Digital Protection For Power System
Digital Protection For Power System
DIGITAL
PROTECTION
FOR
POWER
SYSTEMS
A. T. Johns and
S. K. Salman
DIGITAL
PROTECTION
FOR
POWER
SYSTEMS
Other volumes in this series:
Volume 1 Power circuits breaker theory and design C. H. Flurscheim
(Editor)
Volume 2 Electric fuses A. Wright and P. G. Newbery
Volume 3 Z-transform electromagnetic transient analysis in high-voltage
networks W. Derek Humpage
Volume 4 Industrial microwave heating A. C. Metaxas and R. J. Meredith
Volume 5 Power system economics T. W. Berrie
Volume 6 High voltage direct current transmission J. Arrillaga
Volume 7 Insulators for high voltages J. S. T. Looms
Volume 8 Variable frequency AC motor drive systems D. Finney
Volume 9 Electricity distribution network design E. Lakervi and
E. J. Holmes
Volume 10 SF6 switchgear H. M. Ryan and G. R. Jones
Volume 11 Conduction and induction heating E. J. Davies
Volume 12 Overvoltage protection of low-voltage systems P. Hasse
Volume 13 Statistical techniques for high-voltage engineering W. Hauschild and
W. Mosch
Volume 14 Uninterruptible power supplies J. D. St. Aubyn and J. Platts (Editors)
DIGITAL
PROTECTION
FOR
POWER
SYSTEMS
A. T. Johns and
s. K. Salman
While the authors and the publishers believe that the information and
guidance given in this work is correct, all parties must rely upon their own
skill and judgment when making use of it. Neither the authors nor the
publishers assume any liability to anyone for any loss or damage caused
by any error or omission in the work, whether such error or omission is
the result of negligence or any other cause. Any and all such liability is
disclaimed.
The moral right of the authors to be identified as authors of this work has
been asserted by them in accordance with the Copyright, Designs and
Patents Act 1988.
Page
1 Introduction 1
1.1 Historical background 1
1.2 Performance and operational characteristics of digital
protection 2
1.2.1 Reliability 2
1.2.2 Flexibility 3
1.2.3 Operational performance 3
1.2.4 Cost/benefit considerations 3
1.2.5 Other features and functions 4
1.3 Basic structure of digital relays 4
1.4 References 5
4 Sinusoidal-wave-based algorithms 60
4.1 Introduction 60
4.2 Sample and first-derivative method 60
4.2.1 Basic formulation 60
4.2.2 Calculation of an approximation to the signal
derivatives 62
4.2.3 Error analysis 62
4.2.4 Practical considerations 64
4.3 First- and second-derivative method 65
4.3.1 Mathematical formulation 65
4.4 Two-sample technique 67
4.4.1 Prediction of values of peak (or magnitude) of signal
waveforms 67
4.4.2 Determination of phase angle between waveforms 68
4.5 Three-sample technique 68
4.6 An early relaying scheme 69
4.7 References 70
Index 200
Chapter 1
Introduction
1.2.1 Reliability
Digital relays can be designed to regularly monitor themselves. The process of
monitoring involves executing the relay software in conjunction with a
prespecified data set and comparing the results with those expected from a
properly functioning device. If the response turns out to be different from that
expected, an error is detected and the relay initiates warning signals to the
operator. This feature can be extended by programming the relay to monitor its
peripherals. It should be noted that self monitoring does not in itself directly
improve reliability, but it does provide a means of signifying the operational
state of protection equipment. This in turn has an indirect beneficial effect on
Introduction 3
1.2.2 Flexibility
Digital relays are generally more flexible than conventional devices. For
example, digital relays are programmable, and this in turn makes it possible to
use the same hardware for performing a variety of protection and control
functions by effecting changes in the software. It is also possible that the same
relay can be equipped with multiple characteristics and any revisions or
modifications required by changes in the operational conditions of the system
can be easily accommodated with virtually no changes in the hardware
structure.
The cost of conventional relays has continued to increase during the last two
decades and the cost/benefit ratio has consequently generally increased. On the
other hand, the advancement in microelectronic technology has led to a
substantial reduction in the cost of digital hardware but it must be remembered
that, in particular, it is the cost of the software that often dominates the overall
cost. Situations exist where the cost of software for commercially developed
equipment exceeds that of the hardware by at least an order of magnitude, and
in consequence digitally based equipment costs more than conventional
equipment.
On the other hand, high-volume digital relays, e.g. overcurrent relays, are
relatively cheap because the development costs are spread across many relays
and volume production allows the use of special microchip technology. Sales
volumes and development costs are also important; these vary significantly
according to the degree of functional complexity involved. Overall, it is true to
say that the substantial improvements in performance made possible by the
application of digital technology have resulted in a gradual reduction in the
cost/benefit ratio for digital-protection equipment.
4 Introduction
main applications to date have been in the areas of lines and transformers in
transmission systems. This book will concentrate on the latter applications, but
without reference to specific commercial implementations, details of which are
available from manufacturers' product literature.
1.4 References
1 ROCKEFELLER, G.D.: 'Fault protection with a digital computer', IEEE Trans.,
1969, PAS-88, pp. 438-461
2 MANN, B.J., and MORRISON, I.F.: 'Digital calculation of impedance for trans-
mission line protection', IEEE Trans., 1971, PAS-90, 1971, pp. 270-279
3 GILCHRIST, G.B., ROCKEFELLER, G.D., and UDREN, E.A.: 'High-speed
distance relaying using a digital computer, Part I: System description', IEEE Trans.,
PAS-91, 1972, pp. 1235-1243
4 ROCKEFELLER, G.D., and UDREN, E.A.: 'High-speed distance relaying using a
digital computer, Part II,' ibid, pp. 1244-1258
5 MAKINO, J., and MIKI, Y.: 'Study of operating principles and digital filters for
protective relays with a digital computer', IEEE Publ. No. 75 CH0990-2 PWR,
Paper C75 197 9, IEEE PES Winter Power Meeting, New York, Jan. 1975, pp. 1-8
6 GILBERT, J.G., and SHOVLIN, R.J.: 'High speed transmission line fault impe-
dance calculation using a dedicated minicomputer', IEEE Trans., 1975, PAS-94,
1975, pp. 544-550
7 RAMAMOORTY, M.: 'Application of digital computers to power system protec-
t i o n ' , / Inst. Eng. (India), 1972, 52, pp. 235-238
8 McLAREN, P.G., and REDFERN, M.A.: 'Fourier-series techniques applied to
distance protection', Proc. IEE, 1975, 122, pp. 1295-1300
9 JOHNS, A.T., and MARTIN, M.A.: 'Fundamental digital approach to the distance
protection of EHV transmission lines', Proc. IEE, 1978, 125, pp. 377-384
10 HORTON, J.W.: 'The use of Walsh functions for high-speed digital relaying', IEEE
PES Summer meeting, San Francisco, July 20-25, 1975, Paper A 75 582 7
11 GIRGIS, A.A., and BROWN, R.G.: 'Application of Kalman filtering in computer
relaying', IEEE Trans., 1981, PAS-100, pp. 3387-3397
12 GIRGIS, A.A.: 'A new Kalman filtering based digital distance relay', ibid, 1982,
PAS-101, pp. 3471-3480
13 McINNES, A.D., and MORRISON, I.F.: 'Real time calculation of resistance and
reactance for transmission line protection by digital computer', Elec. Eng. Trans. Inst.
Eng. Australia, 1970, EE7, pp. 16-23
14 RANJBAR, A.M., and CORY, B J.: 'An improved method for the digital protection
of high voltage transmission lines', IEEE Trans. 1975, PAS-94, pp. 544-550
15 BREINGAN, W.D., and CHEN, M.M.: 'The laboratory investigation of a digital
system for the protection of transmission lines', IEEE Trans. 1979, PAS-98, pp. 350-
368
16 GALLEN, T.F., and BREINGAN, W.D.: 'A digital system for directional compari-
son relay', ibid, 1979, PAS-98 pp 948-956
17 SMOLINSKI, W.J.: 'An algorithm for digital impedance calculation using a single
PI section transmission line model', ibid, 1979, PAS-98, pp. 1546-1551
18 TAKAGI, T., BABA, J., VEMURA, K., and SAKAGUCHI, T.: 'Fault protection
based on travelling wave theory, Part I: Theory', IEEE PES Summer Power meeting,
Mexico City, July 1977, Paper No A 77, pp. 750-753
19 TAKAGI, T., BABA, J., VEMURA, K., and SAKAGUCHI, T.: 'Fault protection
based on travelling wave theory, Part II: Sensitivity analysis and laboratory test',
IEEE PES Winter meeting, New York, 1978, Paper A 78, pp. 220-226
20 DOMMEL, H.W., and MICHELS, J.M.: 'High speed relaying using travelling
wave transient analysis', IEEE PES Winter meeting, New York, 1978, Paper A 78,
pp. 214-219
6 Introduction
21 CHAMIA, M , and LIBERMAN, S.: 'Ultra high speed relay for EHV/UHV
transmission lines - Development, design and application', IEEE Trans. 1978,
PAS-97, pp. 2104-2116
22 YEE, M.T., and ESZTERGALYOS, J.: 'Ultra high speed relay for EHV/UHV
transmission lines - Installation, staged fault tests and operational experience', ibid,
1978, PAS-97, pp. 1814-1825
23 JOHNS, A.T.: 'New ultra-high speed directional comparison technique for the
protection of e.h.v. transmission lines', IEE Proc, 1980, 127, pp. 228-238
24 JOHNS, A.T., MARTIN, M.A., BARKER, A., WALKER, E.P., and CROSSLEY,
P.A.: 'A new approach to e.h.v. direction comparison protection using digital signal
processing techniques', IEEE Trans., 1986, PWRD-1, pp. 24-34
25 VITINS, M.: 'A fundamental concept for high speed relay', IEEE Trans., 1981,
PAS-100, pp. 163-173
26 CROSSLEY, P.A., and McLAREN, P.G.: 'Distance protection based on travelling
waves', IEEE Trans., 1983, PAS-102, pp. 2971-2982
27 MANSOUR, M.M., and SWIFT, G.W.: 'A multi-microprocessor based travelling
wave relay - Theory and realization', IEEE Trans., 1986, PWRD-1, pp. 273-279
28 MANSOUR, M.M., and SWIFT, G.W.: 'Design and testing of a multi-
microprocessor travelling wave relay', IEEE Trans., 1986, PWRD-1, pp. 74-82
Chapter 2
Mathematical background to protection
algorithms
2.1 Introduction
Let us assume that numerical values f{xk) of some function y =/( x) are given at
equally spaced values of x, such that x, = #0 + A, x2 = x0 + 2h, . . . , xk = x0 + kh.
f(xk) can be obtained either analytically from some formula or by signal
sampling. Therefore at x = xk
IM=f{%+kh)=£ (2.1)
Using given numerical values fki it is possible to form difference tables
containing differences up to any desired order n.
In general there are three difference functions that can be derived for a given
sample set. These are forward differences, backward differences and central
differences. In digital protection, such differences are commonly used in the
8 Mathematical background to protection algorithms
4=1/V^ A% A 4f A 5/
x*
1 1.0000
-0.2929
2 0.7071 0.1631
-0.1298 -0.1106
3 0.5774 0.0525 0.0827
-0.0774 -0.0279 -0.0656
4 0.5000 0.0246 0.0171
-0.0528 -0.0108
5 0.4472 0.0138
-0.0390
6 0.4083
It follows that, in general, for any positive value of n, the nth order difference
equation takes the form of
i+,~ A (2.2)
By way of illustration, Table 2.1 thus shows the difference Table for the
function/(*) = l/V*.
Backward differences are basically defined in terms of present and preceding
sampled values, so that the first and second backward differences are given by
and
(2.3)
Mathematical background to protection algorithms 9
and
5 5l(5"-|/,- 1 / 2 (2.4)
Since samples are not usually available at half intervals, the easiest way to
implement central difference techniques is by using the samples before and after
the current sample, i.e. **__, and xk+l. Thus, for example, the first central
difference becomes
Given a sampled data set taken at x*_{, xh xk+u . . . , the sequences of numbers
defining the forward and backward difference functions are identical. The
distinction between the three functions essentially lies in the time at which a
given function is available for use in further digital processing and, where a
difference function is used in approximating the differential of an actual signal
(to whatever order), the degree of the resulting approximation. Thus, for
example, the forward difference function is available at the time the k 4- 1 th
sample is available whereas that for the backward approximation is available at
the kih sample instant. Furthermore, the degree of accuracy to which a
particular function can be used to approximate, say, the first differential of a
signal varies according to the waveform involved. For example, with reference
to Figure 2.1, it will be evident that a central difference function using samples
taken at instants corresponding to k + 1/2, k— 1/2 may give a better approxi-
mation to the differential of the particular signal shown than is the case for the
forward or backward difference function. In other words, the slope of a straight
line joining BD in Figure 2.1, which corresponds to the first derivative
calculated using the central difference formula, is close to the slope of the
tangent at C.
In practice, it is not generally possible to assume a particular form to the
waveforms encountered and the accuracy in approximation that can be
achieved with each method varies from point to point as a given waveform is
tracked. Each method therefore has advantages in particular applications. In
any event, the basic difference functions form the basis of more accurate
approximations to signal derivations by using interpolation formulas explained
in Section 2.4.
10 Mathematical background to protection algorithms
Figure 2.1 Forward, backward and centre difference function related to an arbitrary
waveform
Closely associated with difference function operators A, V and d are the shift
operator E and the average operator fi. The shift operator is defined as the
operator that increases the instant at which a function is sampled by one
tabular interval. Thus
E(Jk)=f(Xt+h) (2.5)
A relation between the shift operator and finite difference operators can be
derived as follows:
Mathematical background to protection algorithms 11
A= £~l (2.6)
or
£=1+A (2.7)
from which:
V=l-£-! (2.9)
6 = Em-E'm (2.11)
The average operator ju is defined as follows:
^=l/2(/*+l/2+/,-1/2) (2.12)
or
m
+ E-m) (2.13)
12 Mathematical background to protection algorithms
fk+l=f(xt+h)
Similarly, the value of the function at x = xk+ph is equal to its value at xk shifted
by an interval equal to p times the sampling interval h, so that
(2.14)
Substituting the value of E from eqn. 2.7 into eqn. 2.14 we obtain
. . . +k/,fA"]fk (2.15)
where any value kmj is often known as the mth binomial coefficient, given by
p(p-\)(p~Z)...(p-m+\)
J
ml
It follows that an estimate of the value of a function can be obtained from the
summation of terms given in eqn 2.16:
r i
k
31
n\
where n is the order of the polynomial.
Mathematical background to protection algorithms 13
r(r-l) r(r-l)(r-2)
ft+r =/(** + rh) =£ + rAfk + —^—I2fk ^ ^ &/>+•••
r(r-l)(r-2) . . . (r-«+l)
+- -, -*% (2.17)
Example
Calculate the value of the function f(x) = 1/V# at x = 3.5 from its forward
difference Table 2.1, using eqn. 2.17.
Solution
Let us first estimate the value using only the first two terms of eqn. 2.17 so that
we have^ +r —fk + rAjf, where £ = 3 and r=0.5.
With reference to Table 2.1, / 3 and A/3 are equal to 0.5774 and -0.0774,
respectively, so that
r(r-l)
— ^ — A 2 /,
0.5(0.5-1)
/3.5 = 0.5774 + 0.5(-0.0774) + (0.0246)
= 0.535625
This corresponds to an error of 0.21%, which indicates that the more terms of
eqn. 2.17 are used, the better the accuracy obtained. However, for this example,
it is sufficient to take only the first two terms. In digital protection applications,
eqn. 2.17 can be very useful in determining a numerical estimate of the
differential of a function; this will be discussed in Section 2.4.
14 Mathematical background to protection algorithms
Consider the function shown in Figure 2.1; a backward value of this function,
say/_, =f(xk — h), can be used to obtain its current value/=/(%) by using the
shift operator E. Thus/_, =f(xk- A) = E~lf(xk), o r / , = £/*„, .
In general the current value/ can be found from the function/^ which isp
intervals behind it, which can be succinctly stated mathematically by / =
(2.18)
/>G&+l)(/> + 2) •• • (p + m-1)
ml
It follows that the value of a function can be estimated by eqn. 2.19 by using the
backward difference operator
As previously, eqn. 2.19 can be written in terms of a function of the shift interval
h so that, for a fractional shift sh say, eqn. 2.19 becomes:
, s(s+\)(s
v 2
/+
s(s+\)(s + 2) . . . (s + n-\)
+~ ~{ % v (2.20)
r(r-l) (r+i)(r-i)
2 3
[ (r+l)r(r-l)(r-2)j4
.r^r , r ( r - l ) , 2 r , (r-H)r(r-1)
(2.22)
(2 23)
'
(2.24)
16 Mathematical background to protection algorithms
^ - ^ A 2 + ^ A 3 - - • ^fk (2.25)
VfN+
Setting s = 0 leads to the first derivative, which becomes
V
/*'=/'(%=* > 4 ( +\^+l^ + - • \ft (2.27)
and
( ^ ) (2.28)
Mathematical background to protection algorithms 17
6 30 / ( 2 2 9 )
Example
Use the backward interpolating formula to derive an expression for estimating
the derivative of a sampled function after
(i) two samples are made available
(ii) three samples are made available, but the derivative is to be calculated for
the instant in time that the second sample is available
(iii) repeat (ii) for the second derivative.
Solution
(i) By considering only the first term of eqn. 2.27, we obtain an estimate based
on only two samples; this is given in eqn. 2.30.
1 1
fk=-^fk=-hUk-fk-x ) (2.30)
(ii) Since the derivative is required to be found following the second sample,
let us assume that the last sample is^+,, and hence the second sample
would b e ^ . Now eqn. 2.27 can be expressed in terms offk+l by using the
relationship fk = E~lfk+l which, from eqn. 2.9, can be written in the form
given in eqn. 2.31:
^ = (1-V)j£ +1 (2.31)
By substituting eqn. 2.31 into eqn. 2.27, and simplifying we obtain
(2.32)
Now by considering only the first and second terms, eqn. 2.32 reduces to
eqn. 2.33, which in turn defines, as required, an estimate of the rate of
change of the function at an instant in time midway between the pair of
samples used.
where x{ is the value of the independent variable taken at the fth measurement
and j>; its corresponding dependent variable. Such data may be generated, for
example, by the measurement of current flowing through an electrical element
(independent variable) and its corresponding voltage across the element
(dependent variable). In digital protection, it is common to sample a relaying
voltage at discrete instants in time; in this case, the dependent variable can be
regarded as the sampled value and the independent variable the discrete value
of time at which it is sampled. Let us now determine a function u(x) to
approximate the actual function such that
To define this function fully, it is necessary to choose the values of the constants
OQ, al9 . . . , am such that the function best fits the given data. One criterion,
which is often regarded as giving the best fit, requires that the sum of the
squares of the displacements of the measured points (yt) from the curve defined
by eqn. 2.36 should be a minimum.
Mathematical background to protection algorithms 19
least-square
X yC polynomial curve (u)
Figure 2.2 illustrates the basis of curve fitting using the least-squares method.
Let the vertical displacement of measured points (jy,) from the fitted curve u be
£,, in which case:
£,=.?,-«(*,-) (2.37)
The sum of the squares of errors is thus as given by eqn 2.38.
N N
(2.38)
The necessary condition for S to be a minimum is that its partial derivative with
respect to the parameters a0, a2i . . . etc. should be zero, i.e.
dS
— =0 (2.39)
dak
When eqn. 2.39 is applied to eqn. 2.38, iV equations will be formed, often called
'normal equations'. Thus
dS
-
da0
N
dS
iirr:.... _ . c\
da i
dS
20 Mathematical background to protection algorithms
N 2* 2^
x
t 2LJ
Xi
2Ld Xi
/*iA| a
\
(2.40)
Eqn. 2.40 represents a system of simultaneous equations which can be solved
for the unknown a0, a{, . . . . am.
Example
The following data were obtained experimentally:
-1.0 -0.1 0.2 1.0
1.0
y 1.099 0.808 1.0
Use the least square method to fit a straight line to this data.
Solution
The straight line equation is given by
u = ao + axx
The normal equations for this function are obtained directly from eqn. 2.40.
Setting m = 1, as required for a straight line approximation, the normal matrix
equation is:
» 2^ a0 2* (2.41)
ax
Since there are four data points, the limit on the summations is N=4, and this
gives the following constant values in eqn. 2.41:
*,.=-1.0-0.1+0.2+1.0 = 0.1
xji= (-1-0) (1.0) + (-0.1) (1.099) + (0.2) (0.808) + (1.0) (1.0) = 0.0517
Substituting these values into eqn. 2.41 and solving for a0 and au we obtain
ao = 0.9773 ax = -0.0224
The straight line equation which best fits the data set is therefore given by
u = 0.9773 - 0 . 0 2 2 4 *
In the previous analysis, it has been assumed that the variabley depends only
on a single variable x. The general case is such that the variabley may depend
on more than a single variable. An example that illustrates this case is a three-
phase unbalanced system in which the phase voltage va depends on the three
associated phase currents ia, ib and ic. Therefore, in order to determine a value of
va, the currents ia, ib and ic must be measured. In this case, the voltage va is a
function va=f(ia, ib, ic). Thus each measured voltage is related to the measured
currents by a data set (ial, ibU icU va{), (ia2y ib2, ic2, va2) . . . (iaNy ibN, icN, vcN).
More generally, for a function of z variables, the data set involved
would take the form (*,„ x2U xzu . . . xzUyx), (xl2, xm x32 . . . xz2iy2), . . .
(xlN, x2N, x3N, . . . xzN,yN). The principles involved in deriving a least-squares-
error based function for extended data sets associated with multivariable
functions are similar to those outlined above for the single-variable case. Full
details of the precise methodology involved can be found in References 4 and 5.
(2.42)
0«»*i + 0*2*2 + • * ' + anmxm =yn
A X = Y (2.43)
nXm mX1 nX 1
22 Mathematical background to protection algorithms
where
#22 a
2m x2 J2
A= Y=
Xm
AT A X = i4T F
(2.44)
mXn nXm mX\ mXn nX\
X\
x2
(2.45)
It should be noted that the limits of all the summations in eqn. 2.45 are from
i=\ to n. Eqn. 2.45 in effect defines the normal equations obtained by using the
least-squares method. It can also be seen from eqn. 2.44 that the matrix A1 A is a
square matrix having a dimension of m X m. Therefore, by premultiplying eqn.
2.44 by [i4Ti4]~!, we obtain the vector of unknowns as given by eqn. 2.46.
The matrix [i4Ti4] l A is known as the left pseudoinverse of A and, since eqn.
2.46 is the solution of the least-square error eqn. 2.45, it is therefore clear that
the pseudoinverse approach provides the least-squares solution.
Data obtained from physical measurement often contain errors that can be
conveniently defined as the difference between the correct and measured values.
The correct value means the true, theoretical or average value. In practice, it is
Mathematical background to protection algorithms 23
Example
Let us assume that it is required to find the smoothed value for a measured
three-point data set using linear equations. Let the measured data be denoted
by (^bji)? (*2J J2) a n d (#35^3)- The process of finding the smoothed values of
the given data to fit a linear equation basically consists of the following steps:
(a) The data are fitted, using the least squares method, to the linear function of
eqn. 2.47.
y = aQ-¥a,x (2.47)
This of course involves the determination of the constants a$ and ax to give
the best fit.
(b) The smoothed value Yk can then be determined at the corresponding xk by
using eqn. 2.48.
Yk = ao + alXk (2.48)
To simplify the analysis let us:
(i) Assume that the data are taken at equal spaces in the x variable, such that
xk = kh (2.49)
where h is the interval between values.
(ii) Renumber the measured data in such a way as to extend them symmetri-
cally from negative to positive around a defined middle point. Therefore the
data set becomes (*_,,j_,), (*o,Jo)> (*i,J>i)-
From previous work (see eqn. 2.41), the measured data set can be fitted to eqn.
2.47, using the least-squares method, to obtain the following normal equations:
1=1 /= I
N N N
i= 1 i= I
(kh) (2.52)
24 Mathematical background to protection algorithms
I
2
^ (2.53)
k=-\ k=-l k=-\
%=O>-i+J>o+Ji)/3 (2.54)
From eqn. 2.53 we likewise obtain
Therefore
y_1 = (5j_, + 2 j 0 - j , ) / 6 (2.56)
Similarly the smoothed values YQ and Yx can be found as
Fo=O'-«+J>o+J>.)/3 (2.57)
Yl = (-j.l-¥2y0^5yl)/6 (2.58)
fit) = j + 2 a
« cos na)ot + X *. sin Wo1 (2-59)
where (O0 is the angular fundamental frequency = 2Jtf0 = 2nl T
T is the time period of the fundamental component
na)0 is the nth harmonic angular frequency
t{ is arbitrary
Given the known function f(t) (which in most practical situations is a function
that varies with time), the coefficients OQ, a,, bu . . . an, bn can be determined
from expressions of the form of eqns 2.60-2.62 [10, 11].
2 f'i+r
*o= y / ( 0 dt (2.60)
J /i
Mathematical background to protection algorithms 25
2 f^T
f(t) cos natotdt (2.61)
2 f'.
/(*) sin na)ot dt (2.62)
J /,
^ (2.63)
n=0
where
A0 = j,00 =0 (2.64)
and
where
£ = ( f l . - A ) / 2 , B = ± l , ± 2 . . . . ±00
-I f" r
f(t)dt,n = O
Example
Figure 2.3 shows a periodic function f(t) with a period of T, which is given by
A
Solution
Representation of the function f(t) in the frequency domain requires the
determination of its Fourier coefficients:
From eqn. 2.67
=-f"
+T
* TJ
f(t) , «= ± 1, ± 2, . . . ±00
t,
f(t)dt,n =
Using the definition o£f(t) given in eqn. 2.68 and taking the limits of integration
from — r/2 to r/2, we obtain
1 f'/i
dt, n = ±l, ± 2 , . . . ± o o
Ar-jna>Qt
~T[-jnco
">«>oJ_
0 r/2
f(t)
T -c/2 x/2
Figure 2.3 Illustrative periodic function
Mathematical background to protection algorithms 27
At _ A
T "" 2
Figure 2.4 Fourier coefficients Fn (n — 0, 1, . . .) versus nfor the periodic function shown
in Fig. 2.3 for (a) z/T= 1/2, and (b) x/T= 1/4
At
At
(2.69)
where
sin x
sinc(^) =-
We have seen in previous Sections how any periodic function can be treated
using the Fourier series. However, in practice, non-periodic functions often
arise. This is particularly so in protection, where transient waveforms have to
be processed. Therefore the question arises as to how non-periodic functions
can be treated.
The first step in answering this question is to note that a non-periodic
function can be regarded as one having a periodic time interval T equal to
infinity (J1—» o°). It follows that/(/) = lim r _ 0 0 ^(/), where fp(t) is a replica of the
single function that is repeated after an assumed infinite time. If we assume that
the Fourier coefficients Fn of the function^(^) are known, then fp(t) itself can be
found from substituting eqn. 2.67 into eqn. 2.66. Therefore by assuming
tx = —772, we obtain
I fm
J -Til
or
1 j ; ([Tn.
(2J0)
ZJl
„=_« \J -r/2 F
Now, on allowing T—* °° , the following observations apply:
f(l)=\imfp(t)
Cm y«<v2jr
2(
H= —oo \J —772
>) T (2J1)
or
=
h \" eim da)
" (2-72)
Mathematical background to protection algorithms 29
The term between the brackets in eqn. 2.72 is seen to be a function of frequency
and will be denoted by F((o), i.e.
The frequency function F((t)) is called the Fourier transform off(t). In terms of
the Fourier transform, the function/(/) can thus be written as
jQ>t
da) (2.74)
The time domain function f(t) and the corresponding frequency domain
function F(co) are commonly known as transform pairs, and any non-periodic
function of time can be transformed to a continuous frequency spectrum F(a))
via the Fourier transform integral of eqn. 2.73. The inverse Fourier transform
likewise inverts the process by transforming a continuous frequency spectrum
into the corresponding time variation.
Example
Find the Fourier transform of the rectangular pulse shown in Figure 2.5(a).
Solution
The pulse shown in Figure 2.5(a) can be expressed mathematically as
=[ lo
elsewhere
l
The Fourier transform of PT(t) is obtained from eqn. 2.73 as
PT{w)=(" Pt(l)-j°"dt
J -00
Figure 2.5(^) shows how PT((o) varies with angular frequency. It is interesting to
note that the shorter the pulse, the more quickly the Fourier transform falls to
30 Mathematical background to protection algorithms
Pr(t)
-1/2 x/2
Figure 2.5 (a) Non-periodic (transient) pulse and (b) corresponding Fourier transform
Mathematical background to protection algorithms 31
S=k/T
(b) Multiplication
When two Walsh functions are multiplied together, another Walsh function is
produced such that
where h®k denotes a number whose binary numeral has 0s in those positions
where the binary numerals for h and k are alike and has Is where they are
different. For example
Because 3 (in decimal) =011 (in binary), and 5 (in decimal) = 101 (in binary),
3 ® 5 = 110 (binary) = 6 (decimal).
(c) Symmetry
With reference to Figure 2.6, it can be concluded that
/ T-t\
Wai *, - = - = (-1)* Wai(t, t/T) (2.78)
Thus Wai (A;, t/T) has even or odd symmetry about /=7V2 depending on
whether k is even or odd.
32 Mathematical background to protection algorithms
1.
Wal(O, t/T)
-1
Wa!(1,t/T)
Wal(2, t/T)\1 1
T
Wal(3, t/T)\) I
T
-1
Wa!(4, t/T)
-1
Wal(5, t/T)
-1
1
Wal(6, t/T)
-1
Wa!(7, t/T)
-1
Figure 2.6 The first eight Walsh functions of integral index kfor
Mathematical background to protection algorithms 33
with
jr = 0, 1,2, ,N-\
* = 0, 1,2, , JV—1
where II is the factorisation symbol in which the parameters that follow are
multiplied together. kn j n are binary bits of k, j or
Example
Find the fourth element of the Walsh function having the order of 5 and length
JV=8, i.e., Wal(5, 4).
Solution
therefore
P=3
k = 5 (in decimal)
hence
34 Mathematical background to protection algorithms
Wal(5, 4 ) = J | (
r=0
= (-l)°-(-l)°.(-l)1 (because Jo = ^ = 0)
= -1
j 0 1
*
1 1 1 1 1 0
1 1 -1 -1 -1 1
1 1 -1 1 2
Wal(kJ) = 1 1 1 -1 3
1 -1 -1 -1 4
1 -1 1 1 5
1 -1 -1 1 -1 6
1 -1 -1 1 7
,tlT)
(2.81)
where Wk is the A:th coefficient of the Walsh series and Wai (A:, t/T) is the A:th
order Walsh function. This technique mirrors the previously described method
of constructing a periodic waveform by a number of sinusoidal components
using the Fourier series technique.
Mathematical background to protection algorithms 35
^_
an = V2
V2 [r
F2n = vT7T = -Y /(/) cos nct)ot dt, n= 1, 2, . . .
^ Jo
If the same function f(t) is expanded using Walsh series, we have:
(2.85)
l
W0 = - \ [Fo + V 2 F, sin (o0t + V 2 F 2 cos a)ot + V 2 F 3 sin 2(o0t
. . .]Wal(0, tlT)dt
x=
~f\ J €O S
l, tlT)dt
i r/i+r
Wk = - I [Fo + V 2 F, sin a)ot + V 2 F 2 cos coot + V 2 F 3 sin 2a)ot
36 Mathematical background to protection algorithms
wk Fk
or
W=AF
where
1 f'. + r
V 2 f'i + r
A:, 2m-1 =—— I sil Wal(*, t/T)dt (2.87)
V2 pi
cos(mo)0)t ,t/T)dt,m=l,2,3,...
J '1
Eqn. 2.87 can be simplified by taking the lower time limit / | = 0 , and
normalising the time with respect to the period T, such that
tf=-,dt=Tdtf (2.88)
When eqn. 2.88 is substituted into eqn. 2.87 (replacing the subscript 2m by K
and simplifying) we obtain:
I ^r)/'Wal(*,/')
J
By direct substitution of the Walsh functions Wai (A:, / ' ) , shown in Figure 2.6, it
is possible to calculate the elements of the matrix A having any given size. For
example the elements of a 12 X 16 A-matrix have been evaluated using this
approach to obtain the matrix eqn. 2.90 (see page 37).
As indicated in eqn. 2.86, the relationship between Fourier and Walsh
coefficients is given in matrix form as
W=AF (2.91)
Mathematical background to protection algorithms 37
CD
CM
2 O O O O O O O O O O O O O
.-* O O O O O O O O O O O O O
CD CO
CM
p °1
o d d
o o I o o o I o o o I o o
CO
CD r~5 00
CM • ^O
^ O CM
odooo I ooodooo
CM
-* o o o o
1I o o o o o o o o
^
^ O O O O O O O O O O O O O
o ^ o
22 © S
~ o o o o o oCO I o oO
too o o o
<£5
CD O O O O O OJ O O O O O O O
o
O O O O O O O O O O O O O
o
O O O O O O O O O O O O O
^ S CO
O r^. ^
CD O O I O O O O O O O O O O
O CM
CO t-"^ O
m oooooo* o o o I o o o
O CO
CD CO
Th O O O O O O O O O O O O O
CO
CO O O O O O O O O O O O ! O
O r^ H
CD CO O
o o o o o o o o o o I oo
CO ^
^. t^- t^»
§ <« p
CD d d
o d o o o Io o o I o o o
— o o o o o o o o o o o o
38 Mathematical background to protection algorithms
2.9 References
1 KREYS^IG, E.: 'Advanced engineering mathematics' (5th Edition, John Wiley &
Sons, 1983)
2 WYLIE, C.R.: 'Advanced engineering mathematics' (4th Edition, McGraw-Hill,
1975)
3 LaFARA, R.L.: 'Computer methods for science and engineering' (Intertext Books,
1973)
4 HILDEBRAND, F.B.: 'Introduction to numerical analysis' (2nd Edition,
McGraw-Hill, 1974)
5 GERALD, C.E.: 'Applied numerical analysis' (4th Edition, Addison-Wesley, 1989)
6 APOSTOL, T.M.: 'Calculus—introduction with vectors and analytic geometry',
Vol 1 (3rd Edition, Blaisdell Publishing Company, 1962)
7 RALSTON, A.: 'A first course in numerical analysis' (2nd Edition, McGraw-Hill,
1978)
8 KAPLAN, W. 'Advanced mathematics for engineers' (Addison-Wesley, 1981)
9 STANG, G.: 'Linear algebra and its applications' (Academic Press, New York, 1976)
10 MAYHAN, R.J.: 'Discrete-time and continuous-time linear systems' (Addison-
Wesley, 1984)
11 McGILLEM, C D . , and COOPER, G.R.: 'Continuous and discrete signal and
system analysis' (2nd Edition, Holt, Rinehart & Winston, 1984)
12 BLACKMAN, N.M.: 'Sinusoids versus Walsh functions' Proc. IEEE, 1974, 62,
pp. 346-354
13 SCHREIBER, H.H.: 'Bandwidth requirements for Walsh functions' IEEE Trans.,
1970, IT-16, pp. 491-493
14 KENNET, B.L.N.: 'Note on the finite Walsh transform' IEEE Trans. 1970, IT-16,
pp. 489-491
Chapter 3
Basic elements of digital protection
3.1 Introduction
Operating voltages and currents flowing through a power system are usually at
kilovolt and kiloampere levels. However, for digital processing, it is necessary to
reduce the primary measurands to manageable levels. Therefore, the analogue
signals are converted to digital form, thereby allowing subsequent digital
processing to be performed to determine the circuit state.
In this Chapter the basic principles underlying the conversion of analogue
signals into equivalent digital forms will be explained. We shall also explain the
essentially common features of various digital relaying schemes, other detailed
aspects being discussed in later chapters.
transducer
surge
protective
circuit
signal
conditioning
LP
subsystem
filter
analogue
multiplexer
sample
hold
circuit
A/D
conversion converter
subsystem
digital
multiplexer
digital
processing D/O D/l memory CPU
relay
subsystem
D/A
LP = low pass
A/D = analogue-to-digital
D/A = digital-to-analogue
CPU = central processor unit
D/l = data input
D/O = data output
Figure 3.1 Basic components of a digital relay
Basic elements of digital protection 41
isolating
surge capacitor transformer
to LP filters
output voltage
input voltage
A
1
Figure 3.3 Characteristic of low-pass filters; (a) ideal filter response, (b) practical filter
response
All the above features play a part in the overall dynamic response of digital
relay systems. In particular, in systems where a very high speed decision is
required, it is particularly important to ensure that the low-pass filter is
designed to have a cutoff frequency that gives an overall performance that is not
degraded by long filter delays [6].
inputs 4
output
5
6
7
Figure 3.4 Principle of multiplexing
/*=I/2 (3.2)
W-*T.) (3.3)
f(nTs)d(t-nTs) (3.4)
Basic elements of digital protection 45
f(t) «s(t)
-3T S -2T S -T S 0
• t 11 t,
T s 2TS 3T s 4T s t
M CDs
M,2C0s 03
-3T S -2T S -T S 0
T
Ts 2TS 3TS
A"-
t \ 0
A.
Figure 3.5 Sampling processes; (a) the sampler, (b) representation of the sampler, (c)
functions f(t), s(t) andf(t), (d) Fourier transforms off(t), s(t) andfs(t)
Figure 3.5(c) shows the three time functions/(/), s(t) and f(t). It can be seen
that the sampling function f(t) consists of a train of pulses spaced equally by a
period equal to the sampling interval Ts. The amplitude of each sample is the
same as that of the original signal at the respective sample time nTs. Therefore,
the resulting samples are included within the original signal envelope, as shown
in Figure 3.5(c).
To show how and under what conditions/(/) can be uniquely reconstructed
from^(^), let us examine the spectra of the three time functions f(t), s(t), and
f(t). This is shown in Figure 3.5(</), in which it is assumed that the function
F(a)) contains no frequency above O)m. The spectrum of the sampling function is
thus given by
(3.5)
46 Basic elements of digital protection
Fs(a>)=—F(to)*S(a>) (3.6)
The convolution of F(CD) with each pulse of S((o) produces Fs((o) displaced in
frequency steps equal to na)s. Thus
(3.7)
(3.8)
If this condition is fulfilled, the passing of the signal fs(t) through a low-pass
filter (see Figure 3.5(d)) with a bandwidth such that its cutoff frequency coc is
given by eqn. 3.9, results in a perfect digital reconstruction of the analogue
signal/(/).
(3.9)
i /
\ 1 \\
II ' \ i M 1
V l
I
U/
l
i \\
1 \
I 1 1 1
\ \ \\ 1
^ /
Vc
Figure 3.7 Sample and hold circuit principle (a) switching circuit, (b) analogue input, (c)
control waveform, (d) circuit output
Basic elements of digital protection 49
input 1
• output
input 3
A B
Yi (input 1)<
> Y (output)
Y 2 (input 2){
VR
O
i
R/2
1 O
R/2 2
s2
1
R/2 3
most significant bit (MSB)
1
Figure 3.10 Basis of simple digital-to-analogue converter arrangement
R (3.12)
It is clear from eqn. 3.12 that the output signal voltage Vo is directly
proportional to the binary number W3W2WX Wo- For illustrative purposes assume
that the binary input array corresponds to 15 (decimal), i.e. W3W2WlW0= 1111;
in this case the output is proportional to (23 + 22 + 2* + 2°) = 15 as required.
Similarly, an input word of 0011 would give an output proportional to
(0 + 0 + 2 1 + 2°), corresponding to the input number 3.
52 Basic elements of digital protection
digital output
O W3(23)
u
analog input A/D r\
converter
^\
U
Va
r\ W0(20)
•O
U
4-bit counter
8 4 2 1 digital
output
comparator,
output = 1 whenVj>Vc
D/A converter
analog
input
i
— i
i
4
I
i
-
counter stops at 0101
3 -h Vc
2 -
1 -
i—i—i—i—>
2 3 4 5 6
multiples of clock time step
Figure 3.12 Counter-controlled converter (a) circuit arrangement (b) illustrative response
analogue-integrator
Vo<0
0 Vo>0
W0W1W2W3
digital output
comparators registers
The input ranges from 0 to Fo and is divided into eight reference segments, six
of which have a value of VJ1 the two end segments having a value of Fo/14
each. If an input voltage Vh with magnitude ranging from 0 to Fo/14, is applied
to the converter, then the outputs of all comparators Ci-C 7 are set to logical 0.
The digital representation of this at the output of the converter is seen to be 000.
This is equivalent, as required, to a zero analogue voltage. It will be noted that
an error of Fo/14 is introduced on account of the input being equal to one level
of resolution (F o /14). This is often called a 'quantisation' error.
If the input voltage magnitude is between Fo/14 and 3FO/14, then the output
of all the comparators is zero, except the first one which will be 1, i.e.
CiC2C3C4C5C6C7= 1000000. The output of the comparators is transferred to the
register flipflops at the occurrence of a clock pulse and the register output is
converted by the decoder to a three-bit unipolar binary code.
start/
elimination of low-frequency
" " and higher harmonic
components
Figure 3.15 Flow chart for the software of a digital protective relay
Basic elements of digital protection 57
first of these is the operating principles and performance required, which either
leads to the development of a special algorithm or the implementation of an
existing one. This factor greatly affects the determination of the sampling
frequency, type of hardware structure and the data input hardware system.
The second factor is the digital filtering. Subharmonics, as well as high
harmonic components, can cause false tripping, failure to trip and variation in
protective relay performance. The operating principle and digital filtering must
in general provide for a wide application range and requirements relating to
speed of response to system faults, the latter being influenced by the necessary
computation time.
Figure 3.15 shows an example of a flow-chart for the software of a typical
digital protective relay. The algorithms used and the software required vary
significantly according to the application, and much of the work of later
chapters will be concerned with specific algorithms for meeting a variety of
protection performance and application requirements.
Memory IC memory
auxiliary
relay
unit
§•
^
2
3.7 References
1 WRIGHT, A.: 'Current transformers—their transient and steady state performance'
(Chapman & Hall, 1968)
2 STALWESKI, A., and WELLER, G. E.; 'Novel capacitor divider voltage sensors for
high-voltage transmission systems', Proc. IEE, 1979, 126, pp. 1186-1195
3 'Computer relaying', IEEE Tutorial course 79 EH01487-PWR, 1979
4 GILCHRIST, G.B., ROCKEFELLER, G.D., and UDREN, E.A.: 'High speed
distance relaying using a digital computer, part I—system description', IEEE Trans.
1972, PAS-91, pp. 1235-1243
5 SEKINE, Y., HATATA, M., and YOSHIDA, T.: 'Recent advances in digital
protection', Elect. Power & Energy SysL, 1984, 6, pp. 181-191.
6 JOHNS, A.T., and WALKER, E.P.: 'Genesis and evolution of advanced protection',
Power Eng.J. 1991, 5 (4), pp. 177-187
7 CAHILL, J.J.: 'Digital and microprocessor engineering' (Ellis Horwood, 1982)
8 MAYHAN, R.J.: 'Discrete-time and continuous-time linear systems' (Addison
Wesley, 1984)
9 NATARAJAN, N.A.T.: 'Discrete-time signals and systems' (Reston, 1983)
10 TAUB, H., and SCHILLING, D.: 'Digital integrated electronics' (McGraw-Hill,
1977)
11 STANIER, B.J.: 'Electronic and integrated circuits' (Adam Hilger, 1985)
12 SCHILLING, D., and BELVE, B.: 'Electronic circuits—discrete and integrated'
(McGraw-Hill, 1979)
13 ibid, 3rd edition (McGraw-Hill, 1989)
14 BOOTH, T.L.: 'Digital network and computer systems' (2nd edition), (John Wiley,
1978).
15 HILL, F.J.: 'Introduction to switching theory and logical design' (John Wiley, 1981)
16 MALVINO, A.P.: 'Digital computer electronics' (McGraw-Hill, 1977)
17 MAKINO, J., and MIKI, Y.: 'Study of operating principle and digital filters for
protective relays with digital computers', IEEE PES Winter Power Meeting, New
York, January 1975, 75CHO990-2 PWR, Paper No. C75 197-9, pp. 1-8.
Chapter 4
Sinusoidal-wave-based algorithms
4.1 Introduction
The algorithms covered in this Chapter assume that the post-fault current and
voltage waveforms are sinusoidal. This assumption is not, of course, generally
valid, particularly when EHV or UHV network applications are involved.
However, in practice, the signals processed are often prefiltered and, in lower
voltage distribution systems in particular, the waveforms often very quickly
attain a nominally sinusoidal form. Historically, algorithms developed for use in
applications where the signals processed are nominally sinusoidal were the first
to emerge [ 1, 2]. Most of the early work involved applying the technique to the
calculation of transmission-line fault-loop impedances. However, the methods
are equally applicable to determining the magnitude and phase of relaying
currents for differential protection of lines and plant.
All sinusoidal-based algorithms are designed to predict either the peak or
squared peak of the compared waveforms. They may be loosely classified into
two broad groups: those that use sample and first derivative (or first and second
derivative), and those that use two or three samples to predict the peak or
squared peak values.
When eqns. 4.1 and 4.3 are combined together, we obtain eqn. 4.4, which
defines the peak values of relaying signal ${t.
(4.4)
Similar equations to eqns. 4.1-4.3 can be derived to describe the peak value of
the second relaying waveform s2{t), i.e.
Substituting eqns. 4.12 and 4.13 into 4.9, we thus obtain the argument of the
measured impedance in terms of the instantaneous values of the measurands v
and i, together with their derivatives:
x
-i2in-\ia)Jil) (4.14)
(v+V+..)vk (4.15)
At/2 At/2
Figure 4.1 Calculation of voltage v0 halfway between samples taken at times tk and tk+I
and (4.18)
Av 1
Given that we are dealing with a sinusoidal waveform, the sampled value at
time tk can be written as vk = V sin (O0th which in turn is given by
^=Fsina>0(/0-A*/2) (4.19)
The voltage sampled one time interval later (vk+l) is likewise given by
vk+l=Vsin a)0(t0 + A//2) (4.20)
By substituting eqns. 4.19 and 4.20 into eqns. 4.18 we obtain
os(<Vy/2) (4.21)
sin(a>oA//2)]/A/ (4.22)
Because the value ofa)0At/2 is small, the sine and cosine terms can be readily
approximated by
co0At
cos - y - ~ 1 (4.23)
64 Sinusoidal-wave-based algorithms
a)0At a)0At
sin—~-y- (4.24)
By substituting eqns. 4.23 and 4.24 into 4.21 and 4.22, we obtain
vo^V sin(ttM0) [1 - (a)0At)2/8] (4.25)
and
vf0^a)0Vcos a)oto (4.26)
The estimated value of the peak of the voltage signal using sampled data at
t0 (Fo) can be found using eqn. 4.4:
When this equation is combined with eqns. 4.25 and 4.26, we obtain
or
F o « F[l - (colAt2 sin2 (o0t0)/S] (4.28)
It follows that the percentage error (e%) associated with the estimated value of
the peak of the voltage waveform will be
2
Eo/o =
a)i(At)
— sin2 m ^ x 10Q (4 29 ^
o
Eqn. 4.29 indicates that the error in the estimated peak value is a function of the
system frequency coOy the sampling interval At and the time at which the
estimation takes place. For a sampling interval At of 0.5 ms, the error is seen to
be a maximum when the estimation takes place at the extreme of the waveform,
i.e. at O)Qt0= ±nf, n= 1, 2, . . . On a 50 Hz system, this error is of the order of
0.3% (0.45% for 60 Hz systems).
(4.30)
Eqns. 4.30 use forward, central and backward points, respectively. However, in
most applications of the algorithm it is preferable to use the backward
expression, because a smoothed value can be generated from the data already in
hand, thus speeding up the overall response.
Combining these two equations results in an equation for the square of the peak
of the assumed sinusoidal voltage:
[ ( Y l (4.33)
The corresponding equation for determining an approximation to the peak of
the current is likewise
/* = ! [(,-')*+(-Y1 (4.34)
The square of the magnitude of the measured impedance is
v
= — tan
\a)oi'
It follows that the argument of the measured impedance angle (6) is given by
0 = tan" 1 '
First and second derivatives are commonly determined for use in this algorithm
by using divided differences (see Chapter 2). This is done by substituting the
voltage v and A* for the variables/and h, respectively, into eqns. 2.33 and 2.34,
in which case the following relationships are used:
1
( 4 3 6 )
**-i) *
1
^*+i-2» A + ^_i) (4.37)
where At is the sampling interval, and k + 1, k and k — 1 are subscripts referring
to a set of consecutive samples.
Sinusoidal-wave-based algorithms 67
From eqns. 4.36 and 4.37, it can be seen that the main advantage of this
method is that a constant DC component has no effect since it is cancelled out
from all difference expressions. The effect of the non-zero frequency components
due to any exponentially decaying offsets is consequently minimised, and no
mimic burden or related technique is required in many applications. However,
the disadvantage associated with this algorithm is that higher-frequency
components, due in particular to fault-induced travelling waves, can produce
significant errors in the first- and second-difference equations, thereby causing
large errors in successive impedance estimates. A solution to this problem,
which is successful in many applications, is to use the smoothed form of sample
and first derivative version of the algorithm for processing the voltage signal
and the first and second derivative technique for the current. In this way,
maximum overall "noise" rejection can be obtained, thereby improving the
reliability of successive impedance estimates.
Let vh vk+l be voltage samples measured at times th tk+l respectively, and let At
be the sampling time interval. Then
vk=Vsin a)otk (4.38)
vk+1 = V sin 0)otk+, = V sin a)0(tk + At)
vk+l=Vsin a)otk cos (o0At + Vcos a)otk sin a)0At (4.39)
Substituting eqn. 4.38 into 4.39 and simplifying results in
vk+, — vk cos coQAt
: = V cos (o<fk (4.40)
v ;
sin Q)0At *
2 2
Adding the squares of eqns. 4.38 and 4.40, and noting that sin x + cos x = 1,
we obtain an equation for the square of the peak voltage:
v\ 4- v\+, — 2vkvk+1 cos Q)0At
(sin cooAt)2
The corresponding equation for the peak of the current is simply:
/2 _ 1i + ^1+
illI ~ 244+1
i l l : cos (OnAt
1!_ (A ±n\
(sin
68 Sinusoidal-wave-based algorithms
It can be seen that the values O)0At are fixed for any given nominal system
angular frequency and sampling interval At. They therefore appear as con-
stants in the evaluation of the peak values, which, in effect, simply weight the
sampled values used in the algorithm.
The above equation shows that the phase angle 0 can be determined from a pair
of voltage samples and current samples. Despite the apparent complexity of its
form, the above equation is in fact relatively easy to evaluate, given that, as part
of the overall determination of the measured impedance, the peak values of the
voltage and current measurands (V and I) must be evaluated anyway.
(4-57)
The reactive part (Xf) of the measured impedance is given by X/= (— F/7) sin 0,
and can likewise be readily determined from a three sample voltage and current
measurand set. This is done by means of eqns. 4.54 and 4.55, which leads to:
(4.58)
AC
inputs A/D SPM P2000
signal A/Dsub inter-
, > conditioning system face <
k >
PSU
power >
supply
type- >
unit
writer
computer
sub-
PC system
programmers >
console
CO
contact
outputs
circuit- trip
breaker < relay
trip CKT
Figure 4.2 Schematic diagram for the basic components of the Prodar 70 protection system
(quoted from Reference 3)
4.7 References
1 MANN, B J., and MORRISON, I.F.: 'Digital calculation of impedance for trans-
mission line protection', IEEE Trans. 1971, PAS-90, pp. 270-279
2 MANN, B.J., and MORRISON, I.F.: 'Relaying a three phase transmission line with a
digital computer', ibid, 1971, pp. 742-750
Sinusoidal-wave-based algorithms 11
3 GILCHRIST, G.B., ROCKEFELLER, G.D., and UDREN, E.A.: 'High speed
distance relaying using a digital computer, part I—system description', IEEE Trans.
1972, PAS-91, pp. 1235-1243
4 ROCKEFELLER, G.D., and UDREN, E.A.: 'High-speed distance relaying using a
digital computer, part II—test results', ibid, 1972, PAS-91, pp. 1244-1252
5 MAKINO, J., and MIKI, Y.: 'Study of operating principles and digital filters for
protective relays with digital computers', IEEE PES Winter Power Meeting, New
York, January 1975, 75CHO990-2 PWR, Paper C75, 197-9, pp. 1-8
6 GILBERT, J.G., and SHOVLIN, R J.: 'High speed transmission line fault impedance
calculation using a dedicated computer', IEEE Trans. 1975, PAS-94, pp. 872-883
Chapter 5
Fourier analysis and Walsh function
based techniques
5.1 Introduction
In this Chapter we present digital relaying algorithms which are based on
Fourier and Walsh analyses. From the point of view of this discussion, Fourier
analysis includes Fourier series and Fourier transform-based methods, while
Walsh analysis includes Walsh series only.
The basic assumption used in Fourier and Walsh series based methods is that
the waveform that results from a fault condition (voltage and/or current) is
assumed to be periodic within the interval, say, from t0 to t0 + T, where T is the
period of the fundamental component. This assumption enables the waveform
to be expanded by either Fourier or Walsh series. The fundamental component
is then extracted and then used to calculate either the impedance to the fault or
differential current quantities.
In the case of the Fourier transform method, no assumption as to the nature
of faulted waveform is necessary. Both the voltage and current waveforms
within the data window are transformed to the frequency domain. These
transformed quantities are then used to calculate the apparent impedance to the
fault.
2 .
v(t) sin na)ot dt, n = 1, 2, . . . (5.3)
where a)0 is the angular frequency of the fundamental component and T is its
period.
Eqns. 5.2 and 5.3 show that the fundamental component of a voltage and/or
current waveform can be extracted from the corresponding faulted waveform
simply by setting n = 1.
Now let N be the number of samples per cycle of the fundamental component,
At the sampling time interval, tj=jAt the time of thejth sample and T=NAt
the period of the fundamental component.
The integral that appears in eqn. 5.4 can then be evaluated using the
rectangular method, which leads to
c o s l 4 c o s
^ * • ' ' v^j)
2 ^ /2nj\ 2 "
Vx^% Vj cos I —)=^WxJVj (5.5)
7=0 \ / j=o
where »,-= »((;) IS thejth sample of the voltage waveform, and Wxj is the
weighting factor of thejth sample used to calculate Vn which in turn is given by
2n
WxJ = cos (o0tj = cos — j At
sin cot —
multiplier averager
sampler
cos ©t -
where Wvj is the weighting factor of the jth sample used to calculate Fv, as
defined in eqn. 5.8
WyJ = sin(2Jtj/N), 7 = 0, 1, . . . JV (5.8)
The algorithm described by eqns. 5.5 to 5.8 can also be used to calculate the
real and imaginary parts Ix and ly of the fundamental current component from
current waveform samples, and Figure 5.1 shows a schematic representation of
the algorithm.
The weighting factors defined by eqns. 5.6 and 5.8 can be made to have
values of 0, ± 1 , ± 1 / 2 , and ± V 3 / 2 , which in turn are suitable for computer
application, if the sampling rate is suitably chosen. For example, h*W= 12, then
the values of Wxj and WyJ are as given in Table 5.1. This technique reduces the
Table 5.1 Weighting factors for iV= 12 samples per fundamental cycle
i /
1 0, 12 0 0, 6, 12
V3/2 1, 11 1/2 1,5
1/2 2, 10 V3/2 2.4
0 3, 9 1 3
-V3/2 4,8 -1/2 7, 11
-1/2 5.7 -V3/2 8, 10
-1 6 -1 9
Fourier analysis and Walsh function based techniques 75
output
VJJ(V-V+L{V-
V3 ]
^10 ) I 3.10)
Let Fx 1/2 and Vyt in be the real and imaginary parts of the phasor that
represents the fundamental component derived from a half-cycle window.
Thus, by applying eqns. 5.2 and 5.3 to a half-cycle window, we obtain
po+772
I *o '*
sin a)ot dt (5.12)
(772) I v{t) si
4 N/2
7=1
If N= 12, then the waveform coefficients can be written in the simplified form
If V3
V3 1
(5.15)
i r i V3 i
^,,/2 = 3 ^3 + 2 ("i + " s ) + — ("J + PO (5.16)
By comparing eqns. 5.5 and 5.7 with eqns. 5.13 and 5.14, or eqns. 5.9 and 5.10
with 5.15 and 5.16, it can be seen that the half-cycle window algorithm uses
only half the number of samples of those used in the full-cycle algorithm.
However, although this algorithm appears to be faster than that of the
alternative it has the disadvantage of introducing error, specifically due to any
aperiodic component and/or even harmonics. This can be seen in Figure 5.3
which represents the frequency response of the half-cycle-window algorithm.
The useful property of a rejection of odd harmonics is however retained.
output
input 1.0
0.75
0.5
0.25
3
CO
v(t)=Voc-"'o+Vl (5.18)
where
/0, Vo are the amplitudes of any decaying DC current and voltage components
I\y Vx are the amplitudes of the fundamental components of current and
voltage
Ip, Vp are the amplitudes of any assumed decaying current and voltage
oscillations
r0, Xp are the decaying time constants for any DC and high frequency
oscillations and
O)0, (Op are the frequencies of the fundamental and any high-frequency oscilla-
tions.
Since the data processed falls within a window (of length 7^,) it is convenient to
take the middle of the window as a reference. This can be achieved by shifting
the time scale / t o a new variable /', as shown in Figure 5.4, such that
(5.19)
where
If V'x and Vy are the quantities resulting from the correlation of the shifted
78 Fourier analysis and Walsh function based techniques
I*
2
Figure 5.4 Shifting of the time scale t to the variable t' where t'^t—fh +Tw/2).
(a) = original waveform, (b) = shifted version
* purely sinusoidal waveform is assumed
waveform v(t') with cosine and sine functions having a period equal to the time
span of data window Tw, then
where
The amplitude Vx and phase angle 6VI of the fundamental component of v(t)
can be determined from V'x9 V'y by using the following equations:
(5.22)
The factors k and p depend on data window length Tm and they can be
determined as follows:
r sin nr
(5.23)
where r = a)Ja)0.
Although this algorithm reduces the error due to any aperiodic components,
it causes the error due to high-frequency oscillations to increase. From a
practical point of view, this algorithm is therefore most suitable for applications
where oscillatory components in the currents and voltages are relatively small,
e.g. in distribution system applications.
5.2.3 Fourier-transform-based algorithm [7]
Basic approach
This technique does not impose any assumption as to the nature of the relaying
waveform within any chosen window of information analysed. The basic
approach involves applying the Fourier transform to a data window within the
voltage and current waveforms. The impedance is then calculated from the
voltage and current transforms. By progressively advancing the window, it is
possible to observe the impedance variation with time after a fault, which in
turn is used to determine whether there is a fault within the protected zone.
> • time
A Vi a
/ ^
y\ f V\
time
\ \
1 Ti w \
fault
inception V
v
Figure 5.5 (a) Faulted voltage and current waveform (b) Decomposition of voltage
waveform v into three components vi9 z*j, v3
t>T2
(5.27)
1 t<T2
According to the principles of circuit theory, each of the voltages »,, v2 and vz
would produce a corresponding current according to the following equation:
k=l, 2, 3 (5.28)
(5.29)
Fourier analysis and Walsh function based techniques 81
Now the Fourier transform of the window of voltage and current, v2(t) and i2(t),
can be calculated using eqn. 2.73 to obtain
»=["
v2(ja>) =
J -oo
v2(t)e-jm'dt (5.30)
The limits defined in eqn. 5.26 allow the Fourier transform of the given window
to be reduced:
1 = 1 v(t)e-ja)tdt (5.31)
J T
The Fourier transform of the corresponding current i2(t) is given by
-L i2(t)e-J°"dt
It will be noted that v2(j(o) can be easily evaluated using numerical methods
(5.32)
j (5.35)
By substituting this equation into eqn. 5.34, we obtain
h(t) = M-i(Tl)cru-T>)'t+/„(!) 7-,«.'<7i (5.36)
It must be emphasised that this equation is valid only during the window
bounded by 7*, =^ / ^ T2.
82 Fourier analysis and Walsh function based techniques
= i2(t)e-j(otdt= I 2 i2(t)e'jatdt+ I i
J TX J r, J T2
or
fr2
= [i(0 - KTX )e" ( / - r ' ) / r ] c-^dt
J Ti
which results in J T2
l T
^ ^ (5.40)
where
= i{t)z-jwtdt (5.41)
Computational aspects
Since voltage and current waveforms are presented to the algorithm in the form
of samples, the integration terms that appear in v2(ja)) and i2{j(o) should
therefore be evaluated numerically.
Let At be the sampling interval. Then
T2=TX + TW
= (N+n)At
Fourier analysis and Walsh function based techniques 83
where n is an integer which defines the position of the data window, and Nis the
number of sampling intervals within the data window of length Tw.
By substituting Tu Tw and T2 (defined in eqn. 5.42) into eqn. 5.31, we obtain
= v(t)e*"4t
Jn
where v2n(jco) is the transform of v2(t) for a window starting at Tx = n&t. When
the trapezoidal rule is applied to this equation, we obtain
1
•v[(n + N-
or
_ JV— 1 -i
where vn is the nth voltage sample, and Wn+m is the weighting factor of the
(#-fm)th voltage sample.
When the window is moved by one sample, the transform of the voltage v2(t)
within the new window, i.e. v2(n+^(ja)), can be found recursively by simply
dropping the oldest sample vn and adding the newest sample vn+N+\9 such that
where
At
Q«= —7T [ - W > « - W^+i»«+i + Wm+NOH+N+ Wn+N+lvn+N+l]
Fourier kernels. The previous method can be used to evaluate the current
transform i2(Ja)) defined by eqns. 5.40 and 5.41, which in turn results in
A*r ^ 1
where Mn= Wn+Nin+N~ Wjn, in = nth current sample, i2n(Jo)) = the transform of
i2(t) relating to the window which starts at T—
with
1 [T
FO=T\ v(t)dt
Jo
V2 fT
F2n-X=— v(t) sin na)0t dt (5.47)
Jo
Fourier analysis and Walsh function based techniques 85
where F o is the DC component and F2n, F2n+1 are the RMS values of the real and
imaginary components of the nth harmonic, respectively. As indicated previous-
ly, it is the Fourier series component values at fundamental (or power)
frequency that are used to determine the measured impedance. However, it is
relatively easy to determine these by first evaluating the Walsh component
values. In this respect, it will be recalled that by using the Walsh series, v(t) can
be approximated using eqns. 2.85 to obtain
^) (5.48)
where
At
86 Fourier analysis and Walsh function based techniques
When this is done, the Walsh coefficients Wk can be calculated according to eqn.
5.49, such that
The integral on the right hand side of the above equation can usefully be
evaluated using the trapezoidal rule, which results in
Wal(A,
1
+ v(tN-}) • Wal(*, (N-\)Atf)-¥-v(tN) • Wal(t, JVA*') A* (5.51)
Now let v(tj) = vj, wj(k) = Wal(A, jA/') and fj = v(tj) • Wal(*JA/') = »,• • i^.(A), so
that by substituting into eqn. 5.51 we obtain
(5.52)
=]faf,+ji++, , + • . .J+1
Wt{s) =]faf,+ji +1 + • • . J|-1+
1+,, +^
+^ jf+, | (5.54)
If for example N=8, then either with the help of Figure 2.6 or by using the
discrete representation of the Walsh function given in Section 2.7.3, it can be
seen that
in i i
o(s) = - - * , + * , + , + . . . v7+s + - vB+s (5.55)
Fourier analysis and Walsh function based techniques 87
Wal (1 + f)
vft + 1)
f A
xDF
Figure 5.6 The mechanism of movementfrom right to left of a waveform through a window
that contains W a l ( l , t)
(a) Walsh function window which contains Wal(l, t')
(b) Sampled sinusoidal waveform Vj = v(tj) = v(jAt)
(c) Resulting function f(jM) = v(tj) • Wal(l, /')
(d) The function v(tj+l) = v((j+\)At) resulting from v(tj) after being
moved to the left by one sampling interval
(e) The resulting^, = / ( ( / ' + 1)A/) = »((/+ l)A*) ' W a l ( l , t')
88 Fourier analysis and Walsh function based techniques
Therefore W0(s + 1) can be found recursively by dropping off the oldest sample
and adding the newest sample v9+s to obtain
ir i i i i
-\ --v,--o1+, + -vli+, + -v9+,
J = 0, 1,2, . . . (5.57)
From a computational point of view, not much is gained from calculating W2
recursively since it can be calculated directly as
lfl 1 1
W2(s)=~\-vs + vl+s--v3+s--vt+s-v5+s + v7+s + ~vs+s\ (5.58)
The remaining coefficients can be found in exactly the same way and are given
in eqn. 5.59 below.
in i
in I i
- 1 - vs. - v2+1 + o 4+J + - vs+s
iri i I
(5.59)
j j b , - V2+, + v6+1 - - vs+I
lfl 1 "1
- " ! , - Vi+S + ~ VS+S
lfl
It will be noted that, except for the end-points of the interval (0, 7), the
application of the trapezoidal rule leads to the cancellation of the sample value
where a jump occurs in the appropriate Walsh function. This is because the
area of the function f(tj) = v(tj) • Wal(A:, /j) around any such jump has equal
parts but with opposite sign (see Figure 5.6*).
=2 (5.62)
then
(i) to an accuracy of (+4.28% to —4.8%) the maximum value of g{{t)
(GM) is given as Gmax= 1.0822(|W,| + \W2\) + 0AU\\Wt\ -\W2\\ (5.63)
(ii) the maximum value ofgi(/) occurs in the mth sub-interval, within which the
following formulas are satisfied:
/ \
Wall*,-I (5.65)
then
(ii) the maximum occurs in the mth sub-interval which satisfies the following:
Wai(2, mAl)=S(W2)
• 5(A)
Wal(10, mM')=
[S(W.2)
where
To avoid any confusion that may arise between the Walsh coefficients
ftfk (£ = 0, 1, . . .) for the whole voltage waveform and those that relate only to
the fundamental component of that waveform, we shall denote the Walsh
coefficient due to the fundamental by Wxk (k = 0, 1, . . .). It follows from eqn.
5.62 that the fundamental component can be expressed as
(5.67)
However, the coefficients Wn> v and Wxx v c a n D e found from their corresponding
Fourier coefficients, using the inverse of eqn. 5.50 as
W
"-~0*F- ,5.68)
^ 09F
This is so because F t = 0 for all i except i = 1 and i = 2 (see eqn. 5.46). On the
other hand, Flv and F2v can be expressed in terms of the Walsh coefficients due
to the whole faulted waveform. Assuming a 10 X 10 A matrix, Fu and F2v can be
found according to eqn. 5.60. By substituting the latter equations into eqn. 5.68,
we obtain
It will be noted that Wn^v and PVl2v are equal to Wx and W2 stated in the
amplitude and phase theorems.
5.4 References
1 SLEMON, G.R., ROBERTSON, S.D.T., and RAMAMOORTY, M.: 'High speed
protection of power systems based on improved power systems models', CIGRE,
Paris, June 1986, Paper 31-09
2 RAMAMOORTY, M.: 'Application of digital computers to power system protection',
J. InsL Eng. India, 1972, 52, pp. 235-238
3 McLAREN, P.C, and REDFERN, M.A.: 'Fourier-series techniques applied to
distance protection', Proc. 1EE, 1975, 122, pp. 1301-1305
92 Fourier analysis and Walsh function based techniques
4 PHADKE, A.G., HIBKA, T., and IBRAHIM, M.A.: 'A digital computer system for
EHV substations: analysis and field test', IEEE Trans. 1976, PAS-95, pp. 291-301
5 WISZNIEWSKI, A.: 'Signal recognition in protective relaying' in 'Developments in
power system protection,' IEE Conf. Publ. 185, 1980, pp. 132-136
6 WISZNIEWSKI, A.: 'How to reduce errors of distance fault locating algorithms',
IEEE Trans., 1981, PAS-100, pp. 4815-4820
7 JOHNS, A.T., and MARTIN, M.A.: 'Fundamental digital approach to the distance
protection of EHV transmission lines', Proc. IEE 1978, 125, pp. 377-384
8 DESOER, C.A., and KUH, E.S.: 'Basic circuit theory', (McGraw-Hill, 1969)
9 HORTON, J.W.: 'The use of Walsh functions for high-speed digital relaying', IEEE
PES Summer Meeting, 1975, July 20-25, Paper A 75 582-7
Chapter 6
Least squares based methods
6.1 Introduction
In this Chapter we shall discuss techniques used to fit faulted current and
voltage waveforms, each to a sinusoidal waveform containing a fundamental
component, a decaying/constant DC component and/or harmonics.
These techniques use the least squares (LSQ) method to minimise the fitting
error, and all have the common goal of extracting the fundamental components
of voltage and current waveforms, to calculate the impedance to the fault or the
comparison of current-based signals in digital differential protection.
ma)ot)fdt (6.3)
J
94 Least squares based methods
As discussed in Chapter 2, the best fit occurs when S is a minimum, and this is
satisfied if
dS
— =0 (i=\,2,...,2M+\) (6.4)
dS
dS f
~dK.- =0 = - 2 [v{t)-Ki e'"T
•m J (6.5)
cos m+, sin ma)ot)] cos m(O0t dt
dS f
=0=-2 [v{t)-Kx^
&2m+l J
The limits of integration in the previous equations are taken from /1 to t{ + !F(i.e.
over one fundamental period). After simplification, eqn. 6.5 then reduces to
f/, + 7- ftt + T
K{ e" 2 / / V/= v(t) etlxdt
J /, J /,
2 f'.+T-
K2m = j , I »(0 cos mft>0^ dt (6.6)
J t\
2 f'l+r
K2m+t= -j, I »(0 sin ma)ot dt
J i.
These equations have been derived from eqn. 6.5 using the following identities:
V2 1
cos
eax cos mx dx = — 5 ia mx — m sin mx]eax]x\+2 = 0
a -¥m
1
where
and WXn is the weighting factor of the ith sample used to calculate Ku which is
equal to e'^/I.
If the trapezoidal method is applied to the second eqn. 6.6, we obtain
fl (2n\ (2n\
b f t ) cos ml YJ t\ + v(h) cos ml — I / 2 +•
/2n\ 1 /2JT\ 1
cos m + V C S m 1N
*JV-I) \~f) ^-i o ^ ° \~Tl
or N l
i r ~ i
L «=2 J
where
W2my n ~ weighting factor for nth sample for calculating K2m
(2n\ (2n{n-\)^
= cos m
This algorithm assumes that the current and voltage waveforms contain only
constant DC and fundamental components. Taking the voltage signal v(t) as an
example, then v(t) is assumed to take the form
(6.10)
When these times are transformed to the new reference frame t*, using eqns.
6.11 and 6.12, the sequence 6.13 becomes
N- \\ (N-1
0 >
{ ) A ' • • • ' ~ A t
' • •'
)(J (6.14)
Least squares based methods 97
A 2A
This sequence shows that the sampling time in the new reference t*, is
independent of time. Figure 6.1 explains the meaning of the new time reference
in association with an assumed sinusoidal waveform.
By combining eqns. 6.10 and 6.11, we then obtain
sin[wo{t* +
or
v(t) = Kl + K2 sin(a)ot* (6.15)
where
sin
3!
and
(«>ot*)2
cos a)ot*
2!
Therefore the first equation can be expressed as
v * = a0 + axx + a2x2 (6.16)
98 Least squares based methods
where
x = coQt*
ao = K\ + K2 sin #! = K2 COS /?*
a3=(-K2 cos
Eqn. 6.16 expresses the sampled value in terms of ^-coefficients. In other words,
the monitored waveform, which is assumed to consist of a constant term and a
fundamental component, is fitted by a power series of third order. The
amplitude K2 and phase angle (i of the fundamental component can be
expressed in terms of coefficients ax and % as follows:
(6.17)
= tan
(6.18)
According to eqn. 2.39, the summation of the square of the error S is a minimum
if
dS
When this is applied to eqn. 6.18, we obtain the following set of equations:
Note that the range of summation in eqn. 6.19 is from 1 to N and vh *,- are
evaluated at times corresponding to *,• = / — (N—i)At.
Note also that only ax and a2 are required to determine the amplitude and
phase angle of the fundamental component as per eqn. 6.17. It is, therefore,
sufficient to solve eqn. 6.19 only for a{ and a2. Furthermore, the computational
burden can be greatly reduced if the data are shifted to the new reference frame
t* (which causes x{ to remain constant) and N is chosen to be an odd number.
This results in all sums involving odd powers of x becoming zero. Therefore, by
manipulating eqn. 6.19, the constants a{ and a2 can then be expressed as follows:
a, = M2
(6.20)
a2:== J
where
M{=- I-"
(y A2-
The constants Aft, f = 1, . . . , 4 can be determined beforehand once the number
of points N, and the sampling interval, have been selected. It should be noted
that the limit of the summation is from £ = 1 to i = N.
100 Least squares based methods
The exponential term €~ tl% of eqn. 6.21 can be expanded using Taylor's series,
such that
2 3
t I t I t
e //r==1 + (6 22)
" -;^?"ii? "- -
By considering only the first three terms of this expression, eqn. 6.21 can be
expressed as
/ t2
v(t) = *, - kx - + *i —$ + *2i s
or
t t2
v(i)=k{—kx - + A;, — 5 + (k2X cos 0,) sin cwo/-f (k2X sin 0i) cos a)Qt
s m cos
4- (k23 cos 0 3 ) sin 3a)ot+(k23 #3) 3tt>0/ (6.23)
v(tx) = A;, — kx — 4- A:, r~^4- (A:2! cos 0,) sin a>0/, 4- (k2x sin 0|) cos
+ {k2i cos 03) sin 3ft>o/| + (k23 sin 83) cos 3<w0/
Least squares based methods 101
where Sx is the sample measured at time tx. The coefficients in eqn. 6.24 are
related only to the time at which the samples are taken, and they take the form
aX5 =
After determining the unknown vector X from eqn. 6.27, the magnitude of the
fundamental component of voltage K2 and its phase 0x can be found in terms of
x2 and x3 (see eqns. 6.23, 6.24).
102 Least squares based methods
Z (6.28)
Now K2y, K2/ are the real parts of voltage and current fundamental components,
respectively, and are determined using eqn. 6.8 (with m= 1). Likewise, K3V&nd
K3I are the imaginary parts of voltage and current fundamental components,
respectively, which are determined by using eqn. 6.9 (with m— 1).
The measured impedance can alternatively be determined using the power
series LSQ fit by means of eqn. 6.29:
K2VL (iV
Z=-— r (6.29)
In the case of the power series, K2v, K21 a r e the amplitude of the voltage and
current fundamental components, respectively, and j3v, /?/ are the phase of the
voltage and current fundamental components, respectively. The latter are
determined by eqn. 6.17.
Finally, a measured impedance can be found using the multivariable series as
{ \ j [ \ j w j w
( }
Re[/]+jIm[/] Xy+jx3l '
The values x2F, x3V are the real and imaginary parts of the fundamental of the
voltage waveform and are determined according to eqn. 6.27, together with the
corresponding values derived for the current (;%, x3f).
6.6 References
1 LUCKETT, R.G., MUNDAY, P.J., and MURRAY, B.E.: 'A substation based
computer for control and protection' Developments in power system protection, IEE
Conf. Publ. 125, London, March 1975, pp. 252-260
2 BROOKS, A.W., Jr: 'Distance relaying using least-squares estimates of voltage,
current and impedance', Proc. IEEE PICA Conf, 77CH 1131-2, PWR, May 1977,
pp. 394-402
3 SACHDEV, M.S., and BARIBEAU, M.A.: 'A new algorithm for digital impedance
relays', IEEE Trans., 1979, PAS-98, pp. 2232-2240
Chapter 7
Differential equation based techniques
7.1 Introduction
In this Chapter no special assumption will be made with regard to the content
of faulted current and voltage waveforms. The fundamental approach, which is
common to all algorithms covered in this Chapter, is based on the fact that all
protected equipment can be normally represented by differential equations of
either first or second order. The methods are described by reference to
transmission-line protection, since it is in this application that they are mainly
used. However, the methods can easily be extended to other items of plant. For
the purposes of this Chapter we shall assume the line length is such that shunt
capacitance can either be neglected or can be lumped into a single 'equivalent'
value.
-czb 'trnv-
Figure 7.1 Elemental length of distributed transmission line represented by series resistance
and self and mutual inductances
104 Differential equation based techniques
inductance per unit length between kth and /th phases and ik9 vk are the current
and voltage of kth. phase, then the voltage drop across an infinitesimal length dx
of phase 'a' is:
dvc dia
where Rs, Ls are the series resistance and self inductance per unit length of each
phase, and Lm is the mutual inductance per unit length between any two phases.
By substituting eqn. 7.4 into eqns. 7.1-7.3, we obtain:
dva I d\ dib dic
s s %a m m
~dx~\ dt) Yt^ dt
Now the self and mutual parameters are related to the zero and positive phase
sequence parameters as follows:
(7.6)
and
Differential equation based techniques 105
where Rl9 Lx are positive phase sequence resistance and inductance, respecti-
vely, Lo is the zero sequence inductance and i0 is the zero sequence current.
By combining eqns. 7.5 and 7.6, we then obtain
di0
at
dvb I d\ di0
X
'+(LQ-Ll)— (7.7)
Eqns. 7.1-7.3, 7.5 and 7.7 can be used to calculate the voltage drop between the
fault point and the relay location for different fault types; each of these will be
considered in the following subsections.
or
^ (7.8)
at
where
and
106 Differential equation based techniques
Equations relating to transposed lines can be expressed in the same way, and in
this case eqns. 7.5 are used to obtain the relationship given in eqn. 7.9
diy
va = xRsix + xLs -j (7.9)
where
and
i, = i.+ (LJL,)it+(Lm/L,)it
The alternative sequence component formulation of eqns. (7.7) likewise leads to
div
va = xRjx + xL\jt (7.10)
where
and
dt)
This equation can be reduced to the more succinct form
di
Va~Vb = xRJx + x(La — Lab ) y
-vb = xRaix + x{La-Lab)^- (7.11)
where at
and
{Lb-Lab)ib
When the line is assumed ideally transposed, eqn. 7.10 can be written in terms
of the difference between the currents on each faulted phase, giving
diy
va-vb = xRlix + xLl y (7.12)
The above analysis shows that the behaviour of the transmission line under
fault conditions is governed by a differential equation having the general form of
eqn. 7.13. The measured values of the currents and voltages are taken in the
form of samples and the measurement is usually done simultaneously on all
three phases by using suitable sample-and-hold peripheral equipment as
described previously.
div
v = Rix + L-f (7.13)
at
The values of/? and L in eqn. 7.13 can be calculated by several methods,
described in more detail below.
R | \ y / + L(*
* vv22 -*yl
y l)= | vdt
vdt (7.14)
R | \ y / + L (ffyy44--**yy33 ) = rr ^
^ (7.15)
J ti
It will be evident that the values of measured resistance R and inductance L can
be determined by solving these two equations. However, due to the presence of
harmonics, particularly in the current waveforms during the first cycle after the
occurrence of the fault, it has been found that the values of R and L calculated
using this simple approach are very sensitive to low-frequency harmonics. The
accuracy of calculated values of R and L can, however, be greatly improved if
the limits of integration are so chosen as to eliminate (or filter out) the
unwanted harmonics.
If the highest harmonic contained in the waveform i(t) is N, the last equation
would be reduced to
2 (7.17)
where
and
Now let us integrate eqn. 7.17 from ^ = 0 to t2 = a/co0, and let us call this
integration Ix so that:
c
/, = i(t) dt= cQdt^2j m, cos(ma)^+6m)dt
J /,=o
/, = 0 Jo m=\ Jo
This equation shows that the integration of i(t) with respect to time is equal to
the summation of the integration of its individual harmonic components.
Therefore, let us concentrate our attention on the integration of the mth
harmonic component and its multiples. If InX is the integration of the rath
harmonic over the period from tx = Q to t2 = a/(t)0, then
fa/(o0
cn cos(na)0t+dn)dt
J oo
which gives
J n
[( B) 8n] (7.19)
na)0
It will be evident from eqns. 7.18 and 7.19 that the sum /„, + /n2 = 0. This is of
course true for all harmonics (n). In essence, this means that any rath harmonic
and its multiples can be filtered out from the waveform i(t), by simply adding
two integrals taken once over the limits /, = 0 to t2 = a/a)0 and t3= (7i/n)/a)0 to
Differential equation based techniques 109
= /, cos((o0t
When i(t) is integrated from ^ = 0 to t2z=a/a)0 and from J3= (jt/3)/(O0 to
tA= (jr/3 + a)/ft>0 the corresponding integrations of the third harmonic compo-
nents are equal to the areas shown shaded in Figure 7.2. These areas are equal
to each other but have opposite signs and, when added together, they are
cancelled and therefore effectively filtered out from the original waveform i(t).
Figure 7.2 Physical interpretation of digital filtering by integration over selected limits
110 Differential equation based techniques
However, the number of integrations can be reduced to only two if the value of
a is chosen to be equal to the angle corresponding to a full cycle of the mth
harmonic order (i.e. a = 27t/m). In this way we can ensure the elimination of the
effect of the mth harmonic by the first integration, as the integration over a full
cycle of a sinusoid is always equal to zero. It is then only required to eliminate
the effect of the nth harmonic, which can be done by addition of a second
integration. Thus, for removing two harmonics of order n and m together with
any multiples thereof, the following equation would be used:
C2n/m(i)Q C{nln + 27i/m)/(x)Q
i{t)dt+ i(t)dt (7.21)
Jo J n/nojQ
i—"ums^
v(t) 1 v(t)
R L
(7.25)
2At
Similarly, by using the following sets of samples at tk+l, eqn. 7.23 becomes
+ 2 ~ lk
(7.26)
2At
In matrix form, eqns. 7.25 and 7.26 can be combined to give
h+1 h- I
h R Vk
2A/
(7.27)
*'*+2~~1'*
ik+i L Vk+\
2At \
It is evident that this equation can be written in the following short form:
AP = V (7.28)
where
l
**+l k-\
h 2A/
A=
4+i
2A/ _
112 Differential equation based techniques
(7.30)
2 h) h i h-\)
It must be remembered that this algorithm produces a succession of estimates of
R and L as the sampled data is acquired, and these estimates can be compared
for convergence on any post fault values.
(7.31)
The current ic is the capacitive current flowing through the lumped equivalent
capacitance at the relay location, and it is related simply to the voltage at the
relaying point by noting that ic=Cdv/dt. Eqn. 7.31 thus takes a second-order
form:
di do Sv
Ri + L RC---LC-^ (7.32)
dt dt dt2
H I—^MS^-
To determine R and L from eqn. 7.32 it is necessary to use at least four sets of
voltage and current samples. Therefore, let us assume that at / = tk we have v = vk
and i = i*. Also, by using central finite differences, the current and voltage
derivatives can be determined from the measured samples using eqns. 2.33 and
2.34 as follows:
di
It 2A/
(7.33)
dv
It 2A/
i
(7.34)
dt2 (A/)2
Substituting current and voltage samples at t = tk and their corresponding
derivatives described by eqns. 7.33 and 7.34, we obtain
Rik+l + L RC -LC
2&t 2A< (At)2
Rh+2 -RC — LC
2A/
R
2A/ (A/)2
2to L
2At (A/)2
l
k+3~~h+\
CR
2A/ 2A* (AO 2
CL
2A/ 2A/ (At)2
(7.35)
In short form, eqn. 7.35 becomes:
A
(7.36)
_C DA LCXJ LV2J
114 Differential equation based techniques
Since the vectors X and CX are linearly related, eqn. 7.36 can be reduced to
{A -BBrxC)X= {VI -BDlV2) (7.37)
which can be written in a short form
FX = G (7.38)
The vector X in eqn. 7.38 is readily evaluated by forming the matrices F and G
from sampled values. Again this process can be repeated after each sample set
becomes available, thus producing a succession of estimates of resistance and
inductance between the relaying and fault points.
7.5 References
1 RANJBAR, A.M., and CORY, BJ.: 'Algorithm for distance protection'
Developments in power system protection, IEE Conf. Publ. 125, London, March
1975, pp. 276-283
2 RANJBAR, A.M., and CORY, B J. 'An improved method for the digital protection of
high voltage transmission lines', IEEE Trans. 1975, PAS-94, pp. 544—550
3 GILBERT, J.G., UDREN, E.A., and SACKIN, M.: 'Evaluation of algorithms for
computer relaying', IEEE PES Summer Meeting, 1977 Mexico City, Paper A 77 520-
0
4 SMOLINSKI, W.J.: 'An algorithm for digital impedance calculation using a single PI
section transmission line model, IEEE Trans., 1979, PAS-98, pp. 1546-1551
5 JEYASURAY, B , and SMOLINSKI, W.J.: 'Identification of a best algorithm for
digital distance protection of transmission lines', ibid., 1983, PAS-102, pp. 3358-3359
Chapter 8
Fundamentals of travelling-wave based
protection
8.1 Introduction
In previous chapters, digital algorithms were derived on the assumption that
the current and voltage waveforms are either sinusoids of a single frequency or
sinusoids consisting of decaying/constant DC, fundamental and/or harmonic
components. Many of these assumptions are valid and lead to an acceptable
performance in applications where the line length is limited. However, the
introduction of long EHV/UHV transmission lines and EHV cables into power
systems can produce difficult problems for protective relays. The principal
causes of these problems are:
(i) shunt leakage current due to the increase in shunt capacitance on long
lines and cables
(ii) relatively low frequency transient currents caused by surge travelling
waves following fault inception
(iii) transient currents, at frequencies as low as a few hundred Hertz, caused by
the interaction between the total inductance and capacitance of the system
(iv) a weakly damped transient DC current caused by high system reactance-
to-resistance ratio.
The foregoing considerations, together with a desire to reduce fault-clearance
times to improve the transient stability of electric power systems, has led to
much interest in the development of so-called travelling-wave protection.
In this Chapter, transmission lines are treated as distributed circuits, in order
to explain how travelling-wave phenomena are propagated and detected. The
basic principles of travelling-wave schemes, the formation of relaying signals,
Bergeron's equation and discriminant functions using single-phase line models
are also explained. These principles are then extended to three-phase lines by
decomposing the line into three equivalent single-phase lines using the modal
decomposition approach.
dva
R 'a
dvb
ib < 'b dib
w
dvc
k
1 dx
R a dx L a dx
«a ~ di a , di a = di a1 + d i ^ + di a3
di a iV di.
D
1Cadx
Va
G a dx
R
ib »dx ib - di b , di b = di b1 + di b2 + di^
G b dx
ic - di c , dic = di c1 + di c2 + di c
Gcdx C c dx
dx
unit length of phase k (where k = a, b and c) and Lkm, Ckm are the mutual
inductance and capacitance, respectively, per unit length between kth and mth
phases.
If the line is subjected to transient conditions such as faults, the associated
changes in the voltages and currents within an element dx of the line (which is a
distance x from the relay location) at an instant t are related by the following
equations:
For the 'a' phase:
dia
Cabdx Ccadx —
at at
(8.1)
H
Similar equations can be derived for phases 'b' and 'c':
For phase 'b':
(8.2)
(8.3)
118 Fundamentals of travelling-wave based protection
By replacing vah van vba, vbn vca, vch by their equivalent phase voltages (i.e.
v
ab = va~~vb> • • • , etc.), eqns. 8.1-8.3 can be written:
dv
~dx"
(8.4)
di
~d~x=yv
where
»«" i.
v = «= h
and
a a a '
Lca
L
'Jt L b
°m Jt
a a Lb
z— 'lt
a a a
*
a a Cca
a
~dt ~dt
a a a
Gb+Cbb-
It at
a a a
where
dv
T
ox (8.5)
di
—
dx
Fundamentals of travelling-wave based protection 119
where Z and Y are the same as z and y except that the time operator d/dt has
been replaced by the transform operator p. By combining eqns. 8.5 we finally
obtain
(86)
l?=Pv
dH
-2=PH (8.7)
where
PT=YZ
For the purpose of this and the following Chapters, the formulation of eqns. 8.6
and 8.7 is particularly useful for an appreciation of the important special cases
of single-phase lines and three-phase transposed lines.
where v and i are the voltage and line current, respectively, measured at any
point on the line and Z = /? + Lp, Y= G+ Cp. In this case the previously defined
line parameters R, G, L, C represent single-phase line parameters per unit
length.
When eqns. 8.8 are combined, they produce the scalar form of eqns. 8.6 and
8.7, such that
(8.9)
Vzy=^VZc=- = y and
c
<z [Z
v(x, I) = e-t'T, (1) + c+xp'cF2{t) (8.12)
i(x, t) = (<T*"Fi{t)-e+'*"F2{t))IZ* (8.13)
where
1
C=
VLC
is known as the velocity of propagation and
J
"\c
is known as the surge impedance.
It is possible, with the help of Taylor's theorem, to show that
v(x,t)=FL--)+F2(l + -) (8.14)
x, t) =
Fundamentals of travelling-wave based protection 121
Functions Fx{t — x/c) and F2(t + x/c) thus represent travelling waves in the
forward and backward directions of x, respectively, and eqns. 8.14 and 8.15 may
therefore be written in the simple terms
where v+ =Fl(t — x/c), and v~ =F2{t + x/c) are the forward and backward
voltage component, respectively. The values of i+ and i~ are similarly the
forward and backward components of current.
The physical meaning of the function having the form f(t ±x/c), can be well
understood if consideration is given to the behaviour of this function when the
argument is held constant, i.e. let t — x/c = ki for a forward travelling wave and
/ 4- x/c = k2 for a backward travelling wave. Therefore any change in time t, say
At, requires a change in x equal to Ax such that Ax = cAt for the forward
waveform and Ax = — cAt for the backward waveform. Such considerations
show that a forward waveform travels in the positive direction of xy while the
backward waveform travels in the negative direction of x.
The forward and backward components are related to each other, as seen
from eqns. 8.14 through 8.16, by the characteristic (or surge) impedance of the
line as follows:
v+ = Zoi+
- „ ._ (8-17)
V =—ZQI
It will be seen from the above analysis that the surge impedance (Zo) is a real
number for an assumed lossless line and it is evident from eqns. 8.17 that the
current components are simply a replica of their corresponding voltages. They
also show that, while the forward voltage and current waveforms are of the
same sign, the backward voltage and current waveforms are of opposite sign as
illustrated in Figure 8.2.
Coefficient of reflection
Waves travelling over assumed homogeneous lossless lengths of transmission
line continue to propagate at a uniform velocity c and are unchanged in shape.
However, at points of discontinuity, such as open circuits or other line
terminations, part of the incident wave is reflected back along the line and part
is transmitted into and beyond the discontinuity. The wave impinging on the
discontinuity is often called an incident wave and the two waves to which it
gives rise are normally referred to as reflected and transmitted waves.
Consider Figure 8.3, which shows a transmission line whose sending end is
connected to a source and its receiving end is connected to an impedance Zr.
When the switch S is closed, the voltage wave v travels along the line until it
reaches the discontinuous receiving-end terminal where only part of the wave
passes to the terminal impedance Zr (transmitted wave), and the rest is reflected
back. Let us call the first part v* and the second part v~. Therefore, at the
122 Fundamentals of travelling-wave based protection
receiving end the total voltage and current are equal to the sum of the forward
and backward components of voltage and current:
v = v* + v~
(8.18)
,- = ,•+ + ,-
We recall from the previous Section that the voltage wave and its associated
current are interrelated by the surge (or characteristic) impedance of the line:
(8.19)
v =-
z-z0 (8.20)
where
' zr+z0
7777777777777777777777/ 77777777777777777777777
/77/7///////7//77//77T/
7i
j
Figure 8.3 Successive wave reflection at both ends of a single-phase transmission line
(a) line circuit
(b) wave applied at the source and reflected at the load end
The constant Kr is often called the receiving end coefficient of reflection because
it describes the voltage reflected at a discontinuity in terms of the incident (or
forward wave component).
By combining eqns. 8.19 we similarly obtain
i~ v~ Zr — Z o
or
r =-^+ (8.21)
In a similar manner, the coefficient of reflection at the sending-end can be
derived as:
Z
" '~Z° (8.22)
differential equations with respect to voltages and currents. On the other hand
it is well known in power system analysis that phase quantities can be
decomposed into equivalent symmetrical components such that
where Vp, Ip are the phase voltage and phase current vectors, V^ Is are the
sequence voltage and sequence current vectors and T is the symmetrical
component transformation matrix defined below:
1
vb , v = and T =
1Z_12O°
are tne
K> Vfo K voltages of phases 'a', ' b ' and V , and F o , F,, F2 are the zero,
positive and negative sequence voltages.
It should be noted that, in accordance with the theory of symmetrical
components, phasor values are employed throughout in the above analysis.
It is possible likewise to transform the instantaneous (or time domain) phase
quantities into what are called modal components such that
vp(t)=Svm(t)
(8.23)
iP(t) = Q.L(t)
In this case, the voltage and current vectors vm(t), im(t) are modal component
vectors defined by
f(0)
(I) (t)
(0 I (<)
w. I w
It should be noted that the superscripts (0), (1), and (2) refer to the zero, first
and second modes. The matrices S and Q are the voltage and current
transformation matrices, which are used to transform the phase quantities into
modal components.
The elements of matrices S and Q can be determined using matrix function
theory and the properties of eigenvalues and eigenvectors. Details of the
analyses involved are beyond the scope of this book, but a brief outline of the
method is useful.
By substituting eqns. 8.23 into eqns. 8.6 and 8.7, we obtain
(8.24)
dx2
— = Q-ipTQim (3 25)
It is necessary to choose the matrices S and Q such that S~lPS and Q~lPTQ
become diagonal. This involves finding three scalar quantities (the eigenvalues)
Fundamentals of travelling-wave based protection 125
and a column matrix X (the eigenvectors) for eqns. 8.6 and 8.7. For example,
the eigenvalues and eigenvectors for eqn. 8.6 should satisfy the following
equation:
{P-U)X=0 (8.26)
1 1 1
1 0 -2 (8.27)
1 -1 1^
IF
(8.30)
IF
where y(i) is the modal propagation constant of the kth mode. The zero mode is
126 Fundamentals of travelling-wave based protection
also termed the earth mode, while the second and third modes are termed
'aerial modes'.
These equations show that wave propagation in a three-phase line can be
considered in terms of three separate and independent components, each with
its own propagation constant y{k) and associated surge impedance Z$\
Assuming complete transposition, the surge impedance associated with each
mode can be shown to be:
40> = ( J J ( f m ) J0
2>
4 " = 4 = V(z,-zm)/(Y,+ Ym) = VzJY,
where Z0(k) is the A>mode surge impedance, Zs, Ys are the average sum of all
conductor self impedances and admittances at any frequency, Zm, Ym are the
average sum of all conductor mutual impedances and admittances at any
frequency, Zo, Yo are the zero-phase sequence line impedance and admittance
and Z l5 Yl are the positive-phase sequence line impedance and admittance.
The modal propagation constants are similarly given in terms of phase
sequence impedances and admittances by
(8.32)
(8.33)
When a power system is subjected to a fault condition, the total voltage and
current at any point in the system can be considered as consisting of two
components: one is due to the sinusoidal steady-state condition, the other due to
the application of the fault. The latter component is often called the superim-
posed quantity and is simply equal to the change in the current and/or voltage
due to a disturbance. The idea of superimposed quantities is best explained by
reference to Figure 8.4. Consider point R, where the voltage vRf(t) and current
iRj(t) are the actual quantities present (see Figure 8.4a). These quantities can be
split into the sinusoidal prefault quantities vRS(t) and iRs(t) shown in Figure 8.4£,
and the fault superimposed quantities AvR(t) and AiR(t) shown in Figure 8.4c.
Thus
It should be noted that the sum of the steady-state and superimposed quantities
is equal to the total variation at ail points in the network. The superimposed
Fundamentals of travelling-wave based protection 12 7
network is therefore a network with all source voltages set at zero (short
circuited). The superimposed fault point voltage is therefore zero for ail time up
to the time of the fault disturbance. The voltage of the fictitious superimposed
source on the faulted phase(s) is thus equal in magnitude and opposite in sign to
the prefault voltage at the fault point (i.e. vF(t)).
The superimposed quantities of voltages and currents possess unique
properties in relation to the location of the fault, which will be explained in the
following subsections. These properties form the basis of practical travelling-
wave based protection schemes.
We shall show in this Section that the relative polarities of the superimposed
voltage and current waveforms (or the direction of the motion of their travelling
waves) at the two ends of the line under consideration depend on the fault
location. Consider Figure 8.5, from which it can be seen that the fault produces
superimposed quantities at the relaying points R and S. If an internal fault is
assumed to occur during the positive half cycle of the prefault voltage
waveform, then the signs of the superimposed voltage and current would be
( - ) and ( + ), respectively, at both ends of the line. (Note that reference
o i > ; <i o
S
(t) r
o-
9 is incidence angle
o-
polarity of
superimposed
type of fault corresponding superimposed
network polarity components
AiR AvR Ais Avs
pos.
internal fault
neg.
pos.
external fault
beyond bus R
neg.
pos.
external fault
beyond bus S
neg.
Figure 8.5 Dependency of polarities of superimposed voltage and current quantities on fault
location
directions are those shown in Figure 8.4). However, if the fault occurs during
the negative half cycle of the voltage waveform, the polarities of the superim-
posed voltage and current at the R and S ends would be ( + ) and (—)
respectively. Repeating the analysis for faults at locations external to the line
R—S reveals that the polarities of the superimposed voltage and current would
be similar to each other at one end but are different from each other at the other
end. Figure 8.5 explains in detail all possible cases.
Fundamentals of travelling-wave based protection 129
8.3.2 Interrelation between the superimposed voltage and current quantities versus
fault location [8, 9]
Consider Figure 8.6, which shows a line R-S interconnecting two systems R
and S. The following analysis and associated conclusions are general, but to
simplify the discussion we shall assume that
(i) the line is lossless, and therefore the surge impedance of the line (Zo) and
the wave propagation velocity c are given by
(8.35)
(ii) the surge impedances looking into systems S and R are equal to Zos and Z0R,
respectively, and are such that
— Z0S—Z0/3 (8.36)
Reverse faults
Let us assume that a solid fault occurs immediately behind the relay location at
R. Now because the fault point coincides with the relaying point, the resulting
superimposed voltage at R, AvR(t), is equal and opposite to the steady-state
prefault voltage waveform Vp(t) at end R, i.e. AvR(t) = —Vp(t). As discussed in
Section 8.2.1, this causes a travelling wave to propagate towards end S at a
velocity c defined by eqn. 8.35. The associated superimposed current waveform
AiR(t) also propagates at a velocity c and the superimposed voltage and current
forward reverse
direction direction
777T
ZQS = ZQ/3
AiR(t)
+
)Av F (t) = AvR(t)
distance
fromR
waveform AvR(t) and AiR(t) are interrelated according to eqn. 8.17. In the
period of time from fault inception up to twice the wave transit time from end R
to S {211 c, / = length of line), AvR(t) = Av^(t), AiR(t) = A$(t) and AvR(t) =
Fundamentals of travelling-wave based protection 131
distance
from R
Figure 8.8 Initial propagation of superimposed quantities following a fault in the forward
direction of the relay at R
(a) superimposed fault network
(b) lattice diagram
(c) time variation of superimposed quantities consistent with fault at
negative peak of prefault voltage
superimposed fault voltage vF(t)
superimposed voltage at R
vR(t) = AvF(t-xF/c)/2
superimposed current at R
iR(t) = -3AvF(t-xF/c)/2Zo
132 Fundamentals of travelling-wave based protection
s ( ) r H )
(8.38)
Ag(t) = -KrAiHt)
where
Kr=(Z0S-ZQ)/(Z0S+Z0)
and Avs{t), Ais(t) are the transmitted parts of superimposed voltage and
current waveforms, respectively, at end S, and AvJ(t), Ais(t) are the reflected
parts of the superimposed voltage and current waveforms, respectively, at end
S.
It will be evident that it takes a time of 2 He for the waveform reflected from
end S to arrive back at R after the occurrence of the fault. Figure 8.7 thus
illustrates the behaviour of the system shown in Figure 8.6 due to a fault that
occurs at the positive-to-negative zero crossing of prefault voltage. Figure 8.7(b)
shows the lattice diagram of the superimposed voltages and currents at any
point on the line, and Figure 8.1 (c) shows the resulting superimposed compo-
nents AvR(t), AiR(t) at the relaying point R. It is particularly important to note
that if it is assumed that the fault occurs at time / = 0, then
(i) AvR(t) = AiR(t) = 0 for t < 0
(ii) Although the waveforms of Figure 8.7 (c) are consistent with a fault at zero
voltage point on wave, the relationship of eqn. 8.37 also holds for all time
from fault inception up to twice the wave transit time, i.e. 21/c for any
arbitrary fault point superimposed voltage waveform AvF(t).
(iii) The relationship described in eqn. 8.37 also holds for any value of
terminating surge impedances (ZQRi Zos) and line attenuation. For exam-
ple, if the source surge impedance at end S were exactly equal to that of the
line, there would be no reflection from S, but eqn. 8.48 still holds for all
times
For faults within system R, there is a time delay between fault inception and the
arrival of the superimposed components at the measuring point. Nevertheless,
eqn. 8.37 holds for all times t = 2l/c after the arrival of the superimposed
quantities at R.
Forward faults
Now let us consider the case where faults occur in the forward direction with
respect to the relay at end R. For example, consider a fault at a point F which is
Fundamentals of travelling-wave based protection 133
Now the reflected components of these voltage and current waveforms for an
arbitrary terminating surge impedance Z0R can be found from the correspond-
ing incident waves and the reflection coefficient:
But the total variation of superimposed voltage and current at, and immediately
after, the time when the waves reach end R (t = xF/c) is given by the sum of
transmitted and reflected waves:
Therefore, by finding the difference between eqns. 8.41 and 8.42 and their sum,
we obtain
Eqns. 8.43 show that, unlike the previously considered reverse fault, the
difference AvR(t) — Z0AiR(t) becomes immediately finite on arrival of the
superimposed waves at xFic after fault inception. This conclusion holds for
faults anywhere on the line or within system S.
Figure 8.8 illustrates the behaviour of the superimposed quantities following
a fault on the line R-S of the system of Figure 8.6. The fault occurs at the
negative peak of the prefault voltage. It is evident from the lattice diagram and
134 Fundamentals of travelling-wave based protection
the waveforms of Figures 8.8(£) and (c), that, initially, the travelling waves of
superimposed voltage and current are of opposite polarity.
However, in case of a fault at the end S, there is a time delay between fault
inception and the arrival of the superimposed quantities at R. Although the
waveforms shown in Figure 8.8(c) are consistent with a fault at peak prefault
voltage, the conclusions drawn equally apply for any other arbitrary superim-
posed fault point voltage waveform.
Relaying signals
The difference in the behaviour of the superimposed quantities according to
whether the fault is in the reverse or the forward direction can be used to
determine the direction of the fault with respect to the measuring point. This is
done by the formation of two signals. In terms of primary system values, the
signals used at R are
SlR=AvR(t)~R0MR(t)
(8.44)
S2R=AvR(t)+R0AiR(t)
The corresponding signals used at S which have the reference directions defined
in Figure 8.8(0), are given as
S I S =A^(0-/?oAf 5 (0
(8.45)
S2S= Avs(t) + R0Ais(t)
where Ro is a surge replica resistance whose value is so arranged as to match
closely the line surge impedance Zo.
By examining the signals at end R for reverse faults, with the assumption of
lossless line and perfect matching of the replica resistance Ro, i.e. by substituting
eqn. 8.37 into eqns. 8.44, we obtain
SlR=0
(8.46)
S2R=2AvR(t)
Figure S.9(a) shows how the signals vary following the reverse fault condition
considered previously. It is important to note that S2R becomes finite before SlR.
Figure S.9(a) also shows the sequence in which the signals SlRand S2R rise above
a given threshold value \VS\, (S2R followed by SlR), at times Tx< T2.
Similarly, the signals SlR and S2R for the forward fault described previously
can be found by substituting eqn. 8.43 into eqn. 8.44. By assuming perfect
matching of the replica resistance Ro to the surge impedance Zo, these signals
would be given as
SIR = 2AvF(l-xF/c)
(8.47)
S2R = 2AvF(t-xF/c) (Z0R-Z0)/(Z0R+Z0)
Figure 8.9(6) illustrates the behaviour of the signals SIR, S2R due to the forward
fault condition on the system of Figure 8.8. It can be seen that the setting levels
Vs and — Vs are exceeded almost simultaneously (TlssT2) owing to the very
Fundamentals of travelling-wave based protection 135
\
1
voltage
Vs Vs
/T, xp/c = Ti = T 2
2I/C
j • • • " /
Figure 8.9 Variation of relaying signals at R following faults on the system of Figure 8.6
(a) reverse fault at R
(b) fault on line R-S
R0MR(t)
SlR = AvR{t) - R0MR(t)
S2R = AvR(t) + R0MR(t)
= 0, 1, 2 (8.48)
136 Fundamentals of travelling-wave based protection
8.3.3 Behaviour of relaying signals at the relay and fault locations [10]
The way relaying signals S, and S2 behave at the relay and fault points of
discontinuity can be used to estimate the distance of the fault to a relay.
Consider a forward fault on the line of the system shown in Figure 8.10. The
superimposed voltage at the fault point F is AvF(t) and its associated
superimposed current AvF(t)/Z0 propagates towards end R.
Thus, if the relaying signal SlR (see eqn. 8.44) is formed from the above
voltage and current before and after they are reflected at the relay busbar we
would find its value does not change, and it remains equal to
SlR=2AvF(t-xF/c) (8.49)
The reflected voltage and current waves then propagate back towards the fault,
where they are reflected back again towards end R. Assume the most common
case, where the fault resistance Rj is much smaller than the surge impedance of
the line. Under this condition the coefficient of reflection at the fault point
Kf=(Rf—Z0)/(Rf+Z0)9 would be approximately equal to unity. Therefore,
according to eqns. 8.20 and 8.21, the voltage and current waveforms will be
reflected at the fault with almost the same magnitudes. However, the reflected
voltage would have the same polarity as the incident voltage waveform, while
the reflected current would have opposite polarity. If SlR is recalculated from the
reflected waves, its value would be zero. The important conclusion from this
analysis is that the relaying signal SlR that leaves the relay point toward the fault
would be reflected back to the relay, with its shape and magnitude changed at
the fault point.
A similar analysis can be carried out with regard to S2R, which reveals that it
is changed at the relaying point, in exactly the same manner as the signal SlR. It
follows that, if the time between like changes in SlR and S2R is equal to r, the
distance to the fault xF would be
T/2
xF = (8.50)
c
Figure 8.10 shows the response of a single-phase lossless line to a 1 p.u. voltage
step function applied at the fault point, for values of fault resistance Rj of zero
and of 25% of the characteristic impedance of the line. The coefficients of
reflection are calculated on the basis that inductive sources are connected at
buses 1 and 2, having short-circuit capacities of 35 000 MVA and 10 000 MVA,
respectively.
AvR(t)
RoAiR(t)
RoAiR(t)-AvR(t)
S2R
stored S2R
Figure 8.10 Variation of relaying signal SlR and S2R at the relay and the fault points
(a) system under fault condition with the corresponding lattice diagram
(b) voltage and current waveforms due to a zero resistance fault and the
corresponding signals .S^ and S2R with zero fault resistance
(c) the signals SlR and S2R due to the same fault but with fault resistance
equal to 25% of surge impedance
138 Fundamentals of travelling-wave based protection
L
Ls R f c S
= Vsin(coot
Ls R Lf s
—>—
Figure 8.11 Representation of single-phase system used to explain the elliptical trajectory
based technique
(a) steady-state network
(b) superimposed network
dAiR(t)
(8.51)
or
AiR(t) = A/(cos(tty + 0) -cosO) (8.52)
where
A/=
a)0(Ls
Fundamentals of travelling-wave based protection 139
Again with reference to Figure 8.11(£), the superimposed voltage Az^(/) can
be determined from the superimposed current AiR(t):
dMR(t)
AvR(t)~-Ls~j-~ (8.53)
and XS = O)QLS is the reactance of the source, and Xj= (OoLfis the reactance of the
line to the fault.
Multiplying eqn. 8.52 by a mimic resistance R and combining the resulting
equation with eqn. 8.54 results in
Av2R(t) (RAiR(t)-RAIcos0)2
+
"(Alf (RAW -1 (8 55)
*
By examining eqns. 8.52 to eqn. 8.55, the following conclusions can be drawn:
(i) Eqn. 8.52 shows that the superimposed current component AiR(t) contains
a DC offset equal to A/ cos 0, which depends on the fault inception angle
e.
(ii) Inclusion of resistance in the faulted loop will, of course, cause a decay of
the DC offset. On the other hand, such resistance will slightly reduce the
magnitudes of A V and A/ and introduce an additional phase shift between
the fundamental frequency components of the superimposed quantities.
(iii) Eqn. 8.55 represents an ellipse in the superimposed plane (i.e. the
RAiR(t)-AvR(t) plane), with its centre being shifted in the direction of
RAiR(t) by a constant value of RAI cos 0. This ellipse is known as the fault
trajectory.
Figure 8.12(a) shows the steady-state voltage vSR(t), the superimposed voltage
AvR and the superimposed current AiR due to the fault under consideration.
Figure SA2(b) shows the fault trajectory. It can be seen that immediately after
the fault, AvR will have a finite value, while AiR (under the assumption of a
lossless line) changes gradually from zero. Therefore, the trajectory initially
jumps from the origin of the plane to a point somewhere along the AvR axis. It
will then proceed along the elliptical path of Figure 8.12(^). According to eqn.
8.54, the magnitude of initial jump AvR is a maximum when the fault inception
angle 0 is equal to ± n7t/2, i.e. the prefault voltage is maximum, while AvR is
zero for a zero fault inception angle.
140 Fundamentals of travelling-wave based protection
VSR(t)
AV RAi
RAI
forward fault
RAI COS 6
reverse fault
Figure 8.12 (a) Behaviour of voltage and current waveforms under fault
conditions
(b) corresponding fault trajectory
iR(t) «s(t)
—*—O
vR(t) vs(t)
If we consider the transmission line shown in Figure 8.13, eqn. 8.56 can be
applied to the R and S ends, where x = 0 and x — l, respectively. The following
equations are then obtained:
1
(8.58)
By renaming i(0, /), v(0, t) as iR{t) and vR(t), i(l, t) and v(l, t) as -is(t) and vs(t),
substituting / by (t — l/c) in eqn. 8.57, and combining the resulting equation
with eqn. 8.58, we obtain Bergeron's equation:
1 1
iR(t - r) + — vR(t - r) = - is(t) + — vs(t) (8.59)
Zo Ao
=Lmim
dx
(8.62)
where
L (0) 0 0 c0) 0 0
0 £ (1)
0 cm= 0 C (l) 0
(2) 2)
0 0 z 0 0 c<
In fact, eqns. 8.62 represent independent sets of modal equations. Therefore,
the derivation procedure outlined previously for the single-phase line is
applicable to each of these modal equations. Thus, for a three-phase line,
Bergeron's equation can be derived from eqns. 8.59 and 8.60 as follows:
(8.63)
1
A)
(8.66)
The forward wave/, i.e. Av*(t) and/or Ai*(t)y which propagates towards end
R, and the backward (or reflected) wave b, which is reflected back towards the
fault point F, are shown in Figure 8.14(rf).
By substituting AvF(t) and AiF(t)y given by eqns. 8.65 and 8.66, respectively,
into eqn. 8.64 we obtain
AvR(t-¥T)-Z0AiR(t + T) = -2Vsm(<t)Qt + d) r^t^2t (8.67)
It can be seen that eqn. 8.67, which describes the wave characteristics as seen at
end R, depends on the fault inception angle 6. This dependency can be removed
144 Fundamentals of travelling-wave based protection
by taking the derivative with respect to time of the wave characteristic eqn.
8.67, which in turn gives
IFS
•O
v F (t)
"Ft
-<—
F
o
AiR(t) AiF(t) F
->
Av R (t) Av F (t)
2T
Figure 8.14 Single-phase system used to explain the forward discriminant function
(a) system under steady-state conditions
(b) system under a forward fault condition with respect to a relay at end R
(c) superimposed network
(d) the corresponding forward and backward waveforms,/and b respecti-
vely
Fundamentals of travelling-wave based protection 145
o o
iRS(t)
VRS(t)
-o
V f VRf(t)
Av F (t)
AVR
= -V F (t)
Figure 8.15 Single-phase system used to explain the backward discriminant function
(a) steady-state system
(b) system under a reverse (or backward) fault condition with respect to a
relay at end R
(c) the corresponding superimposed network
(d) the corresponding forward and backward travelling waveforms,/and
b respectively
This equation can be combined with eqn. 8.67 to give eqn. 8.69, which
describes the wave characteristic as seen by a relay located at end R due to a
fault on the line:
1 Yd V
]2
DF= (AvR(t) -Z 0 A^(/)) 2 + — - (AvR(t)-Z0MR(t))
ujQ^at j
= 4 ^ = 8 K22RRMS (8.69)
146 Fundamentals of travelling-wave based protection
where FRMS is the RMS value of the supply voltage and DF is known as the
'forward wave discriminant function'. Repeating the analysis for faults that
occur behind the relay located at end R, as shown in Figure 8.15, results in
another function, which is often called the 'backward wave discriminant
function' DB
* = 0, 1,2 (8.71)
k = 0, 1,2 (8.72)
8.6 References
1 BEWLEY, L.V.: 'Travelling waves on transmission systems (2nd edition)' (John
Wiley, 1951)
2 BICKFORD, J.P., MULLINEUX, N., and REED, J.R.: 'Computation of power
system transients' (IEE Monograph Ser. 18, Peter Peregrinus, 1976)
3 WEDEPOHL, L.M.: 'Application of matrix methods to the solution of travelling-
wave phenomena in polyphase systems', Proc. IEE, 1963, 110, pp. 2200—2212
4 DOMMEL, H.W.: 'Digital computer solution of electromagnetic transients in single
and multiphase networks', IEEE Trans. 1969, PAS-88, pp. 388-396
5 DOMMEL, H.W., and MICHELS, J.M.: 'High speed relaying using travelling
wave transient analysis', PES Winter Meeting, 1978, Paper A78 214-9
6 CHAMIA, M., and LIBERMAN, S.: 'Ultra high speed relay for EHV/UHV
transmission lines—development, design and application', IEEE Trans. 1978,
PAS-97, pp. 2104-2116
7 YEE, M.T., and ESZTERGALYOS, J.: 'Ultra high speed relay for EHV/UHV
transmission lines- installation-staged fault tests and operational experience', ibid,
pp. 1814-1825
Fundamentals of travelling-wave based protection 147
9.1 Introduction
1 1
**(' " T) + — VR(t - T) = - is(t) + - vs(t) (9.1)
A) A)
1 1
- vR{t) = is(t - r) + - vs{t - r) (9.2)
A ^
As explained in Chapter 8, eqn. 9.1 implies that a travelling wave leaving end R
will arrive at end S after a time delay of r, while eqn. 9.2 implies that a
travelling wave leaving S will arrive at R after a similar time delay. For faults
external to the line, eqns. 9.1 and 9.2 are satisfied. However, when a fault occurs
Travelling-wave protective schemes 149
on the line, they are not satisfied, because a discontinuity occurs between each
end. Thus, the equations can be written in the form of eqn. 9.3, in which
£j(/) = o, e2(t) = 0 for an external fault and
£,(/)=£(), ^ W ^ O f° r an
internal fault,
where
1
I & iR(t - r) + is(t) + — [vR(t -T)-v:
-A (9.3)
1
& is(t - T) +«« [vs(t - r) -
150 Travelling-wave protective schemes
1 1
"-iFR(t-T2)+— vF(t-T2) = iR(t-Tl-T2)+ — vR{t-rx-r2)
z0 z, 0
Combining these equations and noting that r, + r2 = r, results in
1
Now it can be seen from Figure 9.1 and eqn. 9.3 that
et(t) = iF(t-r2) (9.4)
Similarly it can be shown that
e2(t) = ip(t-rl) (9.5)
It will be evident that the functions £l(t) and £2(t) effectively describe the
current in the fault path in terms of the voltages and currents at the line ends. A
nonzero value for £{(t) and £2(t) effectively signifies that a fault exists between
the measuring points, a fact that can be used to generate a trip signal. The
formation of the signals £j(/) and £2(t) require the transmission of the voltage
and current waveforms between the line ends to form, in effect, a unit protection
scheme. This technique has recently been implemented for plain feeders [3, 4]
and has been extended to Teed power circuits [5]. Further details of the
implementation are given in Chapter 11.
Three-phase lines
The concept of detecting internal faults using Bergeron's equation can easily be
extended to three-phase lines. Consider the three-phase line shown in Figure
9.2. Assume the line is transposed, lossless and has a series per unit length
inductance matrix L and a shunt capacitance matrix C. The line can be
decomposed into three equivalent single phase lines using eqns. 8.23, such that
VM S V l)
' " '' ,9.6,
i.»=<r'v>
As previously explained, the voltage vector vm(t) defines three separate modal or
component voltages, which can be derived from the actual phase voltages
described by the vector vp{t). The component currents im(t) can be similarly
described in terms of the phase currents. The single-phase theory developed
above can then be applied to each component or mode voltage and current
Travelling-wave protective schemes 151
thereby derived, and the previously derived discontinuity functions of eqn. 9.3
can be written for each mode component. For the kth mode eqn. 9.3 is thus
written as
1
(9.7)
{ k)
2 (t)
(t) == i f ( / -
where Z{ok) is the kth mode surge impedance and x{k) is the kth mode surge
travelling wave time, where A; = 0, 1, 2 for a three-phase line.
Thus the fault detection criteria in a three-phase line can be written as
follows:
• If £{0)(t) = 0 and £{l)(t) = 0 and e{2)(t) = 0, then the line is healthy
• If £{0)(t) ¥=0 or £{])(t) # 0 or £{2)(t) =£0, then an internal fault has occurred.
The physical meaning of £{k) and e^ is that they describe the modal component
values of the current in the fault path, i.e.
£f{t) = if{t-xf), * = 0, 1 and 2
(9.8)
£i*)(/) = ij*>(/-ri* ) ),* = O, 1 and 2
R ^
iR(a) i s (a]
^^ ^
i R (b) is(b)
v R (b) v s (b)
v R (c) vs(c)
where if\t) is the £th modal component of fault path current, t\k) is the A:th
modal surge travelling wave time from the point F to the terminal S and r ^ is
the kth modal surge travelling wave time from the fault point F to the terminal
R.
ryyys nryy\
analogue
interface
analogue ~3
interface _ J E
sss
steady-state
sss
steady-state
frequency suppressors frequency suppressors
i>
amplifiers amplifiers amplifiers amplifiers
local local
trip block block trip
trip trip
information link
Tr Tr
output output
interface interface
surge
replica
circuits
signal-mixing-circuits/analogue interface
superimposed
trip outputs from component block outputs from
0 and 2 channels circuits modes 0 and 2 channels
sequence
detector
(mode-1)
similar
)• arrangement
to end R
blocking relay
R«(t-TVR0«viR«(t = T)
(k) (k) (k)
v R (t)-Ro iR (t) time delay T
(BBD)
differential
amplifier
steady-state
component
true superposition
component
time
7/777
rui OIL
—LTU—inr 77777 (ii)
_iui n_n_
(x)
77/77
rL = 60ms (v)
77777
voltage Vs. On the other hand, an input to the pulse-generating circuits PG1 or
PG2 causes their output to fall to zero for a preset time TL. The latter time
determines how long the detector is latched following a forward or reverse fault,
thus ensuring that any subsequent change in the input signals is ignored.
Similar considerations apply to PG3, but this circuit is arranged so that its
output is normally zero and rises for a time of 60 ms following an input. The
tripping signal has to be delayed to enable the local blocking relay to operate on
receipt of a carrier blocking signal generated by the relay at the other end of the
line following an out-of-zone fault. This is achieved by the inverting delay
circuit which delays the trip signals by the required time.
160 Travelling-wave protective schemes
Operating principle
The operating principles adopted for the digitised UHS scheme are basically
the same as those used in the analogue UHS scheme. For ease of explanation,
consider the superimposed single-phase equivalent circuit shown in Figure 9.7.
The composite superimposed signals used are those defined in eqns. 8.44,
rewritten below. As also explained previously, the actual signals used employ
modal components but for ease of explanation only the equivalent single phase
form of signals repeated in eqn. 9.10 will be considered.
SlR(t) = AvR(t)-RQMR(t)
(9.10)
S2R(t) = AvR(t)+R<£iR(t)
Signal magnitude comparison criteria were used to enable the directional
decision to be based on several consecutive samples of the composite signals of
eqn. 9.10, thereby improving relay security. This is desirable because, for a
reverse fault behind the relay location in Figure 9.7, the simple signal criterion
of eqn. 9.11 only holds for twice the wave transit time (l/c) between the line
ends. In the case of a forward fault at a distance xF from the relay location
considered in Figure 9.7, the signal criterion of eqn. 9.12 holds only for so long
source source
R relay location fault S
AiR(t) F
l
Av R (t) v F (t)
XF
as reflections from within the source network R take to return to the relay
location.
\Sw{t)\>\SXR{t)\ (9.11)
(9-12)
Figure 9.8 Typical superimposed voltage and current for lumped parameter system model
(a) fault at zero voltage
(b) fault at voltage maximum
(c) fault at 145° after voltage zero
Travelling-wave protective schemes 163
-m line
v a ,v b .v c
w. V
forward/reverse
detection signals
scheme logic
breaker
trip TT remote relay
block signals
-O
-0-0 receive send
selection for particular applications. The arrangement of Figure 9.9 forms the
basis of the digital algorithmic relay developed by GEC Alsthom
Measurements, type designated LFDC.
D (2)
1 0 1 1 1 1 1 1 1 1
D(0) 1 1 1 0 0 0 1 1 1 0
Ph 0 1 1 1 1 1 1 1 1 1
"b"
D(2) 1 1 0 1 1 1 1 1 1 1
D(0) 1 1 1 0 0 0 1 1 1 0
Ph D d) 1 0 1 1 1 1 1 1 1 1
"c"
D (2)
0 1 1 1 1 1 1 1 1 1
no
D(i) F l b orD(i) F l a > £ \
no^
V L-L or
3L's fault K D(0)Fla<£
trip "b"
trip "c"
L-L-G
fault
Figure 9.10 Fault detection, directional discrimination, phase selection, andfault classifi-
cation based on the Karrenbauer transform [11]
(a) truth table
(b) process flow chart
166 Travelling-wave protective schemes
forward faults give rise to a clockwise direction of rotation, while reverse faults
give rise to trajectories with anticlockwise rotation.
9.6.2 Sense of trajectories versus fault direction
It has been shown in Section 8.3.4 that the behaviour of the superimposed
voltage and current at the relay point can be described by an elliptical
trajectory in the AvR versus RAiR plane, where R is an arbitrary scale factor.
The trajectory itself is mathematically represented by eqn. 8.55, rewritten
below. (Note that subscript R is omitted in the following analysis.)
Av2(t) [RAi(t) - RAI cos Of
l (9J3)
[RAIf
Figure 9.11 shows the elliptical trajectories due to forward and reverse faults
with respect to a relay located at the R end of the line R-S. Figure 9.11 (a)
illustrates the superimposed component trajectories due to forward faults.
These trajectories possess a clockwise sense of rotation and they depart from the
Av axis, either into quadrant II or IV, depending on whether the fault occurs on
the positive or negative half cycle of the prefault voltage waveform.
On the other hand, Figure 9.11 (b) illustrates the trajectories due to a reverse-
fault condition. Here it can be seen that the trajectories rotate in an
anticlockwise direction and initially enter quadrant I or III, again depending
on which half cycle of the voltage waveform the fault occurs in. Therefore, a
basic technique to determine the direction of the fault using the above
properties of the superimposed component trajectories is to introduce threshold
boundaries in the Av—RAi plane. Once the trajectory has crossed the boundary,
the relay may reset or adjust itself to the postfault sinusoidal voltages and
currents; a typical boundary is shown in Figure 9.12. A simplified method of
boundary checking involves taking linear combinations of Av(t) and RAi(t) and
checking against fixed values. For example, if the linear combination of
RAi(t)-Av(t) first reaches some threshold value, say +£"0 or — Eo, a forward
fault will be indicated. Mathematically this is expressed as
RAi(t)-Av(t) = ±E0 (9.14)
However, if the linear combination of RAi(t) 4- Av(t) first reaches the threshold
values, then a reverse fault will be indicated. More explicitly this can be
expressed as
RAi(t) + Av(t) = ±E0 (9.15)
Boundaries such as hyperbolas, circles, ellipses or piecewise linear curves can
also be used.
RoAi
forward Mult
A\ c^s (9)
Figure 9.12 Typical boundary used to determine fault direction in trajectory-based relay
[12]
(9.16)
1
point of t=0
local reflection remote
source^ R 3 source
9.7 References
1 TAKAGI, T. et al.: 'Fault protection based on travelling wave theory, Part I: theory',
PES Summer Meeting, Mexico, 1977, Paper A77, 750-753
2 TAKAGI, T. et al.: 'Fault protection based on travelling wave theory, Part II:
sensitivity analysis and laboratory test', PES Winter Meeting, 1978, Paper A78, 220-
226
3 AGGARWAL, R.K., and JOHNS, A.T.: 'A differential line protection scheme for
power systems using composite voltage and current measurements', IEEE Trans.,
1989, PWRD-4, pp. 1595-1601
4 AGGARWAL, R.K., and JOHNS, A.T.: CA new differential protection scheme for
power systems using composite voltage and current measurements', 23rd UPEC,
1988, Paper A3.4
5 AGGARWAL, R.K., and JOHNS, A.T.: 'New approach to teed feeder protection
using composite current and voltage signal comparison', Developments in power
system protection (IEE Conf. Publ. 302, 1989), pp. 125-129
6 CHAMIA, M., and LIBERMAN, S.: 'Ultra-high speed relay for EHV/UHV trans-
mission lines—development, design and application', IEEE Trans. 1978, PAS-97,
pp. 2104-2116
7 YEE, M.T., and ESZTERGALYOS, J.: 'Ultra high speed relay for EHV/UHV
transmission lines—installation-staged fault tests and operational experience', ibid,
pp. 1814-1825
8 JOHNS, A.T.: 'New ultra-high-speed directional comparison technique for the
protection of EHV transmission lines', IEE Proc. C 1980, 127, pp. 228-239
9 JOHNS, A.T., MARTIN, M.A., BARKER, A., WALKER, E.P., and CROSSLEY,
P.A.: 'A new approach to EHV directional comparison protection using digital signal
processing techniques', IEEE Trans. 1986, PWRD-1, pp. 24-34
10 MANSOUR, M.M., and SWIFT, G.W.: 'A multi-microprocessor based travelling
wave relay—theory and realization', ibid, pp. 273—279
11 MANSOUR, M.M., and SWIFT, G.W.: 'Design and testing of a multi-
microprocessor travelling-wave relay', ibid, pp. 74—82
12 VITINS, M.: 'A fundamental concept for high speed relaying', IEEE Trans. 1981,
PAS-100, pp. 163-173
Chapter 10
Digital differential protection of
transformers
10.1 Introduction
This Chapter gives a brief general review of the principles of transformer
differential protection. This is followed by an explanation of the application of
digital techniques and the algorithms that have been developed specifically for
application to transformer protection. The algorithms covered include finite-
duration impulse (FIR) filters, least-squares curve fitting, the digital Fourier
algorithm and the flux-restrained current differential algorithm.
Finally, the basic hardware arrangement for implementing digital techniques
to the protection of transformers is described. It is, however, important to note
that closely similar techniques can be applied to the protection of generators,
although, in this case, the transformation ratio of currents is the same on each
side of the protected zone.
transformer
primary and secondary current transformers are connected such that ideally,
under healthy conditions, only circulating currents flow and the differential
signal ID is zero. If an internal fault occurs within the transformer, the balance
between the primary and secondary currents is disturbed and a differential
current signal (//> = Ip — Is) then causes the relay to operate.
10.2.2 Biased differential relaying
If the transformer is equipped with a tap changer, the imbalance between the
primary and secondary currents introduced by the variation of the turns ratio
can be great enough to cause malfunction during external fault conditions.
Imbalance under healthy conditions can also be caused by mismatch between
the current transformers or by saturation. It is therefore important that this
situation should be taken into account in the engineering of transformer
differential protection systems.
A common way of overcoming such problems is to bias the operation of the
relay by deriving a biasing signal equal to the sum of the currents measured on
each side of the transformer. This is illustrated in Figure \0.2(a) together with a
typical operating characteristic (Figure \0.2(b)). It is evident from Figure
\0.2(b) that the differential current required to operate the relay increases with
the bias current. Thus the stability of the relay under external fault conditions is
ensured. For example, if a heavy through fault (healthy) condition exists, the
differential current may take a finite value that is nevertheless much smaller
than the associated bias current. The restraining characteristic of Figure \0.2(b)
is arranged to allow a significant level of differential current to occur without
causing tripping under healthy conditions. Conversely, a faulty condition
causes the bias current to be relatively small, on account of the reversal of the
current measurand Is, thus ensuring that tripping occurs.
In some cases the operating current is expressed as a percentage of the
restraining current, and such relays are commonly known as percentage-biased
differential relays. Many practical designs try to derive differential and bias
signals proportional to the fundamental or power frequency components of
measured current, which is why in Figure \0.2(b) the measurands are
subscripted.
10.2.3 Harmonic-restrained differential relay [5, 6]
The differential protective scheme described above is generally very satisfactory
under normal operating conditions, but is prone to false operation during
energisation of the transformer. Under this condition, an inrush current flows
only through one winding (the winding being energised), which consequently
appears to the relay as an internal fault.
This problem can be overcome by using the fact that inrush currents usually
contain harmonics. Of these, the second harmonic is usually predominant
under all energisation conditions. The protection can therefore be designed so
that it is prevented (restrained) from tripping if the magnitude of the second-
harmonic component of operating current is greater than a certain prescribed
Digital differential protection of transformers 173
A A A A
V V
bias signal
bias signal component (ls)
component (IP)
differential signal
relay operates
relay restrains
restraint/bias current,
=
lm Mp"~Ms
The restraining signal (Im), on the other hand, is commonly formed from the
fundamental components of the primary and secondary components such that
4i=/M+4 (10.2)
The other restraint signals are the second-harmonic restraint Im and fifth-
harmonic restraint current IR5, which are commonly formed from the magni-
tudes of differential currents of any second and fifth harmonics detected. A
commonly used technique involves summing any such harmonic components in
each phase:
/«=|/i».l + |/i»l + |/OTe| (10.3)
and
/*5=|/zJ + | / / J + |/z>5,l (10.4)
where ID2af Imif and ID2c are the differential currents due to second-harmonic
components in phases 'a', 'b' and V, respectively, and ID5ay Imb and ID5c are the
corresponding differential currents due to fifth-harmonic components.
Tripping and restraining signals can be derived using analogue techniques
(as in the case of electromagnetic relays or solid-state relays) but this Chapter
deals with the derivation and use of such signals using digital techniques.
103A FIRfiltercharacteristics
The filters used are assumed to be characterised by means of finite-duration
impulse responses of duration equal to one power-frequency cycle of period T
and having values of +1 or — 1 at any instant during that period. The choice of
± 1 for the impulse response avoids the need tofindthe responses of thefiltersto
arbitrary inputs and eliminates the time-consuming multiplications that would
otherwise be required, although in many modern DSP processors this is not
necessary.
To estimate the magnitude of fundamental and second-harmonic compo-
nents of the input, four filters are required: two for the fundamental and two for
Digital differential protection of transformers 175
the second-harmonic component. Their impulse responses are S{(t), C{(t), S2(t)
and C2(t) (Figure 10.3), which in turn are defined by the following equations:
(ia6)
, 7/2^/^37/4
-1 774^/^772, 3 7 / 4 ^ / ^ 7
+ 1 0^/^7/8, 37/8^/^57/8,
, 57/8=^/^
It will be apparent that £,(/), C,(/) are impulse responses for the sine and cosine
parts of the fundamental component, and S2(t), C2(t) are impulse responses for
the sine and cosine parts of second-harmonic components. Eqn. 2.73 can be
applied to eqns. 10.5-10.8 to find the system frequency responses of the four
filters, which are consequently given by
^ - 2 ™(^
=1 e-J">™(coS(^Pj - 2 c o s ^ + 1^ (10.11)
(^)) (10.12)
T/2 3T/4
-1
-•t
T/4 T/2 3T/4 T
-1
S2(t)
T/2
-1
C2(t)
i
1
>
T/4 T/2 3T/4 1r
-1
Figure 10.3 Impulse response of fundamental and second harmonic FIR filters
Digital differential protection of transformers 17 7
1 2 3 4 5 6
a
0.5
1 2 3 4 5 6
0.5
IF S2 ((o)l
2 3 4 5
c
0.5
IFC2(©)I
1 2 3 4 5 6
d
change. Thus the outputs of filters 1, 2, 3 and 4 are obtained by summing the
samples of their inputs over one cycle. However, the signs of these samples must
be modified in accordance with the impulse responses of the filter under
consideration.
Assume N is the number of samples per cycle of the current i(t) and is chosen
as a multiple of eight. In this case, the time between successive samples is
At = 2Jt/(N(o0) and ik—i{tk) is the Mi sample at any time t = kAt. The result of
time-discrete convolution of the samples ik with the impulse responses defined
by the previous equations will then be given as
(10.13)
(10-14)
k=l
JV/4
(10.15)
k=\
A78
(10.16)
e ^0.093 X/R=\0
0.054 X/R = 20
where XIR is the system reactance-resistance ratio.
Digital differential protection of transformers 179
Imcos(mco0t+dm) (10.18)
(10.19)
or
with
02 = 4 * 2 =-/ 0 /r
aa+2 = cos na)otk) xn+2 = /„ cos 0n
an+7 = sin n(O0tk) xn+7 = In sin 6n
180 Digital differential protection of transformers
A J£_,= « (l0-20)
JVX12 12X1 JVxi
or
X=Bi (10.21)
N
= :=
Ix sin 0 i * 8 ^ . b(8, n)in
(10.22)
I2 cos 82 = x4 = > £(4, n)L
= x9 = ^ b(9, n)in
where xk is the £th element of vector X and b(k, n) is the element of the A:th row
and nth column of matrix B.
Using the foregoing equations, the amplitude of the fundamental and the
second harmonic can be calculated as
x4
Figure 10.5 Schematic representation of the determination of the second harmonic (SH)
ratio
in is the wth sample, b(k, n) is the element of kt\i row and wth column of
matrixB (see eqn. 10.21) and 2 is the sum from n=\ toN, i.e. the number
of samples per fundamental period
eqn. 10.23, are then used to calculate what is called the 'second harmonic ratio',
which is defined by
SHR = - (10.24)
If the SHR is greater than a set value, an inrush current condition is assumed
and tripping is prevented; otherwise an internal fault condition is assumed and
a tripping signal is issued to isolate the transformer.
Figure 10.5 shows a diagrammatic representation of the four digital filters
used to derive the real parts, i.e. x3 and #4, and imaginary parts (x8 and #9) of the
fundamental and second-harmonic components used in calculating the SHR.
The frequency characteristics of the four filters are shown in Figure 10.6 where
it can be seen that the unwanted harmonics are filtered out.
Now assume (as previously) that the current waveform is sampled N times
per period of the fundamental, and let the samples be denoted by ik=i(kAt).
The real and imaginary parts of the nth harmonic (an and bn) can be found by
1 -
Ix3(co)l
IX4(G))l
Figure 10.6 The magnitudes of the frequency responses of digital harmonic extraction
filters
(a) xs(a>)
(b) xB(a>)
(c) x<(a>)
(d) x9(a>)
Digital differential protection of transformers 183
using eqns. 5.5 and 5.7. In terms of current samples starting at the rth sample, an
and bn can then be expressed as
"1 h
2 J*N~l /2jtk\
w= f sin
i v 2 * «hr
(10 26)
-
The magnitude |/£r)| of the wth harmonic and its phase angle 6^ can thus be
calculated as explained previously. Thus
0W = t a i T ' ^ (10.27)
For the transformer protection application, n takes the values of 1, 2 and 5 for
fundamental, second and fifth harmonics, respectively. The result can be
updated iteratively as each new sample becomes available. This is done by
dropping the earliest sample and adding the new sample:
2nr
A r i ) = *<" + - [iN+r-ir] sin nl — \ (10.29)
where ir and %+r are the oldest and newest samples, respectively.
Having determined the fundamental, second- and fifth-harmonic compo-
nents, the transformer protection is then implemented following the principles
outlined in Section 10.2.3.
Other techniques have been reported in the literature to extract fundamental
and other harmonic components [11]. These techniques give closely similar
performance and include the use of rectangular transforms [12], Walsh
functions [13] and the Haar function [14].
(10.30)
dt dt
f'2 (10.31)
J /,
Applying the trapezoidal rule to the integral part of eqn. 10.31 results in
1
(10.32)
If the voltage and current waveforms are sampled such that the sampling
interval is equal to A/, then at the kth sample, eqn. 10.32 can be expressed using
sample notation as
(10.33)
where iPt ky vfiy k are the A:th samples of the primary current and voltage.
Eqn. 10.33 can therefore be used to calculate the mutual linkage flux tp of the
transformer. On the other hand, the differential current can be calculated from
Digital differential protection of transformers 185
the primary and secondary currents. Thus at time th the differential current idk
is given as:
where iPt k is as defined previously and is k is the kth sample of the secondary
current. However, from transformer theory, the differential current idk is equal
to the magnetising current of the transformer. Therefore, if it is assumed that
the flux linkage computed by eqn. 10.33 is a true representation of the actual
flux in the transformer, the samples of the differential current and flux linkage
(idh \pk) are expected to fall on the open-circuit magnetising curve of the
transformer.
In the first phase of the algorithm, the principle of the percentage-biased
differential relay is used to detect internal faults at every sampling interval. The
second phase of the algorithm involves a simultaneous check on the location of
the point (idh xpk). If this point does not fall on the i^-xp curve of the transformer,
a trip signal is issued.
The basis of this approach is explained in Figure 10.8(a), which shows an
open-circuit magnetising characteristic of a transformer and the (idh tpk)
relationship for an internal fault inside the transformer. For the internal fault
condition, the terminal voltage (and hence tpk) is much smaller than is the case
under inrush conditions. Consequently there are distinct regions in the i—xp
plane that define the fault or non-fault status of the transformer.
The above technique works satisfactorily when the residual flux in the core of
the transformer is close to zero. Practically this is not always the case, and
consequently the ij-\j) characteristic of the transformer may vary as shown in
Figure 10.8(6). This situation makes it impossible to differentiate between fault
and no-fault regions in the i-ip plane if the flux linkages are computed
according to eqn. 10.33. This is because the estimated value is subjected to an
error equal to the value of the residual flux linkages.
One way of overcoming this problem is to use a restraining function which is
determined by the slope dxp/di rather than the flux ip itself. From eqn. 10.33 we
then obtain
Figure 10.9 shows that there are two regions in the dtp/di—i plane. The first
region corresponds to a fault condition or an operation in the saturated part of
the magnetisation curve. The second region, which is significantly removed
from the first, designates an operation on the magnetising curve in the
unsaturated part. It has been observed that for internal fault conditions the
current samples and corresponding flux derivative dxp/di remain continuously
in region 1. On the other hand, during inrush conditions, they alternate
between the two regions. This phenomenon can be used to create an index of
restraint kT which is increased each time a sample pair (ik, (dip/di)\k) falls in
186 Digital differential protection of transformers
region 1. However, the index is decreased whenever the sample pair enters
region 2. The index is also constrained to remain always positive. Thus
kr = kr -f l if the differential current indicates a
"dT
CB CB
•m -m
-m- •m-
I—omnmr—I
A
>r'sc
trip signal trip signal;
interfacing interfacing
microprocessor system
circuits circuits
trip
control bus
f A
T
RAM DAC
ROM DAC
12 12
bit timer bit signals for
terminal
host
computer
analogue
signal address bus
amplification
particular algorithm chosen. In the event of an internal fault a trip signal is sent
through the D/A convertor to trip the circuit breakers on both sides of the
transformer.
It will be noted that unlike conventional schemes, no interposing current
transformers are required for phase rotation, as this is readily implemented by
software.
10.8 References
1 KENNEDY, L.F., and HAYWORD, C D . : 'Harmonic-current restrained relays for
differential protection', AIEE 1938, 57, pp. 262-266
2 HARDER, EX., and MARTER, W.E.: 'Principles and practices of relaying in
United States', AIEE, 1948, 67, Pt II, pp. 1005-1022
3 MATTHEWS, C.A.: 'An improved transformer differential relay', AIEE, 1954, 73,
Pt III, pp. 645-650
4 ROCKEFELLER, C D . : 'Fault protection with a digital computer', IEEE Trans.
1969, PAS-88, pp. 438-464
5 EINVALL, C . H , and LINDERS, J.R.: 'A three phase differential relay for
transformer protection', IEEE Trans. 1975, PAS-94, pp. 1971-1978
6 HERMANTO, I., MURTY, Y.V.V., and RAHMAN, M.A.: 'A stand-alone digital
protective relay for power transformers', IEEE Trans. 1991, PWRD-6, pp. 85-95
7 SCHWEITZER, E . O , LARSON, R.R., and FLECHSIG, A.J.: 'An efficient inrush
current detection for digital computer relay protection of transformers', IEEE PES
Winter Meeting, 1977, Paper A7751O-1
8 LARSON, R.R., FLECHSIG, A.J., and SCHWEITZER, E.O.: 'The design and test
of a digital relay for transformer protection', IEEE Trans., 1979, PAS-98, pp. 795-
804
9 DEGENS, A.J.: 'Algorithm for a digital transformer differential protection based on
a least-squares curve fitting', IEE Proc. C, 1981, 128 (3), pp. 155-161
10 DEGENS, A.J.: 'Micro-processor-implemented digital filters for inrush current
detection', Int. J. Elec. Pow. & Energy Syst., 1982, 4, pp. 196-205
11 RAHMAN, M.A., and JEYASURYA, B.: 'A state of the art review of a transformer
protection algorithm'. IEEE Trans., 1988, PWRD-3, pp. 534-544
12 RAHMAN, M.A., and DASH, P.K.: 'Fast algorithm for digital protection of power
transformers', IEE Proc. C, 1982, 129 (2), pp. 79-85
13 RAHMAN, M.A., JEYASURYA, B., and GANGOPADHYAY, A.: 'Digital differ-
ential protection of power transformers based on Walsh functions', Trans. CEA
Engineering & Operating Div. 1985, 24, Paper 85-SP-149
14 JEYASURYA, B., and RAHMAN, M.A.: 'Application of Walsh functions for
microprocessor based transformer protection', IEEE Trans., 1985, EMC-27, pp. 221-
225
15 THORP, J.S., and PHADKE, A.G.: 'A microprocessor based three phase trans-
former differential relay', IEE Trans. 1982, PAS-94, pp. 426-432
16 PHADKE, A.G., and THORP, J.S.: 'A new computer relay flux-restrained current
differential relay for power transformer protection', IEEE Trans., 1983, PAS-102,
pp. 3624-3629
17 DEGENS, A.J., and LANGEDIJK, J.M.: 'Integral approach to the protection of
power transformers by means of a microprocessor', Elect. Pow. & Energy Syst. 1985, 7,
pp. 37-47
Chapter 11
Digital line differential protection
11.1 Introduction
As power systems grow both in size and complexity, it becomes common to use
long and heavily loaded two-terminal lines as well as multi-terminal and tapped
lines [1-4]. This in turn has created difficult problems for their protection. Such
problems include:
(i) Distance relay underreach:
The infeed from another source to a fault point of a multi-terminal line
causes distance relays to measure a higher impedance than the actual line
impedance. This causes a delay in the tripping of the relay, which in turn
may result in incorrect sequential tripping at other terminals.
(ii) Distance relay overreach:
This problem arises when an external fault is fed from a number of
terminals. This causes overreach, in which the distance relay measures a
lower impedance than that of the fault loop. This can again cause false
tripping and difficulty in determining appropriately the time selectivity
between the protected and adjacent lines.
(iii) Effect of load current:
Load current can cause an inaccurate impedance measurement by
distance relays, which in turn can affect the integrity of the decision
reached.
(iv) High resistance faults:
A satisfactory operation of distance relays is not always achieved under
highly resistive fault conditions, e.g. an earth fault through a tree. This
type of fault is in many cases so similar to normal load conditions that the
distance relay may fail to recognise it as a fault condition.
(v) Pilot-wire limitation:
Pilot-wire relays can be applied successfully to overcome many of the
problems described above. However, this is true only if the length of the
line is not greater than typically 20 km because of the resistance and
capacitance associated with the pilot wire. Various means of pilot
compensation have been developed, but in general pilot-wire-based protec-
tion cannot be applied satisfactorily to many transmission circuits.
Recent developments in digital differential protection have been aimed at
overcoming these limitations. These schemes can be broadly classified into
current-based schemes and composite voltage- and current-based schemes.
Digital line differential protection 191
ix(t) iY(t) Y
iF(t)
internal
fault
FM relay characteristics
Although the FM differential relay relies fundamentally on the percentage
differential principle, its percentage restraining characteristics are often slightly
different from the conventional type applied to transformer protection. Figure
11.3 shows the basic characteristics of a typical percentage FM differential
relay. It will be seen that the characteristics consist of two sections, AB and BC.
The slope of section AB, the small current region, is chosen to permit the
detection of small internal fault currents under the existence of heavy load
conditions. It is also effective for detecting small internal fault currents under
high arc resistance conditions. On the other hand, the slope of section BC is
such that it can deal with heavy internal fault conditions. This section is also
effective in preventing false tripping during external fault conditions, which
cause CT saturation and thus generate significant differential-current signal
components.
A A
I/ V \
communication link
D(t)
operating zone
block zone
B(t)
It is important to note that differential and bias signals are commonly derived
separately for each phase of the line. In this way, many of the problems formerly
associated with the summation of phase quantities are overcome in modern
digital line differential protection arrangements.
In its simplest form, the tripping criterion adopted by the relay is that the
tripping signal is initiated when the magnitude of the differential quantity D(t)
exceeds that of the bias quantity B(t) by a certain predefined threshold value Ks:
\D(t)\-KB\B{t)\*Ks (11.3)
However, to avoid the above-mentioned problem, a modified process des-
cribed by eqn. 11.4 is used. In this case the differential current component D(t)
is checked against a time-variant threshold signal S(t):
\D(t)\*S(t) = Ks + KB\B(t)\ (11.4)
The basic arrangement typically used is similar to that shown in Figure 11.2 for
an MMM scheme, although MSS schemes are common when using this
approach.
instead of the time variant signal ±£(/) (see eqn. 11.4), i.e. KB is assumed to be
zero.
Let us first consider the external fault waveform. It can be seen that the
magnitudes of the four samples 2-5 are above the pick-up levels ± Ks. Thus,
internal fault
after the completion of the first operation of the logic, the decision counter KD is
set for an up count. However, since the polarities of these samples alternate
between positive and negative, as shown on the waveform, the second operation
of the logic results in a down count. Consequently KDy for the waveform under
consideration, stays close to zero at all times.
Consider now the internal fault waveform. It can be seen that once the
differential current has exceeded the pick-up level, it stays above that level for
an appreciable time. This means the two operations of the logic related to
comparisons of magnitude and polarities of four successive samples at a time
will allow KD to attain the required value of 4 very rapidly. For this particular
waveform, it can be seen that samples 2—8 successively indicate an up count.
The criteria of adopting a four sample check and a trip signal initiation at a
decision counter output of 4 is based on extensive studies aimed at maximising
relay stability under external fault conditions and at the same time maximising
the sensitivity to internal faults. The adoption of the bias quantity KB\B(t)\ (see
eqn. 11.4) further enhances stability under healthy conditions without signifi-
cantly affecting performance for internal faults. In essence, the trip decision
logic described makes use of further features of the relaying signals to enhance
the integrity of the decision asserted under onerous conditions, which may
otherwise cause relay malfunction.
e
|y Y
but significantly finite under internal fault conditions. Thus, for a three-
terminal line such as that shown in Figure 11.1 we have
{ = 0 healthy conditions
,r , J-.
# 0 internal fault conditions
* = 0> !>2 (11.7)
v
'
For a two-terminal line the signal Sz is equal to zero and the above equation
reduces to
{ = 0 healthy conditions
^ O i rinternal, fault
c..u -conditions
__^:._. * = 0,l,2 (11.8)
where Sx, SY and Sz are the terminal signals at ends X, Y and Z, and k is the
mode number.
11.3.2 Formation of terminal signals
113.2.1 Two-terminal lines [6]
Assuming a distributed-parameter, transposed three-phase transmission line,
the modal voltage and current relationships in the frequency domain can be
described by eqn. 8.30:
2
), - ^ = ( y » ) ^ * = 0, 1, 2 (11.9)
By substituting the values of Kx and K2 obtained from eqns. 11.12 into eqns.
11.13, we finally obtain
^(aO+ZjPM/pV^e^ * = 0, 1, 2 (11.14)
By using the current and voltage reference directions shown in Figure 11.5, i.e.
by reversing the direction of the current IY at end Y, eqn. 11.14 becomes
n\(^)-Z0{(D)lf((o) = ^l[nK(o) + Z^\(o)lf((D)^ * = 0, 1,2 (11.15)
or S f + 5^ = 0 * = 0, 1,2 (11.16)
where Sx{k) is equal to the right-hand side of eqn. 11.15 and S$ is equal to the
left-hand side negated.
113.2.2 Three-terminal lines
The basic theory laid down above for the two-terminal lines can be easily
extended to include teed (or three-terminal) feeders. Consider again Figure
11.1, which shows that a three-terminal line basically consists of three separate
line sections. These are sections XT, YT and ZT. For each of these sections an
equation similar to eqn. 11.14 can be derived in terms of the current and voltage
at the line ends. When the resulting equations are combined together (and
applying the fact that the sum of currents at the tee point T is zero), we obtain
Sf + Sf + S$ = 0 * = 0, 1,2 (H.17)
where
Sf = AxV<£\a>) + BxZW(a>)l$(fi>)
Sf = [AY- Br) Vf(a>) + (Ar+ BY)Z^{(o)I^{w)
Sf = (Az-Bz) Vf(o,) + (Az
Ay=
11,4 References
1 AIEE Working Group of the Line Relay Protection Sub-Committee: 'Protection of
multiterminal and tapped lines', AIEE Fall Meeting, October 1960, Paper CP60-1274
2 'IEEE Study Committee report on protection aspects of multi-terminal lines', IEEE
Report 79, THOO56-2-PWR, 1979
3 KITAGAWA, M. et al.: 'Newly developed FM current-differential carrier relaying
system and its field experience', IEEE PES Winter Meeting, 1978, Paper F78, 291-
297
4 YAMAURA, M., MASUI, M., and OKITA, Y.: 'FM current differential carrier
relaying', Developments in power system protection, (IEE Conf. Publ. 185, 1980,
pp. 156-160)
5 AGGARWAL, R.K., and JOHNS, A.T.: 'The development of a new high speed three-
terminal line protection scheme', IEEE/PES Summer Meeting, 1985, Paper 85 SM
3200-3207
6 AGGARWAL, R.K., and JOHNS, A.T.: 'A differential line protection scheme for
power systems based on composite voltage and current measurements', IEEE/PES
Winter Meeting, 1989, Paper 89WM 053-0 PWRD
7 AGGARWAL, R.K., and JOHNS, A.T.: 'New approach to teed feeder protection
using composite current and voltage signal comparison', Developments in power
system protection (IEE Conf. Publ. 302, 1989, pp. 125-129)
Index
Distance relays 1, 2, 61, 69, 70, 190 Impedance 1, 60-62, 66-69, 72, 79, 101, 102,
Distribution systems 79 126
Dual-slope converter 53—55 variation with time 7
Impulse responses 174-176, 178
Earth mode 126, 135 Inrush detection algorithm 180, 181
EHV applications 152 Integral least squares fit 92-95
Elliptical trajectory based technique 136, 138- Integrators 53, 54
140, 166-168 Internal fault detection 148-170
Errors 13, 46, 56, 62-65, 76, 79, 92, 93, 98, Interpolation formulas 12-15
185 backward 14
central difference 15
forward 12, 13
Fault direction 153, 154, 164-170
Gregory-Newton 13, 14
Fault discrimination 178-181
Laplace-Everett 15
Fault location 4, 61, 127-140, 143-146
Newton-Gauss 15
forward faults 131-135, 144
Stirling 15
reverse faults 129, 130, 132, 134, 135, 145
Fault trajectory 139, 140
Kalman filters 1, 2
Finite difference techniques 7-12
Karrenbauer transformation 125, 164, 165
Finite-duration impulse response (FIR) filter
174-179
Least squares methods 4, 18-22, 92-102,
extraction of fundamental and second-
179-181
harmonic components 175-178
integral 92-95
fault discrimination 178, 179
multi-variable series 99-101
Finite-duration impulse response (FIR) filter
power series 95-99
based algorithms 174-179 LFDC digital relay 164
Flexibility 3 Logic circuits 154, 161, 194-196, 198, 199
Flipflops 56 Lossless single-phase lines 119-123, 140, 141,
Flow charts 56, 57, 164, 165 143-146
Flux-restrained current differential relaying Low-pass filters 40-43, 45, 46, 110
183-187
Forward differences 8-10 Master-master-master (MMM) system 191,
Forward interpolation 12, 13 192, 194
Forward waves 121 Master-slave-slave (MSS) system 192, 194
Fourier analysis 24—30 Memory 3, 40, 57, 70
Fourier analysis based algorithm 72-84, 181 — Microprocessors 187
183 Microwave communication links 192
Fourier-transform based 79-84 Modal analysis 123-126, 135, 142, 146
full-cycle window 73-75 Modal components 124, 194, 196, 197
half-cycle window 75-77 Modal current based protection scheme 193-
sub-cycle window 76-79
196
Fourier coefficients 26, 27, 34-37
Modal equivalent circuits 123
Fourier kernels 84
Modal propagation constant 125, 126, 197
Fourier series 1, 4, 24-27, 72, 73
Modal surge impedance 125, 126, 197
complex form 25
Modal transformation matrices
trigonometric form 24, 25
Clark 125
Fourier transform 1, 28-30, 79-84
Karrenbauer 125
Fourier-transform based algorithm 79—84
Wedepohl 125
Fourier-Walsh transformation matrix 37, 85
Multiplexers 40, 43, 44, 49, 50, 58, 187, 188
Fractional cycle windows 75-79
logic circuits 49, 50
Frequency domain 46, 72, 197
Multi-terminal lines 190-199
Frequency modulation (FM) relays 191-193
Multi-variable series least squares technique
Frequency responses 75-77, 175, 177, 182
99-101
Generator protection 171
Noise 2, 65, 67, 160, 161, 164
Numerical differentiation 16-18
Haar function 183
Nyquist frequency 44
Harmonic-restrained differential relaying 172—
174 Optical fibres 192
Harmonics 92, 99, 100, 107-110, 172-183 Out-of-step blocking 70
202 Index