Professional Documents
Culture Documents
06319090
06319090
I. I NTRODUCTION
c 2012 IEEE
978-1-4673-2673-5/12/$31.00
(2)
(12)
x() + x( /2)
.
x( 1) + x( /2 1)
(13)
what:
=
=0
)
(
1
2
2
Y1 () =
,
x() sen
,
1
=0
A1 () = Y1 ()2 + Y1 ()2 .
(3)
1 =
x(2 1),
(4)
=1
2 =
x(2),
(5)
=1
we have:
= Y1 (),
(6)
()
2
coth
( 2 1 ), (7)
Y1 () = Y1 () +
()
(14)
1
,
2
(1 + ) cos
1
( 2 )]2
[
( )]2 ,
+ sen 2
(16)
()
(17)
Y1
()
(18)
1
2
x() cos
=0
)
,
)
(
)
1 (
2
2
Y1 () =
cos
.
x +
(8)
(9)
=0
=0
( )
Y1 ()
Y1 ( 1) cos 2
( 2)
Y1 () =
.
sen
(10)
1
=0
(11)
(
cos
(19)
A. Indice 1
x() =
[1 y()]2 ,
(25)
=1
(20)
=0
1
cos( 120).
B. Simulated Signals
The simulations used the simplified electrical system proposed in [7] and illustrated in Fig. 1. Data elements with
subscripts 0 and 1 correspond to the values of magnitudes of positive and zero sequence, respectively. Model and
parameters provided in [12] were used to current transformer
(CT).
B. Indice 2
Barra 1
Barra 2
TC
S1
Rel
S2
(21)
According to its definition, 2 is the percentage of maximum overshoot present in the process of estimating the
magnitude of the phasor.
Fonte S1
Linha de transmisso
Fonte S2
V = 1,020 o
l = 180 km
V = 0,9820 o
Z 0 = 1,014 + j18,754 W
Z 0 = 1,127 + j 20,838 W
Z1 = 0,871 + j 25,661 W
Yl .0 = j 2,293 S/km
Z1 = 0,968 + j 28,513 W
C. Indice 3
Figure 1.
(22)
=1
( )
= 1, 2 e 3.
(23)
x() = cos(260) .
(24)
V. P ERFORMANCE E VALUATION
The influence of the sampling frequency in the filters
performances was evaluated by varying the sampling rate
16, 32 and 128 samples per cycle. The signals described
in Section IV-B were simulated with a step of calculating
corresponding to 160 samples per cycle and pre-processed by a
low pass Butterworth filter of third order with cutoff frequency
at 180 Hz, as described em [7].
A. Signal with One DC Offset
In general, the filters showed reduced overshoot and increase
the oscillations number in its steady state response when the
time constant of the dc ofset was increased and maintained at
the sampling frequency, Fig. 2.
Filters with small overshoots are desirable, however, oscillations in their steady state responses should be viewed with
caution, because they reduce its accuracy and convergence
speed.
According to the results shown in Tab. I, the sampling
frequency influences differently the filters. The FCF filter
showed the worst performance when the sampling frequency
was increased, given that its indices 1 and 2 were
increased and 3 didnt show convergence during the six
cycles, randomly chosen for analysis.The filters: MF, Cos
and CosM reduced the overshoot and increased the speed of
convergence, despite intensified fluctuations in their response.
The combination between FCF filter and AMF filter had
the best performance when compared to others filters in the
same sampling frequency as well as being the one to reduce
fluctuations in their response when the sampling frequency
was increased.
4
Filtro Fourier de Um Ciclo
1.2
with convergence speed stable during the variations in sampling rate. Regarding the modified versions, the MF filter
presented the worst rates of performance without converge
during the simulation when was adopted a sample rate of
16 samples per cycle. The analysis results are summarized
in Tab. II.
0.8
0.6
0.4
0.8
Amplitude
Amplitude
0.6
Componente
DC 0,5
de 0,5ciclos
ciclos
Componente
DC de
Componente DC de 5 ciclos
Componente DC de 5 ciclos
0.4
0.2
0
0
1.2
0.2
1
0
0
Tempo (Ciclos)
Filtro Cosseno
1.2
Tempo (Ciclos)
1.2
1.2
2.5
0.4
0.6
0
0
0
0
Tempo (Ciclos)
0.6
0
0
0.4
1.4
0.6
0.2
0
Tempo (Ciclos)
Filtro Fourier de Um Ciclo + FMA
1.2
6
Amplitude
1
0.8
0.6
Tempo (Ciclos)
FCF+AMF
0.8
Tempo (Ciclos)
Table I
CosM
0.4
0.2
0.6
0.4
0
Cos
1.2
Figure 2. Amplitude of the phasor estimated for the testing signal with one
dc offset and sampling rate of 16 samples per cycle.
MF
Tempo (Ciclos)
0.8
Tempo (Ciclos)
FCF
0
0
0.2
Filter
0.4
0.8
0
0
1.2
Amplitude
Filtro Cosseno
1.4
Tempo (Ciclos)
Amplitude
1.2
Tempo (Ciclos)
0
0
1
0.5
0.2
0.2
1
0.6
1.5
0.4
0.4
0.2
Amplitude
0.8
0.8
Amplitude
0.6
Amplitude
Amplitude
0.8
Amplitude
Samples
per
cycle
ID1
ID2 (%)
ID3 (cycle)
16
0,1062
10,4761
Inf.
32
0,2119
10,5739
Inf.
128
0,8437
10,6121
Inf.
16
0,0070
2,7606
Inf.
32
0,0067
1,9105
3,3281
128
0,0129
1,2786
1,3919
16
0,0167
5,2935
Inf.
32
0,0199
4,5422
Inf.
128
0,0468
4,1417
1,4792
16
0,0072
2,7655
Inf.
32
0,0067
1,9118
3,3438
128
0,0127
1,2788
1,3724
16
1, 5689 105
0,1034
1,1354
32
8, 7613 106
0,0547
128
1, 9173 106
0,0144
Figure 3. Amplitude of the phasor estimated for the testing signal with
harmonic and sampling rate of 16 samples per cycle.
Performance indices
Table II
M EAN VALUES OF PERFORMANCE INDICES FOR THE TESTING SIGNAL
WITH HARMONIC .
Filter
FCF
MF
Cos
CosM
FCF+AMF
Samples
per
cycle
ID1
ID2 (%)
ID3 (cycle)
16
0,9375
32
1.0938
128
1.2109
16
0,0070
2,7606
Inf.
32
0,0067
1,9105
3,3281
128
0,0129
1,2786
1,3919
16
0,0167
5,2935
Inf.
32
0,0199
4,5422
Inf.
128
0,0468
4,1417
1,4792
16
0,0072
2,7655
Inf.
32
0,0067
1,9118
3,3438
128
0,0127
1,2788
1,3724
16
1, 5689 105
0,1034
1,1354
32
8, 7613 106
0,0547
128
106
0,0144
Performance indices
1, 9173
C. Simulated signals
The combination between FCF filter and AMF filter showed
the best performance, maintaining its immunity to the effect
of the dc offset in signals closer to those obtained in a real
fault situation, in which multiple exponentially decaying dc
offset may be present, Fig. 4.
Amplitude (kA)
3
2,9
R EFERENCES
2,8
2,7
2,6
2,5
TDFCC
CosM
TDFCCM
TDFCC +FMA
Cos
7
10
11
12
13
14
Tempo (Ciclos)
Table III
M EAN VALUES OF PERFORMANCE INDICES FOR THE TESTING SIMULATED
SIGNAL .
Filter
Samples per
cycle
16
FCF
MF
Cos
CosM
FCF+FMA
Performance indices
ID1
ID2 (%)
ID3 (cycle)
2,9733
10,2164
2,4375
32
5,9531
9,8972
2,0313
128
23,4125
10,2864
1,2344
16
0,2310
2,8270
1,2500
1,3125
32
2,1349
2,5735
128
0,5055
1,4694
1,0547
16
0,4938
4,3026
1,8125
1,3750
32
1,4286
2,9246
128
1,1029
2,3309
0,9453
16
0,2390
2,8711
1,3125
32
2,0832
2,5749
1,3438
128
0,5056
1,4696
1,0547
16
0,0067
0,1593
1,1875
32
1,7364
1,3330
1,1563
128
0,0525
0,0869
0,8594
VI. C ONCLUSION
This paper presented a comparison between filtering algorithms used for estimating phasors in digital relays. We
compared algorithms already in use, as well as those recently
reported in the literature, such as the AMF.
The performance indices allowed to compare the algorithms
on different sampling rates with respect to fluctuations in
the steady state response, overshoot and convergence speed.
How were defined test signals, it allowed us to observe the
performance of the algorithms in different scenarios: only