3.1 INTRODUCTION:
techniques.
38
RELAYS:
and field test of these algorithms have been carried out mostly
here.
of the waves. In 1972, the first digital computer relay used for
distrubances and it did not give any false trip during the six
years of operation in the field [188]. This project proved the
RELAYS:
relaying.
and 3.4 show the system block diagram and the fault processing
a 275 KV, 150 KM transmission line for ten months and the test
block diagram of this system. This system was tested in the field
digital signals between the protected element and the relay. The
PROTECTIVE RELAYING EQUIPMENT
SYNCHRONOUS SIGNAL I
I
I
CURRENT (LOCALi)
z>
OC
o
cc
z
ui
CC
2
Ul
o
»-
Ul
id 0
O
>
_l
<
- Ul
o
(5 <
»- u MICRO-
Uao) aamwsNVdi
WAVE
CHANNELS
QNV SH3AI3D3H IVOlidO
om
-3 2
Ul
lNawdinoa noiivdinowwoo
,-•5-
w
S“z
ui5 o
o
<<>
°-j£
=0
MICRO-COMPUTER
°J<
— Ul o
oc 2 o
Ul° Ul
LINE
I/O INTERFACE
t T~1
4 i
I
L
_ 4.
FIGURE 3-5 BLOCK DIAGRAM OF THE PROTECTIVE RELAYING EQUIPMENT
4-6
w-
7Vr -500 KV
DIGITAL CONVERTER AND TRANSMITTING SIGNAL R EC EIVING AND PROTECTIVE
EQUIPMENT (OUTDOOR) DEVICES (IN RELAY ROOM)
H /S
A /D ENC
OPTICAL
RECEIVER Dl GITAL
DISTANCE!
RELAY
SYSTEM
INFORMATION DIGITAL
OVER
CURREN
RELAY
J1 o p t c a l
'
=jjT R A N S M I-
CLOCK
(7 2 0 H z ) CLOCK
GENERATOR
TO CONVENTIONAL
EQUIPMENT IN USE
performed the protection functions. Due to the high cost and huge
relay point and the fault location. Many algorithms which differ
in both performance and complexity have been proposed [2, 3], but
The cost of computers has become cheaper over the years and
PROTECTION:
in fig. 3.7. The analog relaying signals from power system are
as follows:
evaluation.
The facades are discussed in the following subsections.
samples.
51
reactance reaches near zero from the positive side, but if the
d.c. shift. The algorithms which provide fast trip decision are
discriminants.
a system abnormality.
ALGORITHMS:
188, 11, 197]. The published algorithms can be grouped into two
model. The significant algorithms that fall under the first group
*■
54
204 ] .
101].
84, 190].
voltage and current signals from the complex post fault wave
algorithms are
phase angle a.
v = V Sin ut (3.1)
pk
whose derivative V'is given by
v *= wV Cos wt (3.2)
pk
56
I^ =i2+(i'/w)^ (3.6)
pk
0 = a + tan"'*' ( wi/i1) (3.7)
i
Z = (V2 / I2 )1/2 (3.8)
pk pk
2 , 2 2, 2
I = (i ' / w ) + (i"/co ) (3.12)
pk
and wt = - arc tan ( toi'/i") (3.13)
L
2 2 2 ^ 2 2 2
Thus| %\ = [(V ' / to ) + (V "/ u> ) ] / [(i '/ w)+(i "/ w) ] (3.14)
k k k k
4> = arc tan ( ooi '/i ") - arc tan ( ooV '/V ") (3.15)
L k k k k
Where V ' = (v - V ) /2h (3.16)
k k+1 k -1
L
and V " = (V - 2V + V ) / h (3.17)
k k+1 k k-1
Where h is the sampling interval.
V = V Sin ( wt - $ + 6 ) (3.18)
pk
i = I Sin ( wt + 5 ) (3.19)
pk
at times t , t -A and t
k k k
R can be given as,
/ 4 i r> jf
60
i - V i ) / (i2
X = (V 1 i ) Sin A (3.20)
L k-1 k k k-1 k-1 k-2 k
and R = (2V i V V l ) /
k-1 k-1 k-2 k k k-2
i ) (3.21)
k-2 k
from this approach are close to the true post fault steady state
sections .
N-l
V
s
= (1/N) [ 2
L=1
l V
k-N + L
Sin (2tt/N)L] (3.22)
N-l
V
c
= (1/N) [V
k-N
+ V + 2
k L=1
l V
k-N+L
Cos (2tt/N)L] (3.23)
2 2 2 2
Z = Square root of (V + V ) / (I + I ) (3.24)
sc sc
and sin ( o> Lh) for values of L ranging from 1 to N are stored in
locations, all the previous samples move one location ahead, and
and (3.23). Then the impedance as seen from the relay location
with a line. When X/R ratio is high, the results showed that the
value. The need of the multiplier chip has been avoided by Carr
and Jackson [201] who applied the Fourier transform approach
of sampling rate. This low-pass filter has 1/4 cycle delay time.
i) The use of the R-C low-pass filter will give quarter cycle
delay.
fault.
transient components.
in the line. The estimated pliasor voltages and currents were then
vector form as
X(t)=X+jx (3.26)
d q
where the components X and X can be evaluated in discrete form
d q
from the following equations
N-l
X = (1/N) £ x (Lh) Sin ( wLh) (3.27)
d L=1
67
N-l
and X = (I/N) [ x (Lh) . Cos ( toLh) (3.28)
q L=1
using equations (3.27) and (3.28) for each input signal, the
V = V + j V (3.29)
d q
and I = I + j I (3.30)
d q
equations.
R = [ (V I + V I )] / (I 4* i ) (3.31)
d d q q d q
2 2
and X = [(VI- V i )] / (I + I ) (3.32)
q d d q d q
filters were the voltages and currents from the three phase
The basis for these approaches are the same as for the full-
cycle Fourier scheme, with an additional advantage of shortened
data window.
Cos C co t - B ) (3.33)
P
The first, second and third terms correspond to the steady state,
a periodic (ie. dc) components decaying with time constant ta
V = V + j V (3.35)
1 d q
and I = I + j I (3.36)
1 d q
T /2
/• w
I = (P/T ) I ( t) Sin w id t (3.41)
q w 2
-T /2
w
Where coefficients K and P are,
(1- r ) -n(I-r.2 ) wi
K ; P = .......... ..... and r =-----
r Sin u r Sin it r to2
(3.42)
V V + j V
Id q
I I + j I
1 d q
* V I + V I V I - V I
d d q q q d d q
2 2 2 2
I + I I + I
d q d q
Since Z = R + j X
Therefore R = (V I + V I ) / (I + I ) (3.44)
d d q q d q
2 2
X = (V I -VI) / (I + i ) (3.45)
q d d q d q
Compared with the standard practice of extracting the
oscillatory components.
72
the data window is shortened to half cycle plus one sample for
written as,
N/2
and
N/2
V
q
= (4/N) l V
[k-(N/2)+i]
cos [(2 tt / N) i ] (3.47)
i=l
in equations below,
V = V + j V (3.48)
d q
but with time constant determined by the known X/R ratio of the
line.
on line by using an IBM 5/7 mini computer, and its accuracy was
.for all the signals. Here 12 samples per cycle was selected to
equation,
and iu = di / dt
provide the new set of voltage and current samples for evaluating
phase voltages and currents using their sampled values. The most
be summarized as follows:
N-1
E0 = (1/3) l W k (e ak + e bk + e ck ) (3.51)
k=0
N -1
E
1
= (1/3) 1 (We
k ak
+ W e
k-4 bk
+ W e)
k+4 ck
(3.52)
k=0
and
N-l
E = (1/3) £ (W e + W e +W e) (3.53)
2 k ak k+4 bk k-4 ck
k=0
- j ( 2k tt /n)
where W = [(2j)/nj2]e
k
ii) All the current sequence components are pre-mult ipl ied
&E = AI Z (3.54)
0 0 0
77
A E * H l (3.55)
1 1 1
AE AL Z (3.56)
2 2 1
and AE = Z I (3.57)
1 lpre-fault
where
[I ] - [ I ]
0,1,2 0,1,2
Fault pre-fault
K E / AE (3.58)
0 0 0
K E / AE (3.59)
1 1 1
K E / AE (3.60)
2 2 2
K AE / AE (3.61)
L
K ■ ’ AE / AE e-1(W (3.62)
0 0 1
78
and K ' = 1, if f AE [ ~ [ AE (
2 2 1
= 0 otherwise
K + K K ' + K K '
1 2 2 0 0
D = ......................... (3.63)
1 + K ' + K ' + K ’
0 2 L
this algorithm does not offer any speed advantage over any other
equation
V = Ri + L (di/dt) (3.64)
as,
(V +V )(i -i
)(i -i )
) - (V +V
k-1 k-2
k k-1 k-1 k-2 k k-1
R = ---------------------------------------- (3.65)
(i +i)(i -i )-(i +i )(i-i )
k-1 k k-1 k-2 k-1 k-2 k k-1
and
(V +V )(i -i ) - (V +V )(i +i )
k-1 k-2 k-1 k-2 k-1 k k-1 k-2
L = (h/2) -------- ----------------------------------- (3.66)
(i +i)(i -i ) - (i +i Hi-i )
k-1 k k-1 k-2 k-1 k-2 k k-1
V=Ri+L(di/dt) (3.64)
and
0 ( tt/M) (tt/N)
( tt/M) + ( tt/N)
the d.c. off- set is retained and the errors due to low order
is completely filled with the fault data [4, 188]. The frequency
of a lumped line model and its first derivative are proposed for
V = iR + Li' (3.69)
2
L = (Vi'-v'i) / (i' -i"i) (3.71)
and
2
R = (V'i' - vi") / (i' -i" .i ) (3.72)
The authors have applied the least square criterion curve fitting
section for the protection of the line from the relay location to
components out and then calculating the line impedance from the
filtered signals. The procedure used in deriving this algorithm
(3.80) as below
2 2
V = R i + L (di/dt) - R C (dv/dt) - LC (d V/dt ) (3.80)
equations of the post fault waveforms to obtain the real and the
imaginary components of the fundamental frequency of the current
+ K Sin ( co t + 0 ) (3.81)
n n 1 n
- t/ T
Representing e by the first three terms of its Taylor series
+ax+ax (3.82)
16 6 17 7
a = t
17 1
x = k ; x = k 'Cos 0 ;x = k Sin 0 ;
1*021 131 1
x = k Cos 0 ; x = k Sin 0 ; x = -k /x
4 n n5n n60
x = k / (2 x2)
7 0
in matrix form as
-1
[X] = [A] [V] (3.84)
7-1 7 xm m-1
Where m 72- 7
and
taking the ratio of the power and the square of the current.
eliminate the harmonics and the noise which appear with the
written as
89
T/2
y ( T ) = (1/T) f i (t) . V(t + t ) dt (3.89)
xy 1 J 1
-T/2
2 2
¥ ( t ) = (1/2) (V 1/1). Cos ( 4> + wt ) (3.90)
xy 1 pk L L
Where
t = Time delay
1
The self-correlation of the current wave form can be given as
T/2
¥ ( t ) = (1/T) f i(t).i(t+ t )dt (3.91)
XX 1 J 1
-T/2
¥ = R X tan ux (3.94)
T 2
2 L
90
Y2 Y2
X = -- - - - - - - - - - - - - - - - - (3.96)
L
tan got
1
- tan got
2
correlated with the fault wave form, however are not the
fundamental sine and cosine waves, but odd and even square waves.
91
This algorithm has been developed and tested by Gabr Md. Abd
el-Salam [204]. This algorithm has been used as the main core of
explained. An approach has been used for rejecting the D.C. off
set components from the incoming raw data samples before they
pass through the Walsh function filters. The main effect in this
current and the voltage signals from their sampled data. The real
This algorithm has been tested to evaluate the real and the
voltage and current from their data samples after rejecting the
d.c. off-set components from the current samples. Using this, the
From the test results obtained, the author has reported the
mathematical operations.
ii) It needs less memory location and operation time than full
algorithm.
v) It estimates the transmission line parameter after a half
Haar algorithms with full cycle data window, for extracting the
full cycle window Haar algorithm and in about half a cycle in the
scheme.
i
lines. For such long lines the effect of shunt capacitance must
distributed parameter line model and hence suitable for long EHV
transmission lines.
cycle.
infeed.
99
the basic algorithm requires only two samples. In the same year
R.Ramaswamy et.al. [233] proposed algorithms employing sparse
This paper presented the theory of the technique and the results
checking capability.
system. The fault clearing time depends on the speed of. both the
the travelling waves at the local and remote ends and then uses
first and the most prominent high frequency fault component along
derived from the sampled data available at the local end only. If
the voltage and current changes produced by the fault point are
A V and A i , then,
F F
A V (t)-Z Ai(t)=AV(t+x)-Z Ai (t+ x) (3.97)
FcF F F c A F
For a Line-to-Line fault occuring at time t = 0,
derived as shown
2
D = [ AV (t + x ) - Z Ai (t + x ) ]
F A F c A F
+ 1/ a) ^[d/dt AV (t + x ) - Z d/dt Ai (t+ x )]
A F c A F
2
= V ' (3.101)
for x iz. t ^ 3 x _
F F
The results are also erratic for close-up faults where rapid
-^o
d = AV - Z A i (3.103)
R R
for the forward wave characteristics, and
d = AV + Z A i (3.104)
F R
for the backward wave characteristics, where Z is line
detection.
Hence,
b(t)=r.a(t-2r) (3.107)
F
2 t .
F
107
integrals.
4 x
First integral = j* a(t) b(t) dt (3.108)
t=0
4t
T < T
F
suggested that low pass and d.c. off-set filters are necessary,
Johns [160] and Johns and Aggarwal [150] have reported the
1. The relay was analog circuitry based and did not involve
multiple reflections.
fault.
shown,
no
D = [ a(t) [ - [ b(t) | (3.111)
where a'(t) and b'(t) were time derivatives of a(t) and b(t)
is 2
w V ', where V 'is the peak-pre-fault voltage at the fault
F F
point. After a time of 2 x , the wave arrives back with a maximum
F
slope of 2 f a) V ' where f is the reflection coefficient at the
s F s
Ill
identify very high resistance earth faults has been provided. The
at both the ends of the line stop electronic counters at the ends
fault.
either ends of the line. This may not be the case always.
channel.
signal transmitter.
3.11. CONCLUSIONS:
The year by year development of computerized relays for the
impedance calculation.
117
applications.
of the line.
application.