Validation of power plant transformers

re-energization schemes in case of black-out

by comparison between studies and field tests
measurements
François-Xavier ZGAINSKI, Bruno CAILLAULT, Vincent-Louis RENOUARD

Abstract-- The EHV network and power plant restoration Transmission over-
head lines
plan is a key process of a fast re-energization of customers.
Source Large power Target
The first step after a black-out consists in re-energizing lines Power transformer
and transformers, both on the network and in power plants, as plant
soon as possible, without taking too much risks. To validate the
strategy based on identified schemes, the following methodology
is detailed: first the critical point to deal with is the evaluation, by
using computation tools, of the temporary harmonic overvoltages Fig. 1. Network configuration studied in the case of transformer switching
generated by the transformer re-energization. If the conclusion is The source power plant is a large power plant generally
positive, field tests are performed. Finally, measurements are connected to the 400kV network and is used to re-energize an
compared with simulations results in order to validate the initial
auxiliary transformer. EHV lines are 50 kilometers long, and
feasibility study.
Keywords: black-out, transformer re-energization, field tests. in some cases can reach 250 kilometers. It must be necessary
sometimes to re-energize firstly an intermediate network
I. INTRODUCTION transformer.
In the case presented hereafter, a large network transformer
After a black-out, an initial energizing step is required
(3x357 MVA) has been switched on, located at 50 km of a
before completely restoring the electrical power to consumers.
900 MW power plant. EDF has experience on several schemes
Very weak networks can be used to re-energize, through
where 96 MVA auxiliary power plant transformers are
unloaded EHV lines, one or more other power generating
switched on, with both studies and field tests, but it was the
units. The first steps may consist in:
first time in France that such a large transformer was re-
- identifying the power plants islanded, that can be used
energized with such a network.
to restore power to those out of power in order to
secure them in priority,
II. EVALUATION OF THE TEMPORARY HARMONIC
- re-energizing the auxiliary transformers of the power
OVERVOLTAGES
plants out of power, and supplying the auxiliaries,
- re-connecting as soon as possible power plants together The energization of power transformers may create
through EHV lines, and supplying the load. saturation of the transformer magnetic core and can lead to
The main problems encountered during the first step of re- large harmonic temporary overvoltages due to high inrush
energizing transformers are steady state and transient currents. Key parameters for the analysis of such overvoltages
overvoltages that can damage electrical devices connected to are the following ones :
the network to restore. •• The resonance frequency of the network;
The following figure gives an example of topology of the •• The losses of the network, dependant on its resonance
studied network schemes. frequency, and the load connected to the network;
•• The saturation characteristic of the transformers
feeded by the network;
•• The voltage level at the end of the EHV lines.
In order to minimize both steady state and transient
overvoltages, the setting point of the source power plant
automatic voltage regulator is reduced to its minimum value,
François-Xavier Zgainski, Bruno Caillault, Vincent-Louis i.e. usually 0.9 p.u.
Renouard are working at EDF-DTG (General Technical Department), avenue
de l'Europe, 38000 Grenoble, France (email [email protected],
Furthermore, the levels of overvoltages and inrush currents
[email protected], [email protected]) depend also on the initial conditions of the switching :
•• closing times of the circuit breaker poles;
Presented at the International Conference on Power Systems •• remanent fluxes in the core of the transformer to
Transients (IPST’07) in Lyon, France on June 4-7, 2007
switch-on.
 A. Description of the modeling of the network possible yet to reduce these uncertainty ranges, because of the
The electrical components of the network are modeled differences still remaining between simulations and field tests
using the EMTP tool. measurements ( see part III section C).
The generator is represented by a perfect voltage source In order to consider the uncertainty domain corresponding
with a resistance and an inductance that are obtained by a to these two parameters, which have an influence on the
computation based on the sub-transient values of the network resonance frequency value, 25 couples of values are
inductance and of the time constant. This simplified model is defined. Each couple has the same probability to fit with the
adapted for [100Hz-500Hz] frequency range resonances [7]. real value, and is thus associated with a network frequency
Its simplification is compensated for by taking into account value, which proximity with an harmonic of the 50Hz will lead
uncertainties on the value of the inductance (see section B). to particular resonances.
The line sections are modeled by using matrix PI section For each couple defined, 100 computations are done
model that deals with both electrical and magnetic coupling statistically to deal with the random switching initial
between each electrical conductor. One PI section is used for conditions: closing times of the circuit breaker poles, and
every 25 kilometers line section. In order to represent the remanent fluxes in the transformer core.
losses, parameters are calculated at the network resonance These initial conditions are considered as random
frequency. A good geometrical description of the over-head conditions because they can not generally be imposed and
line towers is provided. controlled. We use measurements in order to have an idea of
The transformers are described by three single phase their extreme values and apply random laws on these
transformers, described by a classical model shown on the conditions. The closing times rules are the following ones : the
following figure, that takes into account the iron losses, the first pole of the circuit breaker is closed anytime on one time
resistances of each conductors, and the short circuits values of period, and the second and third pole are closed with a
the transformer. standard deviation of 20 ms (i.e. one period of 50Hz). The
remanent flux values follow a uniform distribution, and are
RCu2 Ll2 Ll1 RCu1
supposed to reach 0,8p.u. [7]; due to the winding delta
connection always present in the power plant or network
MV side RFe L(λ) EHV side transformers considered, the sum of the three remanent flux
values is assumed to be equal to zero.
Fig. 2. Transformer model used in this study The saturation curve, and especially the Lsat i.e. the final
The non linearity of the transformer magnetic core is slope of this curve, is a key point for the computation of the
modeled by a non linear inductance - characterized by a curve inrush currents but is not very easy to obtain. The transformer
of the flux versus the current (flux(I)) – put in parallel with the manufacturer provides a Lsat slope value with a dispersion
iron loss resistance. For the target transformer, hysteresis is usually considered of +/-20 %. The lower value is taken in the
added, in order to take into account the remanent fluxes in the simulations.
iron core. Each pole of the circuit breaker is represented by an The results of the computations are the estimated
ideal switch in order to statistically modify the different probability to get overvoltages higher and longer than the
closing times of each pole. limits prescribed. The voltages and currents curves are also
analyzed to verify the realism of the computations, as
B. The methodology defined for the studies
illustrated below.
The simulations are made to check that the overvoltages
levels and durations stay under the admissible limits given by C. Results of simulations
the constructor, or, if not, within limits and proportion judged The following figure gives the simulation results obtained
acceptable, considering the strategic importance of the in one of the worse cases computed at the switching of the
transformer, situated in a power plant or on the network, and transformer.
considering the potential different solutions existing for re-
energizing it. 600000
In order to take into account the uncertainties on the key 400000
parameters of the network, a large number of computations V1 (V)
have to be performed. 200000 V2 (V)
V (V)

All the modeling and simulation hypothesis result from 0 V3 (V)

discussions with EDF-R&D about the study methodology [4]
and from the comparisons made between computations and -200000
field tests data. -400000
First of all, the inductance of the generator is taken as a t (s)
+/- 15% value around the direct-axis sub-transient reactance -600000
value (X"d). The phase-to-earth capacitance of the over-head 0 0,2 0,4 0,6 0,8 1 1,2 1,4

lines (Cϕ/t) is supposed to be known at +/- 5%. It hasn’t been Fig. 3. Three 400kV phase to earth voltages simulated in one case.
 The currents and the voltages signals can be distorted
during several seconds. A natural frequency resonance is
superimposed on the fundamental frequency because of the
resonance between, on the one hand, the inductance of the
generator and the non linear inductance of the transformers,
and, on the other hand, the capacitance of the lines. If the
resonance frequency is close to an harmonic of the 50Hz, the
switching can lead to high resonances, during several seconds
(see fig. 4).

600000
EMTP-3.1 (300 curves)
400000

200000 V1 (V)
V (V)

V2 (V)
0
V3 (V)
-200000

-400000
t (s)
-600000
0,4 0,42 0,44 0,46 0,48 0,5

Fig. 4. Zoom on the temporary overvoltages of V1.

The peak voltage values may reach or exceed 2 p.u.
EMTP-RV (300 curves)
The resonance and therefore the risk of high and long Fig. 5. Peak values of the voltages simulated on 4 seconds for 100
overvoltages on the transformers connected on this “weak” simulations with the 2 codes : EMTP-3.1 and EMTP-RV
network depend on the network parameters : the direct-axis It can be also underlined that the minor loop parameters in
sub-transient reactance (X"d) of the generator and the phase-to- the hysteretic reactor model of EMTP-RV [5] have to be
earth capacitance of the over-head line(Cϕ/t). optimized in order to compare with the EMTP 3.1 version.
The frequency depends of course on the length of the line With these fittings, a good agreement was found between the
and on the type of source power plant, but also on the shunt two codes. It can be noted that EMTP-3.1 can be less stable
reactor that can be connected. In our cases, the frequency than the new version: in fig 5, one case simulated –i.e.
values of the networks that are considered stand between 150 corresponding to one set of initial conditions- with EMTP-3.1
and 500 Hz. Thus, the 3rd to the 10th harmonics are in the must be analyzed and verified with other computation
frequency spectrum of the networks studied and can be excited parameters.
by the re-energization of the transformer.
III. FIELD TESTS AND MEASUREMENTS
D. Comparison of the 2 versions of EMTP for
the modeling of transformer switching Field tests are performed in order to check the studies done
with EMTP. Measurements of voltages and currents are
We are now using both EMTP 3.1 [1] and EMTP-RV [2],
performed during the tests.
and trying to analyze the differences in the models of these two
codes : the main problem concerns the hysteresis model that is A. Main objective of the measurements on the
used for the target transformers. Some work has been done on network and general considerations
the parameters values needed in the EMTP-RV model in order When transformer is switched on, we have to check that the
to fit with the original curve flux(I) used in EMTP 3.1, but overvoltages stay under the admissible limits given by the
their determination remains quite difficult due to the few data constructor. Also, particular attention is kept on the presence
provided by the manufacturers. of harmonics. In case of emergency, we have to alert the
In order to compare simulation results between these two network operator and switch off the transformer supplied as
software, we have drawn the following figures that show the quickly as possible.
waves of the peak values of the phase-to-earth voltages for 100 The following figure shows the location of the measuring
simulations computed, which gives 300 curves for each points and the sampling rate chosen for the data acquisition
computation (one per phase). during the tests. Four teams are located along the network in
order to measure and get data from the switching.
 • Switching times
First, the switching times are obtained from the measured
currents. The delay between the maximum of one phase
Control Network Target transformer
voltage and the apparition of a disturbance on the currents is
Systems
20 Hz 5 kHz 5 kHz 5kHz
measured. This delay between the closing times can reach
3V3I 3V3I 3V
almost 80 ms, which is the worse case recorded; delays as high
as 20 ms are however frequent. The closing times can be
Fig. 6. Location of the measurements during field tests
The measurements allow to get interesting data in order to obtained from the recordings with an evaluated precision of
improve the models adopted for the simulations. The +/-1 ms; in some cases their evaluation is less precise.
measurements made on the automatic voltage regulator of the • Remanent fluxes
generator are useful to check the way the generator voltages The remanent fluxes in the core for the three phases of the
are controlled, in case of harmonics; those made on the LV transformer can be obtained via the integration of the voltages
side of the target transformer are necessary for the measured on the transformer windings during its
determination of the remanent fluxes in the transformer; the disconnection. Voltages are usually recorded on measuring
measurements made on the speed governor are used in the voltage transformers directly connected to the low voltage
following phase of supplying power to the auxiliaries. windings.
The measurement device is both a recorder and an The following figure illustrates the calculation of these
oscilloscope. The capacity of our equipment allow to record three fluxes.

PHI1 [V]
quick phenomenon during a long time. The accuracy of the 900

acquisition devices is better than 1%.

600

B. Precision on the measurements

300
The precision on the measurements must be taken into
account in the criteria of comparison between simulations and 0

measurements. -300
The voltages are measured via on site measuring voltage
PHI2 [V]

transformers. The error on the voltages are less than 0,5% for 600

the voltages between 0,8 and 1,2 Un. For the higher values, the
300
error can reach 3% for the phase–to-earth voltages, and 6% for
the phase-to-phase voltages. 0

The inrush currents are generally measured with both on

-300
site protective and measuring current transformers. The
precision depends thus on the tests values reached, compared -600

to the rated values of the current transformers. On measuring

PHI3 [V]

1500

transformers for example, the specified precision has not been 1200

defined for currents values lower than 10% of the current

transformer rated value, which can limit the validity of the 900

comparison with simulations, especially for fitting the damping 600

effects [3].
300

0
61,8 62,1 62,4 62,7 63 63,3
IV. COMPARISON BETWEEN SIMULATIONS AND [s]

MEASUREMENTS Fig. 7. Computation of the remanent fluxes on the three phases of the
transformer
The initial conditions of the tests are evaluated from the
The starting of the integration is determined in order to get
measurements. These data are put in our models, always
steady-state fluxes without any DC component before this
taking into account the dispersion on the network
disconnection. Nevertheless, we have a superimposed 50 Hz
characteristics. Then, a rather good comparison between
AC component in the measured signal probably due to the
measurements and simulations allows to validate the re-
capacitance of the circuit breaker. Because of the remanent
energization scheme. Furthermore, both models adopted and
induced voltages and therefore sinusoidal fluxes after the
their data are improved.
disconnection of the transformer, we consider that the
A. Initial conditions remanent flux values can’t be obtained with a precision better
The initial conditions are obtained via the analysis of the than +/- 10%.
measurements, i.e. the closing times of the three poles of the B. Comparison criterion in order to compare
circuit breaker that switch on the transformer, and the simulations to measurements and validate the models
remanent flux values in the transformer iron core.
The inrush currents give a lot of informations: the initial
conditions (instants of switching on), the peak current values
and their attenuation, and help to build the magnetization •• Influence of the initial conditions :
cycle of the transformer.
-700
The comparison criterion between measurements and
350 400 450 500 550 600 650 700 750 800
simulations data is firstly based on the maxima of inrush -750
currents. A mean square error computation allows to fit the -800
computed values within the uncertainty range adopted. Other -850

I2max(A)
type of comparison can also be used, like the comparison of
-900
the shapes of both the voltages and currents curves (peak
values, attenuation,… ) or the comparison of the harmonic -950
rates and attenuation. -1000
Simulations
C. Sensitivity analysis -1050 Measure
The analysis presented has been done on the peak values -1100
I1max (A)
of two of the three currents, reached at the re-energization of
the transformer. One simulation parameter is changed at a Fig. 10. Influence of the initial conditions on the peak currents computed and
compared to the measured currents
time in order to have an idea of the sensitivity on the results.
Initial conditions, that is to say the remanent fluxes
We have developed a software based on Matlab, that allows
multi parameters analysis and comparison between (supposed to be known in a range of +/-10%) and the closing
simulations and measurements. In the following figures, we times of the circuit breaker (supposed to be known in a range
plot the maximum current on phase 2 (I2max) as a function of of +/-1ms), seem to be in that case the most influent factors.
the maximum current on the phase 1 (I1max) : both
simulation results and measurements are plotted for It is therefore possible to adjust and optimize these
comparison. uncertainties in order to try to fit the current maxima
•• Influence of the uncertainties taken in the models: (measurements and simulations). As an illustration, the
following figures show, on the currents waves, the comparison
-750
between measurements and results of simulation.
I2max(A)

-800350 400 450 500 550 600

-850 1000 J1m(A)
800 J2m(A)
-900 600 J3m(A)
-950 400
200
-1000
I (A)

Simulations Lsat=0.16pu 0
-1050 Measure -200
Simulations Lsat=0.2 pu -400
-1100 Simulations Lsat=0.24 -600
-1150
-800
-1200 -1000
I1max (A)
0,08 0,1 0,12 0,14 0,16 0,18 0,2 0,22 0,24 t (s)
Fig. 8. Influence of the Lsat value on the peak currents computed and
compared to the measured currents Measurements
The previous figure shows that for a chosen set of initial 1000
J1computed (A)
conditions values, increasing the Lsat value within its 800
J2computed (A)
uncertainty range leads of course to decrease the values of the 600
J3computed (A)
400
peak currents. The influence is significant, and it shows that 200
I (A)

the hypothesis of a minimum value for Lsat , made in the study 0

to stay conservative, is not correct. -200
-400
The influence of the generator reactance (L) and of the
-600
over-head line phase-to-earth capacitance (C) values is also -800
taken into account. For one value of Lsat, 5x5=25 couples of -1000
values (L,C) are simulated. They are displayed on fig. 8. It 0,08 0,1 0,12 0,14 0,16 0,18 0,2 0,22 0,24 t (s)
appears not easy to determine precisely the values of L and C Simulations
from the simulations, because their influence on the current Fig. 11. Comparison between the measured currents and the simulated
maxima is low compared to the precision of the measurement currents
transformers. By the means of a Fourier analysis, the A rather good agreement is obtained on the comparison of
uncertainty domain could however be reduced, especially if it the current shapes and maxima; in the present case, the error
is possible during field tests to switch the lines several times. on the maxima is less than 5%. Nevertheless, it is not easy to
determine an unique combination of parameters, inside the
domain defined by the uncertainty values considered.
Some difficulties are also met in representing the current
attenuation few periods after the switching of the transformer. in EMTP”, IPST International Conference on Power Systems Transients, New
Orleans.
I (A)

0 [6] A. Gaudreau, P. Pichet, L. Bolduc, A. Coutu, “No-load losses in

-100 transformer under overexitation/inrush-current conditions : Tests and a new
model”, IEEE Transactions on Power Delivery, Vol.17, N°4, Oct 2002.
-200 [7] CIGRE WG 33-10, “Temporary overvoltages – Test case results”,
-300 Electra n°188, pp70-87, February 2000.
-400
-500 VII. BIOGRAPHIES
-600 François-Xavier ZGAINSKI was born in Oran (Algeria), on may 10, 1970.
-700 He graduated from the “Institut National Polytechnique”, Grenoble in 1993,
J2m(A) and received a PhD Thesis in electrical engineering in 1996.
-800 J2computed (A) His employment experience includes the Cedrat Research Compagny and the
-900 French Navy – Grenoble/Toulon (1993-1996), EDF R&D in Paris (1996-
t (s) 2000), and in 2000, he joined the EDF General Technical Department (DTG)
0,05 0,15 0,25 0,35 0,45
in Grenoble. His special fields of interest include modeling and simulation.
Fig. 12. Comparison between the measured currents with the protective CT His main activity deals with transient simulation of temporary overvoltages
(J2m) and the simulated current (J2 computed) for the phase n°2 for network restoration after a black-out and field tests especially in the case
The results of the simulation are non conservative. of transformers re-energization. He is member of the Society of Electrical and
Electronics Engineers in France.
In this type of networks, the attenuation is hard to model
because of the low level of the load connected to the network
compared to the network losses. Work is still needed for a Bruno CAILLAULT was born in Parthenay (France), on april 2, 1961. He
graduated from the “Institut National Polytechnique”, Nancy, in 1983, and
good modeling of the network losses in the generator, the
received a PhD in mechanics engineering in 1987.
transformers, the lines [6]. Investigations are currently done His employment experience includes the French Atomic Energy Commission
with EDF R&D on the models of these elements, by (1984-1987). In 1987 he joined the EDF General Technical Department
comparison with measurements data. The investigations (DTG) in Grenoble. He is currently technical expert in the “performance of
power plants” team, and is particularly involved in the fields of power system
concern in particular the generator model, in view to analyze restoration studies and tests, and of hydropower generators voltage control
the impact of a complete model for in one hand represent systems, in relation with network ancillary services. He is member of CIGRE
better the losses, and in the other hand model the automatic and of the Society of Electrical and Electronics Engineers in France.
voltage regulator because of the duration of the resonances
Vincent-Louis RENOUARD was born in Amiens (France), on april 3, 1979.
sometimes reaching several seconds. He graduated from the “Institut National Polytechnique”, Nancy, and received
a postgraduate diploma in electrical engineering in 2003.
V. CONCLUSIONS His employment experience includes the Merlin Gerin Electric
Uninterruptible Power Supply in Grenoble (2004-2005), and in 2005, he
This paper was focused on the real case of the re- joined the EDF General Technical Department (DTG) in Grenoble. His
energization of a large network power transformer, from a special fields of interest include modeling, transient simulation of temporary
power plant located at 50 km. After the study, it was decided overvoltages for network restoration after a black-out.

to try the field tests, and measurements were performed. The

comparison between the measurements and the simulation
results, and a sensitivity study, are presented. This paper
underlines the need of field tests in order to understand the
differences still remaining between measurements and
simulations, and to improve the modeling of transient
overvoltages with EMTP. Thus, because of the lower damping
effect shown by the simulations, it appears necessary to
improve the modeling of the losses in the network elements,
i.e. the generator, the lines and the transformers, in presence of
harmonics.

VI. REFERENCES
[1] Hermann Dommel, “EMTP Theory book”; Hermann Dommel, 1992
edited by Microtran Power System Analysis”, Vancouver, Canada.
[2] J. Mahseredjian, S. Dennetière, L. Dubé, B. Khodabakhchian, L. Gérin-
Lajoie, “On a new approach for the simulation of transients in power
systems”, IPST 2005, Montréal.
[3] Technical Notice of The EDF Transmission Network Division on the
measurement transformers. 1991.
[4] M. Rioual, C. Sicre, “Energization of a no-load transformers for power
restoration purposes. Impact of the sensitivity of the parameters”, in Proc.
2001 IPST International Conference on Power Systems Transients, pp 221-
227, Rio de Janeiro, Brasil, June 2001.
[5] Sébastien Dennetière, Jean Mahseredjian, Manuel Martinez, Michel
Rioual, Alain Xémard, “On the implementation of a hysteretic reactor model