differ depending on the ongoing magnetization process. The power losses per cycle of an IVT core made of stacked GO steel sheets can be seen in Fig. 5. The losses per cycle increase nonlinearly up to about 20Hz before they increase more linearly because of the predominant additional eddy current losses. The nonlinear behavior can be explained as follows: The larger the domains of the material, the higher the velocity of the Bloch Walls at an external alternating magnetization. This results in a high additional eddy current loss. With higher fluxes, the additional losses increase as the number of domains in the material decreases. With increasing frequency, a natural multiplication of domains can be observed what leads to a reduction of the wall-velocity and thus a reduction of the additional eddy current losses [12]. On the other hand, the classical eddy current losses which are dependent on the conductivity increase with the square of frequency.
power losses per cycle [Ws]

angle []

0.2

0.15

0.1

0.05

0 0 10 20 30 40 50 60

frequency [Hz]

Power Losses
The magnetic cores of IVT are typically build of stacked or wound silicon-iron steel sheets of nonoriented (NO) or grain-oriented (GO) material. An external magnetization causes macroscopic (dB/dt) and microscopic (moving domain walls) eddy currents which in turn cause power losses in the form of joule heating [10]. These power losses are divided into static and dynamic power losses and are dependent on conductivity and intrinsic material structure. The static losses are assumed to be frequency independent but dependent on flux

Fig. 5 Power losses per cycle of a magnetic core of an IVT made of GO electrical steel including static power losses (dotted line) and classical eddy current losses (dashed line)

The power losses in Fig. 5 are divided into static losses (dotted lines obtained by an extrapolation to zero from low frequency measurements), classical eddy current losses (dashed lines calculated on the basis of the geometry and the conductivity, approximated with [13]) and additional eddy current losses. To be able to obtain the excitation current at power frequency out of low frequency data, (nonlinear) power loss models have to be applied. With knowledge of the power

OMICRON 2013 Instrument Transformer Measurement Forum

Presentation 07.1.4

losses at rated frequency, the additional loss current is calculated and added to the excitation current under quasi static conditions.

obtained from the calculations of the internal voltage drops. The ratio error is calculated with Fu = (Vb-V1)/V1 *100% and the phase-angle error is calculated with = angle (Vb) - angle(V1). (3) (2)

Influence of Leakage Inductance

As mentioned, the sum of the leakage inductance Ls is obtained from a measurement from the secondary side with the primary shorted. As seen in Fig. 1 and Fig. 2, the model requires a separation of Ls to the primary (LS1) and secondary side (LS2) of the equivalent circuit. The leakage inductance Ls is a function of the geometry of the primary and secondary winding.

Ratio Error and Winding Correction

The ratio is be plotted for the overall voltage range. The obtained error of the model calculations is typically defective as the winding correction is not considered hitherto (see Fig. 7). The reduction of primary turns in order to achieve the accuracy for the intended voltage and burden range causes the secondary voltage to be larger at an applied primary voltage or the primary to be smaller at a given secondary voltage. It is proposed to account for the winding correction in two different ways. The first proposal is based on measurements with a reference capacitance connected to the primary side with a low test signal. The shift in the resonance frequency is used to calculate the exact ratio of the ideal transformer of Fig. 1 [9]. The zeropoint Pz of the ordinate is then set to Pz = (wm/wn-1)*100, (4)

-0.5

FU [%]
-1 -1.5 250

200

150
N

100

U/U [%]

50

100

80

60
1 s

40 L /L [%]

10

Fig. 6 Change of ratio error Fu (equal to in Fig. 2) dependent on distribution of the leakage inductance to primary and secondary side and voltage level

Fig. 6 shows the change of ratio error depending on the distribution of the leakage inductance to the primary side (Ls1/Ls in percent %). When the leakage inductance is concentrated to the primary side only, the ratio error is nonlinear over the primary voltage due to the voltage drop caused by the nonlinear excitation current. The higher the proportion of Ls concentrated to the secondary, the more linear gets the behavior. The separation of Ls also affects the phase-angle error, even more as the ratio error. The error-trend over the primary voltage is used to investigate the distribution of the leakage inductances for this model by comparing the modeled data with data from laboratory calibrations. It is valid with minor errors to use Ls=Ls1 for the calculations for a wide range of IVT.

Calculation and Results

With knowledge of the parameters of the equivalent circuit (R1, Ls1, Ls2, R2) and the excitation current and loss angle (Iex, ) the error of the IVT can be calculated related to the Mllinger Diagram for any voltage of interest. The algorithm starts with the calculation of the secondary current I2 at rated secondary voltage Vb and rated burden Zb. Related to Fig. 2, this algorithm to calculate the error of IVT starts at vector Vb. The distance 0-B and thus the error in ratio and phase-angle can be

with the obtained ratio wm and the nominal ratio wn. The second proposal uses an open circuit measurement from the primary side. As the error plot gives information about the transformer error for any voltage, the open circuit error can also be calculated with the model for the overall voltage range. A subsequent open circuit test at a defined voltage obtains an open circuit voltage error. The calculated open circuit error can then be set to the measured open circuit error. The voltages have to be exactly similar for the comparison. The error curve is thus shifted with the obtained correction factor (Fig. 7). The ratio error becomes very large at low voltages and higher voltages above the nominal value because of the small permeability in this region. The excitation current increases significantly and causes a larger voltage drop across the primary impedance. The burden dependent error stays constant.

OMICRON 2013 Instrument Transformer Measurement Forum

Presentation 07.1.5

0 -0.2 -0.4

FU [%]

-0.6 -0.8 -1 -1.2 0

(winding correction)

without winding correction calibration data from laboratory with winding correction 0.5 1 U/Un

1.5

Fig. 7 Ratio error of a 20kV IVT obtained from the model calculations (solid line), with application of the winding correction (dashed line) and comparison to data from a normative laboratory test (triangles) with the leakage inductance concentrated to the primary side

Fig. 9 Error diagram of the 20kV IVT (Sn=50VA)

Phase-angle error
The phase-angle error is obtained by simply subtracting the angle of the voltages Vb (0) and V1 (3). The phase-angle error is not affected by the winding correction. The calculations are performed for the overall voltage range again.
25 20 model calibration data from laboratory

The error graph of the 20kV IVT is exemplary shown in Fig. 9 for nominal burden Sn, a quarter of the nominal burden Sn/4 and different cos. The variation of error due to a change of the applied voltage (0.8*Un..1.2*Un) is also marked in the diagram. When the error graph is once drawn, the error for any burden can be simply obtained for a constant primary voltage and frequency.

Summary and Conclusion

In this paper, the theoretical background of a novel method to obtain the accuracy class of IVT is proposed. The algorithm is based on the common vector diagram of voltage transformers according to Mllinger and Gewecke. The parameters of the equivalent circuit are determined by open and short circuit tests. This method uses low frequencies for the measurement to enable low testing power level and low voltages. Depending on the IVT the capacitive influence is subtracted or neglected. At IVT with a large transformation ratio (approx. 123kV), the capacitive influence has to be subtracted mathematically first. To compensate for the variation of the current caused by ferromagnetic losses, appropriate loss models are applied. Depending on the accuracy of the loss model the excitation current for power frequency is calculated by adding the loss component at 50Hz to the quasi static excitation current. When the parameters, the excitation current and the corresponding loss angle are known, the error in ratio and phase-angle can be calculated. Depending on the distribution of the leakage inductance, the results are accurate what was verified by comparison with calibration data from a testing lab. Further investigation has to be done on the distribution of the leakage inductance.

angle [min]

15 10 5 0

0.5 U/U
n

1.5

Fig. 8 Phase-angle error of a 20kV IVT obtained from the model calculations (solid line) compared to data from a normative laboratory test (triangles)

The phase-angle error becomes large at low voltages (low permeability of the iron especially at iron cores of NO material) and higher voltages (saturation of the iron core). Again, this behavior is caused by the increasing excitation current and the corresponding loss angle at low core impedance.

Error graph
The results of the ratio error (Fig. 7) and phaseangle error (Fig. 8) can be merged to the error diagram (Fig. 9). The winding correction is considered in the error diagram.

OMICRON 2013 Instrument Transformer Measurement Forum

Presentation 07.1.6

Literature
[1] IEC 61869-3:2011 Instrument transformers Additional reqirements for inductive voltage transformers Bergman, Anders: In situ calibration of voltage transformers on the Swedish national grid. PhD Thesis; Upsala, 1994 Raetzke, Stephanie et al.: Condition assessment of instrument transformers using Dielectric Response Analysis. Cigre 2012 Azcarraga, C.G. et al.: On-site Testing of Instrument Transformers; 2006 Annual Report Conference on Electrical Insulation and Dielectric Phenomena Brandao jr., A.F. et al.: On Site Calibration of Inductive Voltage Transformers. 8th WSEAS International Conference on System Science and Simulation in Engeneering, 2009 IEC 61869-3:2011 Instrument transformers Additional reqirements for inductive voltage transformers Braun, A.: Instrument Transformers, PTB, 1984 (rev. translation of the original German edition, E. Zinn: "Messwandler, 1977) Bauer, R.: Die Messwandler, Springer Verlag, 1953 Freiburg, M. and Jenau, F.: Measurement and Computation of the Excitation Curve and Loss Angle of Instrument Voltage Transformers, 48th International Universities Power Engineering Conference (UPEC), Dublin, 2013

About the Authors

Michael Freiburg (born in 1985) is currently working as a research assistant at the Technical University of Dortmund, Germany. His research interests include Diagnostics of High Voltage Equipment and Material Science. He received the engineering degree in 2010 and he is currently working towards the PhD degree at the Technical University of Dortmund in cooperation with OMICRON electronics, Austria. Frank Jenau (born in 1969) studied at the University of Dortmund and received the Engineering degree in 1994. In 2000, he received the doctor`s degree by the Brandenburg Technical University of Cottbus, Germany. From 2000 until 2009, he worked at different industrial companies and positions in the field of high voltage technology and non-conventional measuring systems in Europe and China. At present, he is a Professor in the Department of High Voltage Technology at the Technical University of Dortmund. His research interests cover the characterization of electrical insulation systems, measurement technology and EMC. Florian Predl (12 November 1986) works for OMICRON electronics as an Application Engineer, mainly focusing on instrument transformer applications since September 2007. He graduated at the Federal Higher Technical Institute in Rankweil, Austria with the focus on high frequency technology.

[2]

[3]

[4]

[5]

[6]

[7]

[8] [9]

[10] Krings, A. and Soulard, J.: Overview and Comparison of Iron Loss Models for Electrical Machines, 5th International Conference on Ecological Vehicles and Renewable Energies, Monaco, 2010. [11] Bertotti, G.: General Properties of Power Losses in Soft Ferromagnetic Materials, Magnetics, IEEE Transactions on, vol. 24, no.1, January 1988 [12] Bertotti, G.: Hysteresis in Magnetism, I. Mayergoyz, Ed. Academic Press, 1998 [13] Namjoshi, K.V. et al.: Eddy-Current Power Loss in Toroidal Cores with Rectangular Cross Section, IEEE Trans. On Magnetics, Vol34, NO.3, 1998

OMICRON 2013 Instrument Transformer Measurement Forum