A Very Fast Unblocking Scheme for Distance

Protection to Detect Symmetrical Faults during Power

Swings

Behnam Mahamedi
Deputy of Supervision on Transmission Grid Security
Iran Grid Management Co. (IGMC)
Tehran, Iran
[email protected]

Abstract—Power swing blocking function in distance relays is protection zones of the relay [1]. After blocking a distance
necessary to distinguish between a power swing and a fault. relay by using this method, zero- or negative-sequence
However the distance relay should be fast and reliably unblocked detection based criteria are used to unblock the relay when a
if any fault occurs during a power swing. Although unblocking fault occurs as it is the most common fault detection method in
the relay under asymmetrical fault conditions is straightforward numerical distance relays [2]. It is obviously known that this
based on detecting the zero- or negative-sequence component of fault detection method fails to unblock a distance relay when a
current but symmetrical fault detection during a power swing symmetrical fault occurs since there are neither zero-sequence
presents a challenge since there is no unbalancing. This paper nor negative sequence component. This is the main
presents a very fast detection method used to detect symmetrical
disadvantage of this method. In [3] authors make use of one of
faults occurring during a power swing. Based on a 50 Hz
the feature of swing center voltage to detect symmetrical faults
component getting on three-phase active power after symmetrical
fault inception and using Fast Fourier Transform (FFT), the
occurring during power swings. This method is based on the
proposed detection method can reliably and quickly detect monitoring of SCV variations under power swing and fault
symmetrical faults occurring during power swing in one power condition. However the method assumes some approximations
cycle, i.e. 0.02 second. This detection method is easy to set and generally not correct. Furthermore the proposed method does
immune to the fault inception time and fault location. Power depend on power swing slip frequency which varies
swing and fault conditions are simulated by using software extensively from a power system to another. Reference [4]
PSCAD/EMTDC®. FFT is performed by using On-Line presents another method to detect symmetrical faults occurring
Frequency Scanner block included in the software. during a power swing based on measuring three-phase active
and reactive power. In [4] if the reactance of local source is
Keywords- Distance relay; Fast Fourier Transform (FFT); close to the reactance of remote source then the proposed
power swing; symmetrical faults; three-phase active power method needs that total system resistance, i.e. local and remote
resistances in addition to line resistance would be negligible in
I. INTRODUCTION comparison with line reactance. This condition is not generally
Power swing affects power systems in different ways met in a power system. Some symmetrical fault detection
including damaging to electrical equipment or causing methods during power swings based on the wavelet transform
malfunctions of protective relays. Because of these defective (WT) are proposed [5], [6]. These methods make use of
effects, detecting power swing properly and quickly is very transients getting on phase current [5] and phase voltage [6] at
important. Generally when a power system is recovered from a the fault inception time. The success of these methods
disturbance like a fault it experiences a power swing. During a definitely depend on sampling frequency like other detection
power swing maximum amplitude of current and minimum methods based on transients since fault transients have a wide
amplitude of voltage occurs simultaneously. This causes the frequency band depending on fault location, fault inception
apparent impedance measured by a distance relay to be within time and fault resistance.
its operating zones. Based on this, power swings can be
misinterpreted as a three-phase fault and the relay may trip. This paper presents a new very fast symmetrical fault
Hence power swing blocking function is included in distance detector for distance relays. This detector is based on the
relays to distinguish between a power swing and a fault. Based detection of frequency component of three-phase active power.
on this function the operation of a distance relay can be It will be demonstrated that during a power swing, three-phase
blocked due to a power swing. However, if an asymmetrical or active power has frequency equal to slip frequency whereas
a symmetrical fault occurs during a power swing this function after fault inception time a power frequency component (50 or
must be unblocked and let the relay trip in a proper manner. 60 Hz) will get on three-phase active power. Based on this
The most popular method for detecting a power swing is to feature and using FFT analysis, symmetrical faults can be
measure the rate of change of impedance as it travels into the reliably and quickly detected in one power cycle, i.e. 0.02
second in a 50 Hz system. This method does not need high

978-1-4244-7398-4/10/$26.00 ©2010 IEEE 378 IPEC 2010

sampling frequency and can work with common sampling P3φ =P1 +P2 cos ω1 -ω2 t +P3 sin ω1 -ω2 t (4)
frequency used in numerical distance relay. This detector is
easy to set and immune to the slip frequency, fault location and
fault inception time. PSCAD/EMTDC® software is used to Where, P1, P2 and P3 are constant values.
simulate a power swing. Also FFT is performed by using On- Equation (4) clearly shows that during a power swing the
Line Frequency Scanner block included in the software. The three-phase active power measured by a distance relay
results prove that the presented detector is fast and reliable. oscillates with the angular velocity of ω1-ω2, the difference
between generator frequency and infinite bus bar frequency.
II. THREE-PHASE ACTIVE POWER CALCULATIONS AND However, we haven't yet taken into account the profile of
FAULT DETECTION METHODOLOGY generator frequency during power swings. To do this let us
The proposed method is based on three-phase active power inspect Fig. 1 showing a typical rotor speed of a generator
measurement. So in this section we will present three-phase during a stable power swing in pu. The rotor speed is result of
active power equations under power swing and fault simulating a SMIB system (not shown) by using MATLAB® in
conditions. To do this, we should do some mathematical which the generator experiences a stable power swing
analysis. condition.
A. Calculating Three-phase Active Power during a Power Looking at the profile shown in Fig. 1 and keeping in mind
Swing that this profile is an output of a linear time invariant system
The reason of power swing occurrence in power systems is lead us to the equation below:
the existence of at least two different frequencies which are
-t
close to each other. In fact we can say that a double-frequency ωrotor (pu) t =1+k·e τ · sin 2π·fslip ·t+α (5)
power system does experience a power swing as long as the
difference of these two frequencies remains in a specified
In (5) k and α are constants depending on the conditions at
range. Therefore, the three-phase active power equation during
the inception of power swing. Also τ is damping time constant
this phenomenon differs from normal conditions when there is
of speed profile. fslip is nothing more than rotor speed variation
only one frequency in power systems. To calculate three-phase
frequency called as slip frequency. Typically slip frequency
active power measured by a distance relay during a power
varies from 0.3 to 5 Hz. Generator frequency during a power
swing let's assume a single-machine-infinite-bus (SMIB)
swing is obtained by multiplying ωrotor in pu into ωbase as (6):
system having two different frequencies. This system merely
meets the double-frequency conditions. One frequency is due
-t
to a generator speed deviation from nominal speed. Another ω1 t = 1+k·e τ · sin 2π·fslip ·t+α ·ωbase (6)
forced by the remaining network or infinite bus bar is nominal
power system frequency. In this system the quantities
associated with the generator and infinite bus bar are On the other hand the infinite bus frequency shown by ω2 is
respectively indexed by 1 and 2. For example, ω1 is the angular equal to ωbase. So ω1 -ω2 can be written as follows:
velocity of the generator and ω2 is the one of infinite bus bar.
-t
Base on the double-frequency condition just discussed and ω1 -ω2 =k·e τ · sin 2π·fslip ·t+α ·ωbase (7)
superposition principle the current and voltage of phase A
measured by the relay during a power swing can be written as
follows [7]: By substituting (7) in (4) we will find the equation below
relating three-phase active power to slip frequency, fslip:
iA t =I1 · cos ω1 ·t+θi1 + I2 · cos ω2 ·t+θi2 (1)
uA t =U1 · cos ω1 ·t+θu1 + U2 · cos ω2 ·t+θu2 (2) P P + P cos k · ω ·t·e · sin 2π · f ·t+
α + P sin k · ω ·t·e · sin 2π · f ·t+α (8)
Where, I1 and I2 are the magnitude of current measured by
the relay injected by generator and infinite bus bar respectively.
For small enough value of x we have:
θi1 and θi2 are the phase angle of corresponding currents at the
power swing inception time. U1 and U2 are the voltage
magnitude at the relay location. θu1 and θu2 are the phase angle
of corresponding voltage at the power swing inception time.
Current and voltage for phase B and C are the same as (1) and
(2) with considering balance conditions during power swings.
Three-phase active power is given by:

P3φ =uA t ·iA t +uB t ·iB t +uC t ·iC t (3)

Substituting currents and voltages in (3) and using some

trigonometric equations leads us to the three-phase active Fig. 1 A typical rotor speed during a power swing
power equation during power swings:

379
 sin x~x (9) we have the following equation:

Vm
cos x ~1- (10) k1A =iA 0 - sin αA (15)
2
R + Lω 2

Therefore, if the argument in (8) is small as it is [8] then

three-phase active power will be written as below: For phase B and phase C currents we have similar
equations as follows:
P P +P ·t ·e · cos 2π · 2f · t + 2α +P · · R Vm
e−tτ·sin2π·fslip·t+α (11) iB t =k1B ·e-Lt + sin ωt+αB (16)
R2 +(Lω)2

Equation (11) shows that three-phase active power has R Vm

frequency components including fslip and 2fslip during a power iC t =k1C ·e-Lt + sin ωt+αC (17)
swing. Assuming maximum fslip as 5 Hz, high frequency R +(Lω)2
2

component of three-phase active power would be 10 Hz.

However, in most cases the amplitude of frequency component Where,
of 2fslip value is negligible and we can say that three-phase
active power oscillates with fslip. This feature is very important αB =αA -
2π
(18)
to distinguish a fault from power swings. 3

B. Calculating Three-phase Active Power after Symmetrical αC =αA +

2π
(19)
Fault Inception 3

Here we investigate the behavior of three-phase active Vm

power when a symmetrical fault occurs. To do this, at the first k1B =iB 0 - sin αB (20)
2
R + Lω 2
we should extract phase current equations when a three-phase
fault occurs. Fig. 2 shows the model of phase A in a power
Vm
system when a three-phase fault occurs. Based on the model, k1C =iC 0 - sin αC (21)
after a three-phase fault occurrence we have (12) for phase A R2 + Lω 2
current:
We have taken into account these constants since the three-
diA
RiA +L =Vm cos ωt+θA (12) phase active power profile after fault inception is definitely
dt dependant to them. Using (3) and (13)-(21) and after some
trigonometric calculations we will have three-phase active
Where, R and L are respectively resistance and reactance
power measured by relay after fault inception time as follows:
from relay location to fault point. Also ω is equal to 2πf, (f is
power frequency, 50 or 60 Hz). Solving the above equation we R
will have (13) for phase A current: P P+V e L k A · cos ωt + θA + k B · cos ωt +
R θA − +k C · cos ωt + θA + (22)
Vm
iA t =k1A ·e-Lt + sin ωt+αA (13)
R2 +(Lω)2
Where, P is a constant value.
In (13) αA is obtained from equation below: Equation (22) clearly shows that after a symmetrical fault
inception time three-phase active power has damping
R cos θA +ωL sin θA oscillations with nominal frequency, i.e. 50 Hz in a 50 Hz
tan αA = (14) power system.
ωL cos θA -R sin θA

Also k1A is a constant value obtained from initial condition, C. Proposed Symmetrical Fault Detection Methodology
i.e. phase A current at the fault inception time. To obtain k1A Now we are in the position where we can propose our fault
detection methodology. As it has just been explained, during a
power swing three-phase active power oscillates with slip
frequency, i.e. fslip value whereas after fault inception time it
will oscillate with nominal frequency (50 or 60 Hz). On the
other hand slip frequency is much smaller than power
frequency. So if the relay performs FFT analysis with
fundamental frequency of 50 Hz then the Fourier coefficient
will be considerable after fault inception time whereas there is
negligible Fourier coefficient during a power swing. So if the
Fourier coefficient is considerable then the relay detects a
three-phase fault occurrence and it will trip in a proper manner.
Fig. 2 Phase A model after symmetrical fault inception

380
 respectively shown in Fig. 6, Fig. 7 and Fig. 8. For the sake of
clarity we zoom on three-phase active power profile at the fault
inception time (case 6) as shown in Fig. 9. As can be seen
three-phase active power will oscillate with frequency of 50 Hz
following symmetrical fault inception. In section II we
mentioned that FFT is performed every cycle. Based on this,
Fig. 3 Simulated system
Fourier coefficients under time interval [4.5, 5.1] for cases 3, 6
and 9 are respectively shown in Fig. 10, Fig. 11 and Fig. 12. As
In the proposed method, fault detection delay is 0.02 shown the coefficients are of considerable value at t=5.02
second (for a 50 Hz power system) as the required date window second, i.e. 0.02 seconds following the faults whereas they are
width is 0.02 second and FFT is performed every cycle. This negligible during the power swing. Table II presents the
time delay for relay unblocking is perfectly acceptable in HV Fourier coefficients value after 0.02 second from fault
and EHV power systems and in fact the method is very fast. inception time for all cases. Minimum value for the coefficients
is 28.3 whereas the maximum value during power swing is
III. SIMULATION RESULTS equal to 5. This considerable difference between FFT
A. System Description and Power Swing Simulation
The simulated system is shown in Fig. 3. In this system, a
generator is connected to grid (infinite bus) by a 400 kV
double-circuit line. Line CB3-CB4 has 50 km length. Length of
other lines and the rated values of generator and transformer
have been depicted in Fig. 3. A power swing condition is
simulated by using the model of system in PSCAD/EMTDC®.
The sampling frequency is 1 kHz which is common to
implement in digital relays. In the predicted scenario causing
power swing a single-phase fault (Fault1) occurs at 0.5 second
on bus bar A. Because there is no bus bar protection the fault is
cleared after 0.4 second by the operation of zone-2 of relays R3
and R6 (not shown) which send a trip signal to breakers CB3
and CB6 respectively. This sends the system into a power
swing. Under the power swing condition the three-phase active
power variations measured by relay R1 is studied.
Rotor speed variation during power swing conditions is
shown in Fig. 4. The current, voltage and three-phase active
power waveforms seen by the relay are shown in Fig. 5.
Looking at Fig. 4 and Fig. 5(c) we recognize that the frequency
of three-phase active power is equal to the rotor speed variation
frequency or slip frequency. This is the result we have proof
mathematically in section II.
B. Fault Simulation and FFT Coefficients
Now we should consider symmetrical fault effect on the
three-phase active power by using simulation. To do this we
simulate a three-phase fault on the line CB1-CB2 when it is
experiencing the power swing. To evaluate the detection
methodology, different cases including different fault inception Fig. 5 Current, voltage and three-phase active power during the power
time and different fault location are considered according to swing
cases in table I. For case 4, 5 and 6 three-phase active power
variations during the power swing and after fault inception is TABLE I. CASES SIMULATED TO TEST DETECTION METHODOLOGY
Fault inception Fault distance
Case
time (s) (km)
1 4.5
2 4.75 5
3 5
4 4.5
5 4.75 125
6 5
7 4.5
8 4.75 245
Fig. 4 Rotor speed during the power swing 9 5

381
coefficients values during power swing and those after
symmetrical faults shows that the proposed method can work
for any symmetrical fault. There is no matter when and where
symmetrical faults occur. This detector is easy to set since we
must only set the threshold of FFT coefficient by which
symmetrical fault can be detectable.
IV. CONCLUSION
This paper introduces some interesting features of three-
phase active power during a power swing and after a Fig. 9 Three-phase active power oscillation following a symmetrical fault
symmetrical fault inception. It is demonstrated that during (case 6)
power swings the oscillation frequency of three-phase active
power is equal to slip frequency whereas after symmetrical
fault inception the power oscillates with power frequency (50
or 60 Hz). The proposed method makes use of these frequency
components of three-phase active power to detect symmetrical
faults. The advantages of this approach over other reported
methods are several. To detect symmetrical faults this method
does not need high sampling frequency and works with
common sampling frequency used in distance relays. This
method does not depend on any calculations that use Fig. 10 FFT coefficients during power swing and after symmetrical fault
approximations of the system conditions (such as equal source occurring at t=5 and distance 5 km (case 3)
voltages at SE and RE). Since FFT is performed only one time
in every cycle, this method does not impose any significant
calculating burden but at the same time can detect a
symmetrical fault very quickly.

Fig. 11 FFT coefficients during power swing and after symmetrical fault
occurring at t=5 and distance 125 km (case 6)

Fig. 6 Three-phase active power during power swing and after a

symmetrical fault occurring at t=4.5 and distance 125 km (case 4)

Fig. 12 FFT coefficients during power swing and after symmetrical fault
occurring at t=5 and distance 245 km (case 9)

TABLE II. FFT COEFFICIENTS FOR ALL SIMULATED

Fig. 7 Three-phase active power during power swing and after a CASES 0.02 SECOND AFTER FAULT INCEPTION
symmetrical fault occurring at t=4.75 and distance 125 km (case 5)
Case Fourier Coeff. (MW)
1 31.96
2 28.3
3 30.64
4 67.93
5 67.44
6 71.67
7 77.3
8 73.25
Fig. 8 Three-phase active power during power swing and after a 9 74.8
symmetrical fault occurring at t=5 and distance 125 km (case 6)

382
 ACKNOWLEDGMENT [4] L. Xiangning, G. Yan, L. Pei, “A novel scheme to identify symmetrical
faults occurring during power swings,” IEEE Trans. Power Delivery,
The author would like to extend his gratitude to Dr. Davood vol. 23, no. 1, pp. 73-78, Jan. 2008.
Farokhzad, Director of Deputy of Supervision on Transmission [5] L. Xiangning, L. Pei, Ch. Shijie, “Effective transmission line fault
Grid Security, IGMC, for his kindly encouragement to pursue detection during power swing with wavelet transform,” in Proc. IEEE
this research. Power Eng. Soc. General Meeting, 2000, pp. 1950–1955.
[6] S.M. Brahma, “Distance relay with Out-Of-Step blocking function using
REFERENCES wavelet transform,” IEEE Trans. Power Delivery, vol. 22, no. 3, pp.
1360-1366, Jul. 2007.
[1] J. L. Blackburn, Protective relaying principles and applications, 2nd ed.,
New York: Marcel Dekker, 1998. [7] C. Wei, P. Zhencun, Y. Zhihao, Z. Rong, “The method of detecting
asynchronous power swing based on the variation of active power,”
[2] ABB, Technical reference manual REL 670, 2007, [online], Available: International Conference on Power System Management and Control,
www.abb.com. 2002, pp. 297-302.
[3] B. Su, et al., “Fast detector of symmetrical fault during power swing for [8] B. Mahamedi, “Power swing detection in power systems using wavelet
distance relay,” in Proc. IEEE Power Eng. Soc. General Meeting, 2005, transform,” M.S. thesis, Dept. Electrical. Eng., Shahid Beheshti Univ.,
pp. 1836–1841. Tehran, Iran, 2010.

383