Professional Documents
Culture Documents
. (4)
In the other case, if the impedance falls out of the circle from
Fig. 1, the phase difference between V
1
and V
2
is:
|| <90
. (5)
Eqs. (2) and (3) can be written in the time-domain as:
v
1
(k) = u
pol
(k-delay), (6)
v
2
(k) = u
r
(k-delay) +d i
r
(k), (7)
where k represents kth sample at time t, delay is an integer that
changes the angle of the MHO characteristic . That indicates that
in the time-domain approach based on the phase comparator with
border angles 90
m
, (8)
where m represents number of samples per fundamental cycle T
(Table 2 is given for m=32). Parameter delay also represents a
time shift T
= delay
T
m
= T
360
. (9)
The time shift T
.
2.1. Time-domain phase comparator
In this paper we applied a phase comparator based on the aver-
agepower, presentedin[29]. Thephasecomparator uses anaverage
value of the product of input signals p(k) =v
1
(k) v
2
(k) for determin-
ing their phase difference. It is well-known that the product of two
fundamental harmonics has two components: a component with
the double frequency (the reactive power component), and a con-
stant component that depends on the phase difference between
signals (the active power component). Negative values of the con-
stant (active power) component indicate that the phase difference
|| >90
t
tTp
v
1
() v
2
() d. (10)
For the practical implementation of Eq. (10), two data windows of
length m are needed. These data windows contain discrete values
v
1
(k) and v
2
(k) calculated from the raw samples i
r
(k), u
pol
(k) and
u
r
(k) through Eqs. (6) and (7). The average power of v
1
(k) v
2
(k),
within one cycle T
p
, can be determined by the sum:
TDHC =
2
m
m/2
k=1
v
1
(k) v
2
(k). (11)
Eq. (11) represents the time-domain half-cycle algorithm. The sec-
ond version of the time-domain algorithmis based on the full-cycle
sumwithin one cycle T:
TDFC =
1
m
m
k=1
v
1
(k) v
2
(k). (12)
In the time-domain approach the MHO characteristic is realized
making use of Eqs. (6), (7) and (11), or alternatively (6), (7) and (12).
The same operating characteristic in the phasor-domain approach
requires computing the phasors U
r
, U
pol
and I
r
, then calculating V
1
and V
2
through Eqs. (2) and (3), and nally their phase difference
Fig. 4. Block diagramof the time-domain and the phasor-domain algorithms.
=arg(V
1
) arg(V
2
). Computing each of the phasors requires the
same computational burdenas the time-domainphase comparison
in Eq. (12). This indicates that the time-domain approach allows us
to realize a distance relay algorithmwith a smaller computational
burden than in the phasor-domain approach.
3. Block diagram
The proposed algorithms are based on the time-domain
approach utilizing Eqs. (6), (7), (11) and (12). Characteristics of the
proposed algorithms are compared to characteristics of a phasor-
domain algorithm, based on the well-known full cycle discrete
Fourier transformation and Eqs. (2) and (3) in the phasor-domain.
The block diagram of the complete system for testing the algo-
rithms is shown in Fig. 4.
Input signals from the protected line are obtained from a sim-
ulator that calculates electromagnetic transient behavior of the
power system. In our testing we used a sampling rate f
s
=1600Hz,
or m=32 samples per fundamental cycle. Fault type detection is
implemented according to Table 1.
The anti-aliasing lter (F1) is an analog low-pass lter that is
used to minimize aliasing effects as well as to attenuate the high
frequency components. We used simulation of an analog second
order Butterworth low-pass lter with cutoff frequency of 93.6Hz,
stop-band cutoff frequency of 380Hz and stop-band attenuation
28dB (F1 in Fig. 5).
Inorder toreduce decaying DCoffset ini
r
andimpact of the tran-
sient response of capacitive coupled voltage transformers (CCVTs)
[31], we propose using Characteristic Harmonic Digital Filter
Fig. 5. Frequency responses of the anti-aliasing lter (F1) and CharmDF (F2).
S.J. Zubic, M.B. Djuric / Electric Power Systems Research 92 (2012) 2028 23
Table 3
The relationship between the sampling rates mand the phase differences .
m 16 20 32 64
78.75
81.00
84.37
87.19
,
R
f
=0).
LOAD =5
=10
=15
=20
,
R
f
=0).
LOAD =5
=10
=15
=20
), fault resistance
R
f
=0and SIR=0.3. We can notice that for some fault resistances
it is possible overreach, but for increasing and R
f
reach is smaller.
According to Table 4 we chose the cross-polarized MHO character-
istic with =45
, 45
and 0
and R
f
=0.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.08 0.38 0.35
30% 0.41 0.53 0.53
50% 0.42 0.58 0.58
70% 0.52 0.98 0.98
80% 0.6 1.07 1.06
Fig. 20. Impact of the CCVT transient on the algorithms output when CharmDF is
applied.
Fig. 21. Operating time of the compared algorithms for different source impedance
ratios (SIR).
S.J. Zubic, M.B. Djuric / Electric Power Systems Research 92 (2012) 2028 27
Table 8
Operating time in cycles for an inception angle of 45
and R
f
=0.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.1 0.16 0.16
30% 0.15 0.56 0.55
50% 0.25 0.6 0.6
70% 0.55 0.74 0.74
80% 0.6 1.05 1
Table 9
Operating time in cycles for an inception angle of 0
and R
f
=0.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.21 0.28 0.26
30% 0.28 0.62 0.6
50% 0.35 0.74 0.72
70% 0.63 0.82 0.82
80% 0.72 1.15 1.1
Table 10
Operating time in cycles for an inception angle of 90
and R
f
=20.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.1 0.39 0.38
30% 0.4 0.52 0.52
50% 0.42 0.78 0.7
70% 0.51 0.96 0.96
80% 0.92 1.42 1.41
Table 11
Operating time in cycles for an inception angle of 45
and R
f
=20.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.12 0.27 0.26
30% 0.22 0.55 0.55
50% 0.48 0.65 0.64
70% 0.6 1 1
80% 0.7 1.15 1.2
Table 12
Operating time in cycles for an inception angle of 0
and R
f
=20.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.23 0.37 0.37
30% 0.3 0.65 0.65
50% 0.57 0.76 0.76
70% 0.68 1.1 1.08
80% 0.82 1.3 1.3
Table 13
Operating time in cycles for an inception angle of 90
and R
f
=40.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.28 0.42 0.41
30% 0.41 0.75 0.68
50% 0.45 0.9 0.9
70% nop nop nop
80% nop nop nop
Table 14
Operating time in cycles for an inception angle of 45
and R
f
=40.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.15 0.48 0.47
30% 0.45 0.62 0.62
50% 0.52 0.95 0.91
70% nop nop nop
80% nop nop nop
Table 15
Operating time in cycles for an inception angle of 0
and R
f
=40.
Fault location Time domain algorithm Phasor domain
PDFC
TDHC TDFC
10% 0.28 0.61 0.6
30% 0.52 0.72 0.72
50% 0.66 1.06 1.04
70% nop nop nop
80% nop nop nop
fundamental cycles, while nop is used for cases when an algorithm
did not operate. Tables 715 show that the nop cases occurred for
the biggest R
f
according to Table 4 for =45