Вы находитесь на странице: 1из 8

A Algorithm for the Three-Pole Controlled Auto-Reclosing of Shunt

Compensated Transmission Lines with a Optimization for the second and


third Pole



G. Pilz*, P. Schegner
Technical University of
Dresden (Germany)



C. Wallner
Siemens AG Berlin,
(Germany)



H. M. Muhr, S. Pack
University of Graz
(Austria)

Abstract
This paper presents an algorithm for auto-reclosing of shunt compensated transmission lines. The
algorithm uses signal analysis methods to estimate the future voltage shape over the circuit-breaker.
Based on this voltage shape switching instants are estimated in the zero crossings. The minimization
of the probability of a prearcing is reduced be adding time shift parameters. These parameters take the
limited rise of strength recovery of the circuit breaker and the mutual influence of the coupling of the
poles by a three-pole auto-reclosing into consideration. The advantage of the descript method is
significant reduction of the switching overvoltage.
Keywords
Controlled Closing, Switching Overvoltage, Shunt Compensated Transmission Lines, Pattern
Recognition, Prony
1 Introduction


During the dead time of an auto-reclosing of a shunt compensated transmission line transient
phenomena on the line side are excited. The transient phenomena depend on the parameter of the
transmission line and the level of compensation. The reclosing after the dead time may generate
overvoltages, which are up to four times higher than the nominal amplitude of the line-to-earth
voltage. Normally a lightning arrester or a closing resistor had to be installed to limit the overvoltage.
A powerful reduction of the switching overvoltage is also possible by reclosing the transmission line
in a optimal switching instant. To do this it is necessary to analyze the voltage over the main switching
contacts, to predict the voltage shape and to identify a optimal switching instant. The voltage on the
bus-bar side of the circuit-breaker is the nominal voltage, which is sinusoidal with constant amplitude
and frequency. The parameters of the voltage on the line side of the circuit-breaker differ in frequency,
damping ratio and amplitude. The waveform of this voltage depends on the design of the transmission
line and the connected compensation equipment. The resonant circuit (shunt reactor and transmission
line capacitance) and the induced voltage from the other phases define the resulting line-to-earth
voltage of each phase. The frequency of voltages of the different phases can differ. During the dead
time of a single-pole auto-reclosing two phases are still feeded the nominal voltage, that means
operated with the nominal frequency. During a three-pole auto-reclosing the frequency of each phase
is mainly defined by the resonant circuit of the individual phases. The real time processing ability is a

pilz@ieeh.et.tu-dresden.de





21, rue d'Artois, F-75008 Paris
http://www.cigre.org
CIGR
Session 2004
A3-115
2

Figure 1: Estimation of the future voltage (80%
compensation degree) with pattern
further criterion for the development of an algorithm for controlled closing. The maximum calculation
time is limited by the dead time of the auto-reclosing cycle, which depends on the voltage level.
Two algorithms for prediction of the future voltage are presented in chapter two. The first algorithm
is based on the pattern recognition. This method analyzes the envelope curve of the voltage signal.
Therefore it is not possible to reclose in the zero crossing of the voltage, which is the optimal instant
for this switching task. The reclosing operation can take place only in a nodal point of the voltage. The
state of the art applied methods are using pattern recognition algorithms [1], [2]. The second algorithm
is based on the Prony Method. The method composes the voltage signal by a sum of sinusoidal,
exponential damped functions. The parameters of these functions are the amplitude, the damping
factor, the frequency and the phase shift. The lowest recognizable frequency is independent of the
evaluation window of the Prony Method. Another advantage of the Prony Method is a high resolution
in the frequency domain. Signals with small frequency differences will be exactly determined. Via the
estimation of the signal components it is not only possible to calculate accurately the voltage zero
crossing across the switching contacts of the circuit-breaker as possible switching instant, but also the
rate of voltage rise in this point. So it is now possible to shift the switching instant, if the rise of
voltage in the zero crossing is greater than the rise of the characteristic curve of strength recovery of
the circuit-breaker. The switching overvoltage will be minimized. The connection between the rise of
voltage at the zero crossing and the strength recovery of the circuit-breaker is explained in chapter
three. In addition the selection of the optimal sequence of the switching instants of a three-pole auto-
reclosing will be in this chapter discussed. The selection of the switching instant of the second and
third pole is in this case complicated, because after closing the first pole the estimated zero crossings
for the second and third poles are no longer valid. A recalculation of the future voltage shape of these
two poles is not possible, because postpone between the closing of the three phases should be minimal.
The first algorithm is optimized for a standard processor. The second algorithm is more complex and
calculates optimized switching instants for the second and third pole. This leads to an additional
reduction of the switching overvoltage.
In chapter four the results of the simulations are presented. First the simulations of switching
compensated transmission lines with ATP are illustrated. A sensitivity analysis could be accomplished
by varying the parameters of the transmission lines (e.g. degree of compensation, length of the
transmission line, one or three pole auto-reclosing). The calculated waveforms are the input data for
the programmed real-time algorithms. The programming takes place in MATLAB. The results of the
real time calculation are the switching instants. With this result a complete simulation of the auto-
reclosing is carried out. The breaker operating time and the limited rise of characteristic curve of the
circuit-breaker are considered. The quality of a controlled reclosing is determined by the size of
calculated switching voltage.
2 Estimation of the Future Voltage
2.1 Pattern Recognition
The algorithm for the estimation of the future
voltage must be work in a loop. This loop is
divided into individual time windows. The
voltage over the circuit-breaker is scanned in
the first window. The envelope of the scanned
voltage is calculated and likewise the voltage
over the circuit-breaker is scanned in the
second window. The calculation of the
envelopes for the second scanning, the shift
of the first envelopes at the window length
and the comparison of the two envelopes take
place in the third window. The voltage over
the circuit-breaker is just as again scanned. If a correspondence between the envelopes of the first and
second window is detected, a controlled switching instant could be calculated. The envelope of the
second data window is shifted at two window lengths for this purpose and a node of the voltage is
defined as switching instant. If no correspondence between the both envelopes is detected, the
envelope of the first window will be deleted and the envelope of the second window is used as
3

Figure 2: Estimation of the future voltage (30%
compensation degree) with pattern
recognition, (a) window length
100ms, (b) window length 90ms
reference sample. This reference sample must
then be compared with the envelope of the
third window. A valid result, the controlled
closing operation or the expiration of a defined
time slot interrupts this loop. The time slot
defines, how long the control device may delay
the closing signal to the circuit-breaker.
The absolute value of the voltage shape of
an 80% compensation transmission line during
the dead time an auto-reclosing is represented
in Fig.1 (thin black line). The strong vertical
lines mark the borders of the individual
windows (length 100ms). In this example the
comparison between the envelope of the first
(green squares) and the second window is
positive. Therefore the envelope of the second
window (blue circle) is shifted to the time of
the fourth window. In this future window the
switching instant at the time index 0,39s is
selected. To give the possibility to compare the
result, the absolute value of the voltage for this
fourth window is also shown in the figure.
In Figure 2a the same task is illustrated for a compensation degree of 30%. The results of the
estimation are insufficient. But if the individual window length were made smaller (length 90ms, see
Figure 2b), the result for a controlled closing would be sufficient. The strong dependence of the results
on the selected window length is a big disadvantage of this method, because the estimation must take
place with a fix window length for all possible compensations. Further disadvantages of this method
are the neglect of the damping of the voltage and the ability only to estimate switching instants in the
nodal point of the voltage. Especially the switching instants are too inaccurate for a three-pole auto-
reclosing, because the influence of the individual poles among themselves can not be taken into
consideration. A better estimation of the future voltage is necessary for this task. The advantages of
this method are the lower demand of power of microprocessors and the allowed high noise to signal
ratio.
2.2 Prony Method
The estimation of the future voltage via the Prony Method works also in a loop with data windows.
These windows have likewise a constant length. The voltage over the circuit-breaker is scanned in
each individual window. The N data samples x[1],,x[n],,x[N] with the constant sample interval T
of the first window are the starting points of the calculation. This vector can be described with the
following summation with a p-term complex exponential model.
( )( ) [ ]

=
+ + =
p
k
k k k k
j T n f j A n x
1
1 2 exp ] [
(1)
In this equation A
k
is the amplitude,
k
the damping factor in s
-1
, f
k
the sinusoidal frequency in Hz
and
k
the sinusoidal initial phase in radians. If the data samples are real, the complex exponential
must occur in complex conjugate pairs of equal amplitude. The eq. (1) may expressed in the form

=
p
k
n
k k
z h n x
1
1
] [ ,
(2)
where the complex constants h
k
and z
k
are defined as
( ) j A h
k k
exp = ,
(3)
( ) [ ] T f j z
k k k
2 exp + = .
(4)
The goal for the Prony Method is to minimize the squared error over all sampled data with respect
to the complex parameters and the number of exponents. A detailed representation of the algorithm is
in Appendix A and in [3].
4
Figure 3: Estimation of the future voltage with
Prony Method, (a) 80% compensation
degree, (b) 30% compensation degree

Figure 4: Connection between prearcing voltage
characteristic curve of the circuit-breaker
and the voltage
During the second window in parallel
the estimation of the parameter for the first
window and data sampling are carried out.
It is necessary to examine the quality of the
computed parameters. Two possibilities for
the examination of the accuracy of the
estimated parameters exist. In the first
method the signal shape of the first
window is recalculated based on the
estimated parameters of the signal. By a
comparison between the measured and the
recalculated signal, the quality of the
estimation could be estimated. In the
second method the estimated parameters
are used to recalculate and compare the
signal of the second window. This
comparison could take place during the
third window. In opposition to the first
method, the output of switching instants
could only be done for the fourth window.
The significance of the second method is
greater. The zero crossing of the future
voltage can be used as switching instants.
The estimation of the future voltage via the Prony Method for the two examples from chapter 2.1 is
represented in Fig.3. The input vector (green squares) and the estimated voltage (blue circle) are
illustrated in the first respectively in the third and fourth window. The quality of estimation is very
good.
The correct calculation of signal components with periods greater than the length of the window is a
big advantage of this method. Also the exact estimation of the future shape of the voltage, including
the correct zero crossings, is very important for one-phase and especially for three-phase auto-
reclosing. Thereby it is possible, to reduce the switching voltage to the theoretically minimum value.
The demands on the signal to noise ratio of the input signal and the necessary power of the
processor are much higher for this method. The algorithm gives only incorrect results in present of a
high noise level. Analog-to-digital converters are one possible source of the noise. Algorithms for the
reduction influence of the noise are given in [4] and [5].
3 Algorithm for Selection of the Optimal Switching Instants
Even with accurate estimated zero crossings
of the voltage a prearcing occurs, which leads
to high switching voltage. Fig.4 illustrates
the connection between the prarcing voltage
characteristic curve of the circuit-breaker and
the voltage. The value of the prearcing
voltage is decreasing during a closing
operation. If the rise of voltage in the zero
crossing is steeper than the curve of
prearcing voltage characteristic, a prearcing
is the consequence (curve of prearcing
voltage characteristic S
1
in Fig.4). The
switching voltage can be minimized by a
time shift of the estimated switching instant from the zero crossing (S
2
in Fig.4). It is necessary to
develop an algorithm, which respect to this relation.

5

Figure 5: Time shift parameters for circuit breakers
with prearcing voltage characteristic
curve of 75kV/ms, 100kV/ms and
125kV/ms
Other point is the fact that the system status for the second and third closing pole of a three-pole
auto-reclosing during the reclosing operation is changing. Therefore the estimated zero crossings are
no more valid. A new estimation is not possible. The algorithm for selection of the optimal switching
instants must consider this problem also.
3.1 Single-Pole Auto-Reclosing
The estimated future voltage shape will be divided into possible switching windows. The zero crossing
with the smallest rise of the voltage will be selected within each window. These zero crossings are
then the possible switching instants. The length of the windows must meet two criteria. The length of
the window should be not too short, because otherwise no zero crossings with small rises are existing
in the window. On the other hand the window length should not too large, because otherwise no fast
reaction to a closing command is possible.
Two algorithms are illustrated to minimize the switching voltage under consideration of the limited
rise of the prearcing voltage characteristic curve. By an addition of a fixed time parameter to the
estimated zero crossing the switching instant will be shifted in positive direction. The probability of a
prearcing decreases with the magnitude of the time shift parameter. But the smallest possible
switching voltage will also increase. Anyway a small probability of prearcing will be left, if the time
shift parameter is not great enough. The addition of a variable time shift parameter is a better solution.
The rise of voltage in the zero crossing is the parameter for the size of the variable time shift
parameter. The calculation of this parameter is easy with the knowledge of the exact future voltage
shape. The connection between the rise of voltage and the variable time shift parameter will be
estimated by a sensitivity analysis. A large number of opening operations of compensated transmission
lines were simulated with ATP. Thereby the input parameters for the simulation (e.g. compensation
degree, length of transmission line) were changed in a wide range. Then a virtual curve of strength
recovery was lined up at all zero crossings. If the intersection point between the virtual curve of
prearcing voltage characteristic and the voltage is before the zero crossing, this curve will be time
shifted as long as the intersection point is after the zero crossing. The rise of voltage at the zero
crossing and the shift parameter are stored.
The results of the simulations are illustrated in Fig.5. The time shift parameter for a circuit-breaker
with a rise of the prearcing voltage characteristic curve of 100kV/ms (system level 380kV) in relation
to the rise of the voltage at the zero crossing is represented with red crosses. No time shift is necessary
up to a rise of prearcing voltage characteristic curve of 60kV/ms. The picture shows, that the
magnitude of the time shift parameter depends not only on the rise of voltage in the zero crossing,
because the voltage on the line side consist of several harmonics. The accuracy of the descript method
is sufficient for a real time application. The approximated straight line at the left side of the scatterplot
gives the minimum time shift parameter for a circuit-breaker with a prearcing voltage characteristic
curve of 100kV/ms. No time shift parameter will added if the rise of voltage at the zero crossing is
smaller then 60kV/ms. The value of the time shift parameter for rises which are greater then 60kV/ms
will be calculated with the parameter of the approximated straight line.
The approximated lines for circuit-breakers with prearcing voltage characteristic curve of 75kV/ms
and 125kV/ms are also shown in the figure.
With a steeper characteristic the size of the
necessary time shift will be smaller.
3.2 Three-Pole Auto-Reclosing
The first algorithm named method A is an
extension of the algorithm for the single-pole
auto-reclosing. The estimation of the window
with the smallest rise at the zero crossing and
the adding of a first time shift parameter take
place for all three poles. In addition the
switching sequence of the poles during the
reclosing operation had to be estimated and
then a second time shift had to be added to
6
Figure 6: Method A for selection of the swit-
ching instants
Figure 7: Method B for selection of the
switching instants
the switching instants off the second and third switching pole.
An example is shown in Fig.6. The phase L3 is switch at first. The time shift parameters in the poles
L1 and L2 are consisting of two parts. The calculation of the first time shift parameter is based on the
limited rise of the prearcing voltage characteristic curve of the circuit-breaker and the calculation of
the second time shift parameter is based on the sequence of the pole switching.
This algorithm calculates not an optimum for the second and third switching pole. Especially if the
differences between the individual switching instants are great, then high switching voltages could
occur. A second algorithm named method B is optimized. Again the estimation of the windows with
the future voltage shape is carried out. Then the zero crossing with the smallest rise of voltage for all
three poles will be estimated. The pole with the smallest rise at the zero crossing is selected as the
third switching pole, if there are zero crossings in the two other poles before this estimated zero
crossing. The zero crossings in the two other poles are then the first and second switching instants. If
there are not preliminary zero crossings in the two other poles, the pole with the second smallest rise
of voltage in the zero crossing will be selected as the third switching pole. This loop will be executed
up to a valid result.
Fig.7 shows an example for that method. The zero crossing with the smallest rise is in pole L3. The
preliminary zero crossings in pole L1 (second switching pole) and L2 (first switching pole) determine
the other two switching instants. Again a time shift parameters are added to the found switching
instants. This is done identically to method A, dependent on the sequence of the poles and limited rise
of the prearcing voltage characteristic curve of the circuit-breaker.
The small time differences between the switching instants are the advantage of the method. But the
second method has higher demands on the power of the processor than the first method.
4 Results of the Simulation
Simulations of auto-reclosing were carried out in ATP for the examination of the algorithms. The
parameters of the transmission line (e.g. length, transposition, damping), short-circuit power of the
supply system, compensation degree and the position of the failure was varied. Disconnections were
simulated in a first step. The voltage shapes over the circuit-breaker are the input data for the real-time
algorithm. The toolbox xPC from MATLAB was used as real-time environment. A Pentium PC with
350MHz was the target hardware. This processor meets the demands for a single pole auto-reclosing.
A higher performance is needed for a three-pole auto-reclosing. This switching task was simulated in
an offline modus. The only difference between the target hardware and a real control device is the
connection to the input and output signals. In the used real-time environment the input datas are stored
in a data file. The implementation of a I/O-interface will be released in later version. The estimated
switching instants are the outputs of the algorithms. A complete simulation of an auto-reclosing
including the controlled closing is now possible in ATP. The circuit-breaker will modelled in
consideration of his operating time and his prearcing voltage characteristic curve. The operating time
7
Figure 8: Example for the simulation of a
controlled closing during a three-pole
auto-reclosing
Figure 9: Results of single-pole auto-reclosings
with different time shift parameters
Figure 10: Results of three-pole auto-reclosings
with method A and B
was defined with 50ms and the rise of the
prearcing voltage characteristic curve with
100kV/ms for a voltage level of 380kV. The
calculated switching voltage is the
performance index of the algorithms.
Fig.8 shows a simulation for a controlled
reclosing of a three-pole auto-reclosing with
the algorithm of method A. The switching
voltage of the first switching pole is
approximately 60kV. The voltage over the
circuit-breaker breaks down after the
intersection of the voltage line with the
prearcing voltage characteristic curve of the
circuit-breaker. Also the induction of transients from L1 in L2 and L3 is shown. The time shift
parameter of the second switching pole was sufficient, but for the third switching pole the time shift
parameter was too small.
4.1 Single-Pole Auto-Reclosing
Fig.9 shows the results of the single-pole auto-
reclosing with different time shift parameters.
The switching voltage is presented on the x-
axis and the relative number of the estimated
switching variations on the y-axis. There are
round about 10000 auto-reclosings of different
transmission line systems had been simulated.
The switching voltage of a controlled reclosing
without a time shift parameter is marked with a
red line. Very low switching voltages are only
achieved for 30% for all reclosings. The rise of
voltage in the estimated zero crossings are
greater then the rise of the prearcing voltage characteristic curve of the circuit-breaker for the
remaining reclosings. The maximum switching voltage is 630kV.
If the estimated switching instant is shifted with a fix time shift parameter a very low switching
voltage could not be achieved for all simulations. High switching voltages were estimated only for
12% (time shift of 1ms blue line) respectively for 9% (time shift of 1.5ms green line) of all
reclosings. The best results will be achieved with a variable time shift. No prearcing in all reclosings
was detected. The maximum switching voltage was 70kV, which is nearly a tenth of the maximum
switching voltage for the other simulations.
4.2 Three-Pole Auto-Reclosing
Fig.10 shows the results for the three-pole
auto-reclosing. The switching voltages for the
first switching pole (red) are similar for both
methods, because there are no great differences
between the methods for this first pole. The
switching voltages of the second (blue) and
third (green) switching pole are in general
greater. The switching voltages of these poles
for method A (solid line) are higher then for
method B (dashed line). It is a greater time
shift parameter for the second and third
switching pole in method A necessary then for
method B to prevent a prearcing. Thats the reason for different switching voltages. On the other hand,
a greater switching window for method B is necessary to get these results. If the switching windows
8
for both methods are equal, the results for the second and third switching pole will be similar. Both
methods will minimize the switching voltage and the influence of the coupling of the three-poles.
5 Conclusion
The advantages of the exact estimation of the future voltage for a controlled closing operation of a
shunt compensated transmission line during an auto-reclosing have been shown in this paper. The
closing operations in the optimal switching instant ensure a very low switching voltage. The
estimation of the exact future voltage shape minimized the probability of a prearcing. The minimum
switching voltage by a single-pole auto-reclosing will be limited by the characteristics of the circuit-
breaker. The algorithms minimize the effects of coupling between the three poles during the three-pole
auto-reclosing. The real-time capability of the algorithm could be verified.
The next goals are the implementation of the algorithms on a testing device and checks on real
transmission lines.
Appendix A
A complete derivation of the Prony Method is in [3]. Thats why only a short introduction of the
estimation of the both parameter h and z is on this place. The N data samples x[1],,x[n],,x[N] with
the constant sample interval T are the starting point.
[ ] [ ] [ ]
[ ] [ ] [ ]
[ ] [ ] [ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]

+
+
=


+

p x
p x
p x
p a
a
a
p x p x p x
x p x p x
x p x p x
2
2
1
2
1
2 2 1 2
2 1
1 1
M M
L
M O M M
L
L
(5)
This vector will be split in a matrix and a vector. The rule of the splitting is in eq.6. The parameter p
specifies thereby the number of individual signals in the input signal. The coefficients of a polynomial
are in the vector a. This vector will be calculated in the first step. The nulls of the polynomial are the
parameter z, which are estimated in the second step.
[ ]
[ ]
[ ]


N x
x
x
h
h
h
z z z
z z z
z z z
p
N
p
N N
p
p
M M
L
M O M M
L
L
2
1
2
1
1 1
2
1
1
1 1
2
1
1
0 0
2
0
1

(6)
The matrix Z must be prepared for the estimation of the vector h. Then the estimation of the future
voltage with the parameter z and h is possible.
References
[1] Froehlich, K.; Hoelzl, C.; Stanek, M.; Carvalho, A.C.; Hofbauer, W.; Hoegg, P.; Avent, B.L.;
Peelo, D.F.; Sawada, J.H., Controlled closing on shunt reactor compensated transmission lines. I.
Closing control device development, Power Delivery, IEEE Transactions on ,Volume: 12 , Issue:
2 , April 1997, Pages:734 740
[2] Cho, K.B., Kim, J.B., Shim, J.B., Park, J.W.,Development of an Intelligent Autoreclosing Concept
using Neuro-Fuzzy Technique An Optimal Controlled Switching for Power System Operations,
CIGRE 1998 Session 13
[3] Marple, S. Lawrence, Digital Spectral Analysis, London: Prentice Hall International, 1987
[4] Pilz, G., Schegner, P., Wallner, C., Analysis of Noisy Voltage Signal with a high Resolution of
Frequency for the Closing of Transmission Lines PSCC 2002, Sevilla
[5] ivanovi, R., Schegner, P., Seifert, O., Pilz, G. Identification of the Resonant-Grounded System
Parameters by Evaluating Fault Measurement Records, IEEE Transactions on Power Delivery,
submitted, accepted (it will be published)

Вам также может понравиться