0% found this document useful (0 votes)
13 views44 pages

Chapter 5

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
13 views44 pages

Chapter 5

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 44

Chapter 10

Digital differential protection of transformers

10.1 Introduction

This chapter gives a brief general review of the principles of transformer differential
protection (TDP). This is followed by an explanation of the application of digital
techniques and the algorithms that have been developed specifically for the appli-
cation of transformer protection. The algorithms covered include finite-­duration
impulse response (FIR) filters, least-­squares (LSQ) curve fitting, the digital Fourier
algorithm, and the flux-­restrained current differential algorithm.
Finally, the basic hardware arrangement for implementing digital techniques
for 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 pro-
tected zone.

10.2 Principles of transformer protection

10.2.1 Basic principles
The most commonly encountered transformer protection arrangement is based
on the differential current principle [1–4]. This can be illustrated by reference to
Figure 10.1, in which the primary and secondary currents (Ip and Is) are compared
after being reduced by their corresponding current transformers (CTs). The primary
and secondary CTs 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 cur-
rents is disturbed and a differential current signal (ID = 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. The imbalance under
healthy conditions can also be caused by the mismatch between the CTs or by their
258 Digital protection for power systems

Figure 10.1   Basic principle of TDP

saturation. It is therefore important that this situation should be taken into account in
the engineering of TDP 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 10.2(a) together with a
typical operating characteristic (Figure 10.2(b)). It is evident from Figure 10.2(b)
that the differential current required to operate the relay increases with the bias cur-
rent. 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 10.2(b) is arranged to allow a sig-
nificant 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 trip-
ping occurs.
In some cases, the operating current is expressed as a percentage of the restrain-
ing 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 10.2(b) the measurands are subscripted.
In modern microprocessor-­based TDP relay, the restraining characteristic can
be generated with different shapes. Figure 10.2(c) shows an example of such shapes,
which basically consists of two straight lines with slops S1 and S2 connected by a
smooth curve section. The characteristic of this shape will be extensively used in the
discussion of section 10.7 related to the enhancement of TDP to improve its security
and dependability.

10.2.3 Harmonic-restrained differential relay


The differential protective scheme described in section 10.2.2 is generally very sat-
isfactory under normal operating conditions but is prone to false operation during
energization of the transformer. Under this condition, an inrush current flows only
through one winding (the winding being energized), which consequently appears to
the relay as an internal fault.
Digital differential protection of transformers 259

Figure 10.2  Biased differential relay (a) basic arrangement, (b) typical


operating characteristic, and (c) operating characteristic formed
from two linear sections with slopes S1 and S2 connected by a
smooth curve
260 Digital protection for power systems

This problem can be overcome by using the fact that inrush currents usually
contain harmonics [5, 6]. Of these, the second harmonic is usually predominant
under all energization conditions. The protection can therefore be designed so that
it is prevented (restrained) from tripping if the magnitude of the second-­harmonic
component of the operating current is greater than a certain prescribed percent-
age of the fundamental component. Fifth harmonics, which are generated under
over-­excitation conditions (caused, for example, by temporary overvoltages), are
also used in some equipment to prevent the relay from tripping under energization
conditions.
It will be apparent that harmonic-­restrained current differential transformer
protection basically uses signals formed from fundamental (or power frequency)
and harmonic components. The power-­frequency components predominate when
an internal fault occurs, and the harmonic and bias signals are used to restrain the
protection during faults, switching operations, and over-­excitation conditions.
With reference to Figure 10.2(a), the differential current ID is commonly formed
from the fundamental frequency components of the primary and secondary currents
Ilp and I1s, respectively, such that
‍ ID = I1p  I1s‍ (10.1)
The restraining signal (IR1), on the other hand, is commonly formed from the funda-
mental components of the primary and secondary components such that
‍ IR1 = I1p + I1s‍ (10.2)
The other restraint signals are the second-­harmonic restraint IR2 and fifth-­harmonic
restraint current IR5, which are commonly formed from the magnitudes of differential
currents of any second and fifth harmonics detected. A commonly used technique
involves summing any such harmonic components in each phase:
‍ IR2 = |ID2a | + |ID2b | + |ID2c |‍ (10.3)
and
‍ IR5 = |ID5a | + |ID5b | + |ID5c |‍ (10.4)
where ID2a, ID2b, and ID2c are the differential currents due to second-­harmonic com-
ponents in phases “a,” “b,” and “c,” respectively, and ID5a, ID5b, and ID5c are the cor-
responding differential currents due to fifth-­harmonic components.
Tripping and restraining signals can be derived using analog 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.

10.3 FIR filter-based algorithms

These algorithms are based on using FIR filters to estimate the magnitude of the
fundamental and second-­harmonic components [7, 8]. The estimates are then used
to calculate the ratio of the magnitude of the second harmonic to fundamental. If
this ratio is found to be greater than a certain value, an inrush current condition
Digital differential protection of transformers 261

is assumed. This algorithm is often designed to avoid multiplication and division


as these operations are rather unwieldy to perform at both software and hardware
levels.

10.3.1 FIR filter characteristics


The filters used are assumed to be characterized by means of FIRs 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 to find the responses of the filters to 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 components of
the input, four filters are required: two for the fundamental and two for the second-­
harmonic component. Their impulse responses are S1(t), C1(t), S2(t), and C2(t)
(Figure 10.3), which in turn are defined by the following equations:
8
 <+1 0  t  T/2
S1 t = (10.5)
:1 T/2  t  T
‍ 8 ‍
 < +1 0  t  T/4, 3T/4  t  T
C1 t = (10.6)
:1 T/4  t  3T/4
‍ 8 ‍
 <+1 0  t  T/4, T/2  t  3T/4
S2 t = (10.7)
:1 T/4  t  T/2, 3T/4  t  T
‍ 8 

 <+1 0  t  T/8, 3T/8  t  5T/8, 7T/8  t  T
C2 t = (10.8)
:1 T/8  t  3T/8, 5T/8  t  7T/8
‍ ‍
It will be apparent that S1(t) and C1(t) are impulse responses for the sine and cosine
parts of the fundamental component, and S2(t) and C2(t) are impulse responses for
the sine and cosine parts of the second-­harmonic components. Equation (2.73) can
be applied to (10.5)–(10.8) to find the system frequency responses of the four filters,
which are consequently given in
   
  2 j!T/2 !T
FS1 ! = e cos 1 (10.9)
‍ j! 2 ‍
    
  2 j!T/2 !T !T
FC1 ! = e cos  2 sin (10.10)
‍ j! 2 4 ‍
     
  2 j!T/2 !T !T
FS2 ! = e cos  2 cos +1 (10.11)
‍ j! 2 4 ‍
      
  2 j!T/2 !T 3!T !T
FC2 ! = e sin  2 sin + 2 sin (10.12)
‍ j! 2 8 8 ‍
262 Digital protection for power systems

Figure 10.3   Impulse response of fundamental and second-­harmonic FIR filters


Digital differential protection of transformers 263

where T = 2π/ω0 is one period of the system frequency f0.


The amplitudes of the above frequency responses plotted against the ratio ω/ω0
are shown in Figure 10.4.

10.3.2 Extraction of fundamental and second-harmonic


components
To extract the fundamental component of a current, it is necessary to determine
the responses of the filters to the transformer current. The responses of filters 1–4
to transformer current after one cycle of power frequency (t = T) can be found by
convolving the impulse responses of these filters with the transformer current (as an
input to the filters) and evaluating at time t = T.
As always in digital protection, the current inputs to the filters are in the form
of samples, and since the impulse responses of the FIR filters have the values of
either +1 or −1 at any instant of time [as defined by (10.5)–(10.8)], the multiplica-
tion operation required by the convolution process is simply a sign change. Thus, the
outputs of filters 1–4 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 Δt =
2π/(Nω0), and ik = i(tk) is the kth sample at any time t = kΔt. The result of time-­
discrete convolution of the samples ik with the impulse responses defined by the
previous equations will then be given as
  N/2
P 
S1 t = ik  ik+N/2 (10.13)
‍ k=1 ‍
  N/4
P   
C1 t = ik  ik+N/4 + ik+N/2 + ik+3N/4 (10.14)
‍ k=1 ‍
  N/4
P 
S2 t = ik  ik+N/4 + ik+N/2  ik+3N/4 (10.15)
‍ k=1 ‍
 P N/8     
C2 t = ik  ik+N/8 + ik+N/4 + ik+3N/8 + ik+N/2
k=1
   (10.16)
‍  ik+5N/8 + ik+3N/4 + ik+7N/8 ‍

10.3.3 Discrimination between inrush and internal fault currents


The criteria used to distinguish between inrush currents and currents due to internal
faults are based on evaluating the ratio ε, which is calculated from the ratio of the
larger of the two components of each pair of filter outputs as described in (10.13)–
(10.16) such that
 
max |S2 |, |C2 |
"=   (10.17)
‍ max |S1 |, |C1 | ‍
264 Digital protection for power systems

Figure 10.4  Magnitudes of the frequency responses of fundamental and second-­


harmonic filters: (a) filter 1, (b) filter 2, (c) filter 3, and (d) filter 4
Digital differential protection of transformers 265

An internal fault, as distinct from a current inrush condition, is characterized by a


high value of the fundamental component of differential current and a relatively low
value of second harmonic. Conversely, inrush conditions are associated with rela-
tively high second-­harmonic current components. It has been suggested [7] that the
values of ε for the detection of internal faults are
0  "  0.146 for X/R=5
0  "  0.093 X/R = 10
‍ 0  "  0.054 X/R = 20 ‍
where X/R is the system reactance-­resistance ratio.
A typical threshold value ε0, which is convenient from a computational point
of view, is ε0 = 0.125. In such a case, any value ε that is greater than 0.125 would
be taken to indicate an inrush condition and would inhibit the operation of the
protection.

10.4 LSQ curve fitting-based algorithms

LSQ curve fitting methods have been discussed in detail in Chapter 6. These
methods are basically used to extract the fundamental and/or any other harmonic
­components [9, 10]. Since the relative values of second-­harmonic and fundamental
components of the differential current play an important role in the TDP, LSQ tech-
niques can be used either to detect the second harmonic in an inrush current (which
is later used to block the differential protection) or to find the ratio between the fun-
damental and second-­harmonic components of the differential current, which in turn
is used to differentiate between inrush current and internal faults.

10.4.1 Basic assumptions and algorithm derivation


Since inrush currents can in fact contain significant harmonics up to the fifth, the
differential current can be described by a waveform containing a decaying DC com-
ponent fundamental and harmonic components up to the fifth order:
 P5  
i t = I0 et/ + Im cos m!0 t + m (10.18)
‍ m=1 ‍
where τ is the time constant of any decaying DC component.
If the decaying DC component is approximated by the first two terms of its
Taylor expansion, then the current sample i(tk) measured at tk can be expressed as
    P5   P5  
i tk = I0  I0 / tk + Im cos m cos m!0 tk + Im sin m sin m!0 tk
‍ m=1 m=1 ‍
 (10.19)
or
‍ ik = a1 x1 + a2 x2 +    + a11 x11 + a12 x12‍
266 Digital protection for power systems

with
a1 = 1 x1 = I0
a 2 = tk x2 = I0 /
) ) .
an+2 = cos n!0 tk xn+2 = In cos n
n = 1, : : : , 5 n = 1, : : : , 5
‍ an+7 = sin n!0 tk xn+7 = In sin n ‍
To solve for the unknowns xn(n = 1, . . ., 12), m equations can be constructed from N
current samples. The resulting equations can be written in matrix form:
A „ƒ‚…
„ƒ‚… X = „ƒ‚…
i
(10.20)
‍ N12 121 N1 ‍
or
‍ X = Bi ‍ (10.21)
where B = {AT · A}−1 is the pseudo-­inverse of A and AT is the transpose of matrix A
(see section 2.5.1.1).
It follows that the real and imaginary parts of the fundamental and second har-
monic can be calculated from (10.21):
PN
I1 cos 1 = x3 = b(3, n)in
n=1
P
N
I2 sin 1 = x8 = b(8, n)in
n=1
PN (10.22)
I2 cos 2 = x4 = b(4, n)in
n=1
P
N
I2 sin 2 = x9 = b(9, n)in
‍ n=1 ‍
where xk is the kth element of vector X and b(k, n) is the element of the kth row and
nth column of matrix B.
Using the foregoing equations, the amplitude of the fundamental and the second
harmonic can be calculated as
q 2  2
In = In cos n + In sin n n = 1, 2 (10.23)
‍ ‍
Similar techniques can be applied in extracting, where necessary, the fifth-­harmonic
components.

10.4.2 Basis of discrimination between inrush and internal fault


currents
Discrimination between an inrush and an internal fault condition is often based on
an inrush-­detection algorithm. This algorithm is based on comparing the second-­
harmonic and fundamental components I2 and I1, which are present in the differential
current. The values of these currents, determined according to (10.23), are then used
to calculate what is called the “second-­harmonic ratio (SHR),” which is defined by
I2
SHR =
‍ I 1 ‍ (10.24)
Digital differential protection of transformers 267

x4
� b (4, n)i�

�𝑥𝑥𝑥𝑥� � + 𝑥𝑥𝑥𝑥� � I2
x9
� b (9, n)i�
SHR
+
x3
� b (3, n)i� I1

�𝑥𝑥𝑥𝑥� � + 𝑥𝑥𝑥𝑥� �
x8
� b (8, n)i�
i�

Figure 10.5  Schematic representation of the determination of the second


harmonic (SH) ratio where in is the nth sample, b(k, n) is the
element of the kth row and nth column of matrix B [see (10.21)],
and ∑ is the sum from n = 1 to N, i.e., the number of samples per
fundamental period

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 x4, and imaginary parts (x8 and x9) of the fun-
damental and second-­harmonic components used in calculating the SHR. The fre-
quency characteristics of the four filters are shown in Figure 10.6 where it can be
seen that the unwanted harmonics are filtered out.

10.5 Fourier-based algorithm

10.5.1 Filtering of harmonics
This algorithm is based on the fact that the fundamental, second-, and fifth-­harmonic
components contained within an inrush current can be digitally extracted using the
Fourier approach given in detail in section 5.2.1 [11].
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(kΔt). The real
and imaginary parts of the nth harmonic (an and bn) can be found by using (5.5) and
(5.7). In terms of current samples starting at the rth sample, an and bn can then be
expressed as follows:
268 Digital protection for power systems

Figure 10.6  The magnitudes of the frequency responses of digital harmonic


extraction filters: (a) x3(ω), (b) x8(ω), (c) x4(ω), and (d) x9(ω)

 
2 X
r+N1
 2k
anr = ik cos n (10.25)
‍ N k=r N ‍
 
2 X
r+N1
 2k
bnr = ik sin n . (10.26)
‍ N k=r N ‍
 
r r
The magnitude |I‍ n |‍of the nth harmonic and its phase angle ‍ n ‍can thus be calcu-
lated as explained previously. Thus
Digital differential protection of transformers 269
r  2    2

r r
|I | =
n
r
an + bn

. (10.27)
 bnr
 n
r
= tan 1

r
‍ an ‍
For the transformer protection application, n takes the values of 1, 2, and 5 for fun-
damental, second, and fifth harmonics, respectively. The result can be updated itera-
tively as each new sample becomes available. This is done by dropping the earliest
sample and adding the new sample:
 
   2  2r
anr+1 = anr + iN+r  ir cos n (10.28)
‍ N N ‍
 
   2  2r
bnr+1 = bnr + iN+r  ir sin n (10.29)
‍ N N ‍

where ir and iN+r are the oldest and newest samples, respectively.
Having determined the fundamental, second-, and fifth-­harmonic components,
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 perfor-
mance and include the use of rectangular transforms [12], Walsh functions [13], and
the Haar function [14].

10.6 Flux-restrained current differential relay

This algorithm basically uses the flux-­current relation of the transformer to obtain
the restraint function [15, 16]. If the flux could be estimated correctly, it would pro-
vide a sound basis for detecting over excitation as well as magnetizing inrush condi-
tions. An important feature of this technique is that it requires fewer computations
than that based on the Fourier analysis.

10.6.1 Development of the algorithm


Consider a simple two-­winding transformer (basically the same analysis can be
applied to a multi-­winding transformer) such as that shown in Figure 10.7. It will
be assumed that winding resistance is negligible. The relation between the primary
applied voltage vp, the primary current ip, and the mutual flux linkages ψ of the
transformer is given by
 
dip t d t 
Lp + = vp t (10.30)
‍ dt dt ‍
where Lp is the leakage inductance of the primary winding. By rearranging (10.30)
and integrating from t1 to t2, we obtain the flux linkages at these times such that
270 Digital protection for power systems

Figure 10.7   Two-­winding single-­phase transformer

    
t2     
ˆ
t2  t1 = vp t dt  Lp ip t2  ip t1 (10.31)
‍ t1 ‍
Applying the trapezoidal rule to the integral part of (10.31) results in
    1          
‍ t2 ' t1 + 2 t2  t1 vp t2  vp t1  Lp ip t2  ip t1 . ‍ (10.32)
If the voltage and current waveforms are sampled such that the sampling interval is
equal to Δt, then at the kth sample, (10.32) can be expressed using sample notation
as
1    
k = k1 + t vp,k  vp,k1  Lp ip,k  ip,k1 (10.33)
‍ 2 ‍
where ip,k and vp,k are the kth samples of the primary current and voltage respectively.
Equation (10.33) can therefore be used to calculate the mutual linkage flux ψ of
the transformer. On the other hand, the differential current can be calculated from
the primary and secondary currents. Thus, at time tk, the differential current idk is
given as follows:
‍ idk = ip,k  is,k ‍ (10.34)
where ip,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 mag-
netizing current of the transformer. Therefore, if it is assumed that the flux linkage
computed by (10.33) is a true representation of the actual flux in the transformer, the
samples of the differential current and flux linkage (idk, ψk) are expected to fall on the
open-­circuit magnetizing curve of the transformer.
In the first phase of the algorithm, the principle of the percentage-­biased dif-
ferential 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
(idk, ψk). If this point does not fall on the id − ψ 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 magnetizing characteristic of a transformer and the (idk, ψk) relationship for
an internal fault inside the transformer. For the internal fault condition, the terminal
Digital differential protection of transformers 271

Figure 10.8  Transformer magnetizing curve: (a) fault and non-­fault region and
(b) effect of remanent flux

voltage (and hence ψk) is much smaller than is the case under inrush conditions.
Consequently, there are distinct regions in the i—ψ plane that define the fault or
non-­fault status of the transformer.
The aforementioned 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 id − ψ characteristic of the transformer may vary, as shown
272 Digital protection for power systems

Figure 10.9   Fault and non-­fault regions in the dψ/di–i plane

in Figure 10.8(b). This situation makes it impossible to differentiate between fault


and no-­fault regions in the i–ψ plane if the flux linkages are computed according
to (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 that is
determined by the slope dψ/di rather than the flux ψ itself. From (10.33), we then
obtain
 
d k  k1 1 Vp,k + Vp,k1
 = t  Lp (10.35)
‍ di k ik  i k1 2 ip,k  ip,k1 ‍
Figure 10.9 shows that there are two regions in the dψ/di–i plane. The first region
corresponds to a fault condition or an operation in the saturated part of the mag-
netization curve. The second region, which is significantly removed from the first,
designates an operation on the magnetizing curve in the unsaturated part. It has been
observed that for internal fault conditions, the current samples and corresponding
flux derivative dψ/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 kr which is increased each time a sample pair (ik,
(dψ/di)|k) falls in 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 + 1 if the differentials current indicates a trip condition and if the pair


[ik , (d /di)|k ] is in region 1
( .
kr  1 if kr > 0, and the pair [ik , (d /di)|k ] is in region 2
kr =
‍ kr if kr = 0 ‍
 (10.36)
Digital differential protection of transformers 273

The value of kr is found to increase almost monotonically for fault conditions while
it never reaches greater than a threshold value kr,max for all non-­fault conditions.
However, the threshold kr,max is found to depend, among other things, on the sam-
pling rate. Therefore, kr,rmax must be determined experimentally.

10.7 Enhancement of TDP to improve its security and


dependability

As discussed in section 10.2.2, one of the causes of the imbalance between the trans-
former’s primary and secondary currents is the saturation of CTs. Such saturation of
CTs can be caused by some external faults with high DC offset and high X/R system
time constant [17]. This in turn can lead to the differential/restraint currents ratio to
be above the preset characteristic driving it into the operating region. Consequently,
the TDP operates and causes unnecessary tripping of the transformer.
However, modern TDP is built-­in microprocessor-­based relays such as the prac-
tical TDP commercially known as 87T. This makes it possible to perform several
tasks including signal processing, filtering, currents compensation, and computation
of differential and restraint currents. This in turn enables TDP to perform with a cer-
tain level of sensitivity, security, and dependability during faults. Microprocessor-­
based relays also make it possible to use algorithms that use simple characteristic
to distinguish between external and internal faults, as well as more complex algo-
rithms, which involve more protection principles. The advantage of having a more
complex algorithm is their ability to provide better security during external faults by
tolerating CT saturation and consequently inhibiting the operation of the protection.

10.7.1 Types of faults to which power transformers are subjected


Power transformers are subject to internal as well as external faults, as shown in
Figure 10.10(a). Internal faults include faults to which transformer windings, tank,
insulation, bushings, and tap changer are subjected to and all other faults that occur
within the TDP zone defined by the CTs. On the other hand, external faults, as the
name implies, include all types of faults that occur outside the TDP zone. Obviously,
some of these external faults may occur close to the transformer location, while oth-
ers are far from it.
The CTs are easily saturated if the magnitude of fault currents is accompanied
by full offset and long DC time constant. Additionally, the CT saturation becomes
even worse if the impedance of the burden connected at its secondary side is high.
Figure 10.10(b) shows examples of ID/IR trajectories of internal and external faults
with and without CT saturation. As shown in Figure 10.10(b), trajectory No. 1 is
due to the differential/restraint currents ratio during an external fault with light CT
saturation with the highest point falling into the safe no-­operation (Block) region
below slope S2. Trajectory No. 2 is due to an external fault with severe CT satura-
tion, whereby the distortion caused by the CT saturation causes the corresponding
ID/IR trajectory to cross the characteristic into the operating region. This obviously
leads to mal-­operation of TDP causing unnecessary tripping of the transformer it
274 Digital protection for power systems

Figure 10.10  Differential/restraint trajectories during internal and external


faults with and without CT saturation (after Reference [17])

protects. Trajectory No. 3 is due to an internal fault that caused light CT saturation,
while trajectory No. 4, which is a straight line, is due to an internal fault without CT
saturation.

10.7.2 Securing TDP against external faults


Securing TDP against external faults consists of two steps [17]. The first step is con-
cerned with detecting an external fault, while the second step deals with determining if
Digital differential protection of transformers 275

an internal fault is developed during the occurrence of the external fault by checking the
per-­phase angle between the primary and secondary windings of the transformer. The
second step is executed if the actual differential/restraint current ratio due to the external
fault falls into the operation region (see Figure 10.10(b)) as explained in (i) below. The
outcome of the second step confirms whether the fault has evolved into an internal fault
or not. In case the fault is an external fault, the TDP 87T is prevented from operation.

(i) Execution of external fault detection

When a fault occurs, the detection of external faults is achieved by the follow-
ing steps.

1. Determine the resulting maximum restraint current (IRmax) value.


2. Compare IRmax with the restraint current threshold (RCT) value, where RCT =
IRthr. The latter is computed from the operating IR-­IDcharacteristic of the relay as
the average value of the restraint currents corresponding to the breakpoints BP1
and BP2. Thus, IRthr falls between the values corresponding to these breakpoints
(see Figure 10.11).
3. Determine the differential/restraint currents ratio during the fault and check if
it falls below the average slope of sections S1 and S2 of IR-­IDcharacteristic (see
Figure 10.11). The point of the average slope and the average breakpoint fall

Figure 10.11  Logic flags during internal and external faults (after Reference
[17]). FTY = Fault type (flag), OPE = Operate (flag), RCT =
Restraint current threshold (flag)
276 Digital protection for power systems

around the middle of the smooth intermediate curve connecting sections S1 and
S2 (see Figure 10.11).
4. If IRmax is equal to or greater than IRthr, trigger the RCT flag, i.e., set RCT = 1, and
maintain it in this position for a sufficient time to allow the fault IR-­ID trajectory
to be fully developed, i.e., either crosses the characteristic into the operating
region, in which the case trigger operate (OPE) flag, i.e., set OPE = 1, or stay in
the Block non-­operation zone, otherwise, if IRmax is less than IRthr set RCT = 0.
5. If RCT = 1 and OPE = 1, initiate fault type (FTY) flag check [see section (ii) ]

(ii) Fault type flag check

The fault type (FTY) flag is set to either 0 or 1 (see Figure 10.11) depending
on whether the fault is external or internal, respectively. In case of external fault,
whereby the angle between the phase voltages of the corresponding primary and
secondary compensated currents windings is usually greater than 90°, FTY = 0. On
the other hand, if the fault is internal, whereby the angle between the phase voltages
of the corresponding primary and secondary compensated currents windings is usu-
ally less than 90°, FTY =1.
It will be noted that IR-­ID trajectory due to an external fault always crosses
the RCT (IRthr) value whether such a fault causes severe or light saturation (see
Figure 10.11), which leads to setting RCT=1 [as per (i)]. However, the fault type
flag check is initiated only for the case of external fault that causes severe CT satu-
ration and consequently causes its corresponding R IR-­ID trajectory to cross into the
operating region (trajectory No. 2 in Figure 10.10). Finally, the operate (OPE) flag
is triggered, i.e., OPE = 1 as soon as the IR-­ID trajectory due to a fault (whether inter-
nal or external) crosses the characteristic into the operating region.
Figure 10.12 shows the security logic that can be applied to transformer differen-
tial protection relay type 87T against external faults.
Based on the aforementioned discussion, the two trajectories of Figure 10.10
related to internal faults do not trigger the RCT flag, and the protection does not need
to initiate FTY check in order to produce operation and trip action.
Figure 10.13 shows a flow chart for the implementation of securing the trans-
former differential protection against external faults using the various logic flags
discussed above.

10.8 Wavelet-based transformer protection methods

As discussed in Chapter 2, wavelet transform (WT) is a powerful tool for analyz-


ing transient signals. WT can identify the time localization of different frequency
components of a given signal. The analyzing functions (or wavelets) used by WT
adjust their time widths according to their frequency such that high-­frequency wave-
lets have narrow windows, while low-­frequency wavelets have wide windows. This
makes them powerful in discriminating between the transformer’s internal fault and
inrush currents. In studying transformer protection using wavelets, it has been found
Digital differential protection of transformers 277

Figure 10.12  Security logic that can be applied to relay type 87T (after
Reference [17])

that the window for internal fault currents is narrow, while that for inrush currents
is very wide [18].
It has been shown in Chapter 2 that discrete WT can be used to decompose a
transient signal into its frequency components using the “Quadrature mirror filters,”
i.e., convolving the signal with filter coefficients, called low-­pass filter decomposi-
tion (LFD) and high-­pass filter decomposition (HFD) coefficients. This technique
has been used in discriminating between transformer inrush magnetizing currents
and fault currents [19].
Thus, the process of discriminating between the transformer’s inrush currents
and internal faults using wavelet analysis involves the decomposition of transformer
three-­phase currents signals ia, ib, and ic, simply by passing each of these currents
through two quadrature mirror filters, i.e., convolving the signal with filter coef-
ficients, namely, LFD and HFD coefficients, and by down-­sampling the result, they
emerge as two components called low-­frequency and high-­frequency coefficients.
In multilevel wavelet analysis, the decomposition process is iterated, with
successive low-­frequency components being decomposed in turn, so that the sig-
nal under consideration is broken down into many lower-­resolution components.
Figure 10.14 shows the four-­level wavelet decomposition of the line currents ia, ib,
and ic suggested in Reference [19]. The figure shows the application of the four-­level
decomposition process using wavelet analysis to the line current ia. Obviously, the
same process using the same figure can be applied to line currents ib and ic by simply
replacing ia by ib and then by ic.
This technique has been tested on a power system modeled as a two-­source
power system interconnected by a circuit consisting of a three-­phase transformer
278 Digital protection for power systems

Figure 10.13  Flow chart of executing various logics during internal and


external faults that ensures securing the TDP against external
faults

and a transmission line, as shown in Figure 10.15. The conducted tests included
(i) magnetizing inrush currents without a fault, (ii) phases BC to ground fault at
bus 2, (iii) simultaneous magnetizing inrush current and phase A to ground fault at
the transmission line middle point (TLMP), (iv) simultaneous magnetizing inrush
Digital differential protection of transformers 279

ia(t) Details =
Coefficients Coefficients
vector x ia(t) vector length
LD1ia(t) HD1ia(t)

HF Level 1
HD1ia L(ia)
LD2ia(t) HD2ia(t)

HF Level 2 L(HD1ia)
LD3ia(t) HD3ia(t) HD2ia

L(HD2ia)
HD4ia(t) HF Level 3
LD4ia(t) HD3ia

L(HD3ia)
HF Level 4
HD4ia
L(HD4ia)

LF Level 4
LD4ia
L(LD4ia)

LDi = Low-pass filter decomposition coefficient at ith level


HDi = High-pass filter decomposition coefficient at ith level
L(HDiia) = Length of HF decomposition coefficient at ith level
L(LD4ia) = Length of LF decomposition coefficient at 4th level

Figure 10.14  Four-­level wavelet decomposition of transformer line currents ia,


ib, and ic (after Reference [19])

current and a fault between phases B–C at TLMP with 10 Ω and 100 Ω fault resis-
tances, and (v) simultaneous magnetizing inrush current and phases BC to ground
fault at TLMP.
This technique proved to be successful in differentiating between magnetizing
inrush and fault conditions in less than half power frequency cycle.
It is important to note that the mother wavelet used in the investigation is
Daubechies (db4). It has been reported that Daubechies (db) wavelet is one of
280 Digital protection for power systems

Figure 10.15  Two-­source power system used in the application of wavelet


analysis to discriminate between transformer’s inrush currents
and faults (after Reference [19])

the most widely applied mother wavelets suitable for a wide range of power sys-
tem applications [20]. However, it has been reported in the literature that other
mother wavelets have also been used in investigating the application of wavelet
analysis to the protection of power transformers. For example, the investigators in
Reference [21] have employed Coiflets 6 (Coif 6). Wavelet packet transform has
also been applied to the protection of power transformer [22].

10.9 Basic hardware of microprocessor-based transformer


protection

Figure 10.16(a) shows the connection of a microprocessor-­based protective


scheme to a three-­phase, two-­winding Δ/Y transformer [23]. The three-­phase cur-
rents at the primary and secondary sides are first reduced using CTs of suitable
ratio and then connected to the microprocessor-­based system. Figure 10.16(b)
shows the basic interfacing arrangement to the microprocessor system where the
currents obtained from the secondary side of the CTs are fed to analog amplifiers
for signal amplification. The currents are then sampled by sample-­and-­hold (SH)
circuits and digitized by an analog-­to-­digital converter (ADC) before being pre-
sented to the microprocessor. The SH, multiplexer (MUX), and A/D and D/A con-
venors are controlled by software. At the microprocessor, the data are processed
using the particular algorithm chosen. In the event of an internal fault, a trip signal
is sent through the D/A converter to trip the circuit breakers on both sides of the
transformer.
It will be noted that unlike conventional schemes, no interposing CTs are
required for phase rotation, as this is readily implemented by software.
Digital differential protection of transformers 281

Figure 10.16  Microprocessor-­based protective scheme for the protection of


a three-­phase, two-­winding transformer: (a) connection and
(b) typical interfacing arrangement. CB = circuit breaker; SH =
sample and hold; MUX = multiplexer; ADC = analog-­to-­digital
converter; DAC = digital-­to-­analog converter
282 Digital protection for power systems

10.10 Summary

This chapter is devoted to discussing the digital differential protection of transform-


ers. Principles of transformer protection are covered in section 10.2, which include
basic principles, biased differential relaying, and harmonic-­restrained differential
relay.
The methods/algorithms that can be used for digital differential protec-
tion of transformers covered in this chapter include (i) FIR filter-­based algo-
rithms under which the topics discussed are FIR filter characteristics, extraction
of fundamental and second-­harmonic components, and the discrimination
between inrush and internal fault currents, (ii) LSQ curve fitting-­based algo-
rithms whereby the basic assumptions and algorithm derivation, and the basis of
discrimination between inrush and internal fault currents have been discussed,
(iii) Fourier-­based algorithm, which is used to filter out harmonics contained in
the faulted signal, (iv) Flux-­restrained current differential relay, which is based
on an algorithm that uses the flux-­current relation of the transformer to obtain
the restraint function, and (v) wavelet-­based transformer protection methods,
whereby WT is employed to discriminate between transformer’s internal fault
and inrush currents.
The enhancement of TDP to improve its security and dependability has also
been discussed under section 10.7. This is particularly devoted to discussing secur-
ing TDP against external faults having in mind that modern digital relays are built
in a microprocessor-­based hardware. The basic hardware of microprocessor-­based
transformer protection is discussed under section 10.9.

References

[1] Kennedy L.F., Hayword C.D. ‘Harmonic-­current restrained relays for differ-
ential protection’. AIEE. 1938;57:262–66.
[2] Harder E.L., Marter W.E. ‘Principles and practices of relaying in the unit-
ed states’. Transactions of the American Institute of Electrical Engineers.
1948;67(2):1005–23.
[3] Matthews C.A. ‘An improved transformer differential relay’. AIEE. 1954;73:
645–50.
[4] Rockefeller G. ‘Fault protection with a digital computer’. IEEE Transactions
on Power Apparatus and Systems. 1969;PAS-­88(4):438–64.
[5] Einvall C.-H., Linders J.R. ‘A three-­phase differential relay for transform-
er protection’. IEEE Transactions on Power Apparatus and Systems. 1975;
94(6):1971–80.
[6] Hermanto I., Murty Y.V.V., Rahman M.A. ‘A stand-­alone digital protective
relay for power transformers’. IEEE Trans. 1991;PWRD-­6:85–95.
[7] Schweitzer E.O., Larson R.R., Flechsig A.J. ‘An efficient inrush current
detection for digital computer relay protection of transformers’. IEEE PES
Winter Meeting, Paper A77510-­1; 1977.
Digital differential protection of transformers 283

[8] Larson R., Flechsig A., Schweitzer E. ‘The design and test of a digital re-
lay for transformer protection’. IEEE Transactions on Power Apparatus and
Systems. 1979;PAS-­98(3):795–804.
[9] Degens A.J. ‘Algorithm for a digital transformer differential protection based
on a least-­squares curve-­fitting’. IEE Proceedings C Generation, Transmission
and Distribution. 1981;128(3):155–61.
[10] Degens A.J. ‘Microprocessor-­implemented digital filters for inrush current
detection’. International Journal of Electrical Power & Energy Systems.
1982;4(3):196–205.
[11] Rahman M.A., Jeyasurya B. ‘A state-­of-­the-­art review of transformer pro-
tection algorithms’. IEEE Transactions on Power Delivery. 1988;3(2):
534–44.
[12] Rahman M.A., Dash P.K. ‘Fast algorithm for digital protection of power trans-
formers’. IEE Proceedings C Generation, Transmission and Distribution.
1982;129(2):79–85.
[13] Rahman M.A., Jeyasurya B., Gangopadhyay A. ‘Digital differential protection
of power transformers based on walsh functions’. Trans. CEA Engineering &
Operating Div. 24, Paper 85-­SP-­149; 1985.
[14] Jeyasurya B., Rahman M. ‘Application of walsh functions for microprocessor-­
based transformer protection’. IEEE Transactions on Electromagnetic
Compatibility. 1985;EMC-­27(4):221–25.
[15] Thorp J.S., Phadke A.G. ‘A microprocessor based three phase transformer
differential relay’. IEE Transactions. 1982;PAS-­94:426–32.
[16] Phadke A.G., Thorp J.S. ‘A new computer relay flux-­restrained current
differential relay for power transformer protection’. IEEE Transactions.
1983;PAS-­102:3624–29.
[17] Sevov L., Khan U., Zhang Z. ‘Enhancing power transformer differential
protection to improve security and dependability’. IEEE Transactions on
Industry Applications. 2017;53(3):2642–49.
[18] Sendilkumar S., Mathur B.L., Henry J. Differential protection for power
transformer using wavelet transform and PNN [online]. World Academy
of Science, Engineering and Technology; 2010. Available from http://cit-
eseerx.​ist.psu.edu/viewdoc/download?doi=10.1.1.309.6497&rep=rep1&type
=pdf
[19] Youssef O.A.S. ‘A wavelet-­based technique for discrimination between faults
and magnetizing inrush currents in transformers’. IEEE Transactions on
Power Delivery. 2003;18(1):170–76.
[20] Kim C.H., Aggarwal R. ‘Wavelet transforms in power systems – part 1. gen-
eral introduction to the wavelet transforms’. Power Engineering Journal.
2000;14(2):81–87.
[21] Ozgnenel O., Onbilgin G., Kocaman C. Transformer protection using the wave-
let transform [online]. Ondokuz Mayis University Turkey; 2016. Available from
https://pdfs.semanticscholar.org/055e/ac2c85b2b32eb9bb6551eb5cb0b64d-
9303ab.pdf
284 Digital protection for power systems

[22] Saleh S.A., Rahman M.A. ‘Modeling and protection of a three-­phase power
transformer using wavelet packet transform’. IEEE Transactions on Power
Delivery. 2005;20(2):1273–82.
[23] Degens A.J., Langedijk J.J.M. ‘Integral approach to the protection of power
transformers by means of a microprocessor’. International Journal of Electrical
Power & Energy Systems. 1985;7(1):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:

1. 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.

2. Distance relay overreach:

This problem arises when an external fault is fed from a number of ter-
minals. 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.

3. 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.

4. 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 recognize it as a fault condition.
286 Digital protection for power systems

5. 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 com-
pensation 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 over-


coming these limitations. These schemes can be broadly classified into current-­
based schemes and composite voltage- and current-­based schemes.

11.2 Current-based differential schemes

11.2.1 Basic principles of line current differential protection


The basic principle of line current differential protection used for two-­terminal and
multi-­terminal lines is essentially the same as the percentage-­differential protection
applied to transformers, which was discussed in Chapter 10 [3–5].
Consider the three-­terminal line shown in Figure 11.1. An instantaneous dif-
ferential current signal D(t) and a bias (or restrain) current signal B(t) are typically
formed using the instantaneous currents at the three ends such that:
   
‍ D t = ix t + iy t + iz t ‍ (11.1)

‍ B t = ix  iy  i z ‍ (11.2)
where ix(t), iy(t) and iz(t) are the instantaneous currents measured at ends X, Y and Z,
respectively.

Figure 11.1 Basic three-­terminal line configuration


Digital line differential protection 287

Under healthy conditions, ideally, the magnitude of the differential quantity D(t)
should be zero. In practice, it has a small value both under normal operating con-
ditions and external faults, due to line charging current and other errors such as a
mismatch of current transformers. However, once an internal fault occurs, the dif-
ferential quantity D(t) approximates to the fault current iF(t), i.e., the current flowing
out of the fault point.

11.2.2 Frequency modulation current differential protective scheme


In frequency-­modulation (FM) current differential protective schemes, the
instantaneous values of the currents at each end are frequency modulated and
typically transferred to the other terminal(s) via a voice channel of a microwave
communication network, as shown in Figure 11.2 [3, 4]. Every terminal can be
arranged to perform the functions of transmitting, receiving and initiating trip-
ping signals when a fault is detected, and this is commonly called a master–mas-
ter–master (MMM) system (assuming a three-­terminal line). The other way of
arranging the system is to install a master in one terminal only, which receives
current data from other terminals (slaves) and generates tripping signals that are
sent to the slave terminals. Such a system is called a master–slave–slave (MSS)
system.
The transmission of the instantaneous current values of each terminal is usually
achieved by allocating a frequency band, from 300 to 3400 Hz for each current. The
current at the secondary side of the CT at each terminal is converted into a voltage,
which, in turn, is converted into a frequency signal in the appropriate voice band
before it is transmitted to remote terminals. Either a microwave communication link
or optical fibres can be used to transmit the necessary signals. At the receiving end,
the current signal is recovered by conversion of the frequency-­modulated signal to
a voltage signal.

Figure 11.2   Basic construction of FM relay system


288 Digital protection for power systems

Figure 11.3  ercentage restraining characteristic of an FM current differential


P
relay

11.2.2.1 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.
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.
Digital line differential protection 289

11.2.3 Modal current-based protection scheme


The modal current-based protection scheme is proposed by [5].

11.2.3.1 Operating principles and tripping criteria


The operating principles of this relay also depend on the percentage-­differential cur-
rent concept. However, the differential quantity D(t) and the bias quantity B(t) are
derived using the instantaneous values of modal currents at the line terminals. Let
us first consider these principles using total time variations for a three-­terminal line.
With reference to Figure 11.1, the two quantities D(t) and B(t) at the master end
are given in (11.1) and (11.2), respectively. It will be noted that, under temporary
two-­terminal operation involving transmission between ends Y and Z, or conditions
where there is a complete loss of infeed at end X, the differential signal D(t) and the
bias signal B(t) will be equal in magnitude. This is of no consequence for internal
faults, but with external faults causing an abnormally high degree of CT saturation
at ends Y and Z, it could cause sufficient unbiased differential current to be processed
and thereby lead to relay instability. This problem can be overcome by adding a
logic check within the tripping decision process, as will be explained later.
In its simplest form, the tripping criterion adopted by the relay is that the trip-
ping 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 described by
(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.

11.2.3.2 Modal components of differential and bias quantities


As previously discussed in Chapter 8, assuming an ideally transposed line, phase
quantities can be transformed into what are referred to as modal quantities. This has
the advantage of producing relay measurements that do not require complex trans-
formations, and therefore do not delay signal components in their derivation. The
first modal component of differential and bias quantities D1(t) and B1(t) is formed as
the difference of the ‘a’ and ‘c’ phase currents, such that:
              
D1 t = iax t  icx t + iay t  icy t + iaz t  icz t
               (11.5)
‍ B1 t = iax t  icx t + iay t  icy t  iaz t  icz t ‍

The second modal component-­based signals D2(t) and B2(t) are commonly found
from the difference between ‘a’ and ‘b’ phase currents:
290 Digital protection for power systems
              
D2 t = iax t  ibx t + iay t  iby t + iaz t  ibz t
               (11.6)
‍ B2 t = iax t  ibx t + iay t  iby t  iaz t  ibz t ‍

where iax, ibx, icx are the ‘a’, ‘b’, ‘c’ phase currents at end X, iay, iby, icy are the phase
currents at end Y, and iaz, ibz, icz are the corresponding currents at end Z.
Since in practice there are no fault conditions that can simultaneously give rise
to zero-­valued modal current signals, tripping is initiated for internal faults by at
least one modal channel. This form of signal compression reduces the communi-
cation channel requirement from three (for separate phase-­by-­phase comparison)
to two.

11.2.3.3 Trip-decision logic process


A typical trip-­decision logic process can best be explained by reference to
Figure 11.4, where (a) shows a flow diagram of the decision process algo-
rithm. Figure 11.4b shows typical differential current waveforms for external
and internal fault conditions. The former is small and comprises mainly spill
current, which is predominantly high frequency, whereas the latter is large and
predominantly power frequency. To simplify the explanation, the pickup level
is taken as ±KS instead of the time variant signal ±S(t) (11.4), i.e., KB is assumed
to be zero.
Let us first consider the external fault waveform. It can be seen that the magni-
tudes of the four samples (two to five) are above the pickup levels ±Ks. Thus, 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, KD, 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 dif-
ferential current has exceeded the pickup level, it stays above that level for an
appreciable time. This means the two operations of the logic related to compari-
sons 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 are 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 11.4) further
enhances stability under healthy conditions without significantly affecting per-
formance 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.
Digital line differential protection 291

Figure 11.4  Typical extended digital decision logic process [5]. (a) Decision
logic algorithm; (b) typical waveforms.

11.3 Composite voltage- and current-based scheme

Although current differential schemes are relatively fast in terms of fault-­clearance


times, they have the common problem of becoming insensitive to low levels of fault
292 Digital protection for power systems

current because of their total dependence on current. This arises from the standard
practice of providing a bias signal proportional to through current to ensure security
under conditions where a significant differential signal is generated under healthy
conditions, e.g., due to shunt capacitance charging currents or current transformer
errors. A second problem is that certain system operating conditions may result in
the difference between the differential and restraint signals being small for internal
faults, which means a reduction in dependability of the protection.
To overcome these problems, an approach has recently been adopted that
involves deriving differential signals that are functions of both the voltages and cur-
rents measured at each end of either plain or teed lines [6, 7]. Consequently, a bias
signal to compensate for capacitance spill currents is no longer required. This, in
turn, increases the relay sensitivity to a much lower level of fault currents than pro-
vided by schemes that are totally dependent on current measurement.

11.3.1 Basic operating principles


The basic operating principle relies on deriving signals proportional to the instanta-
neous values of modal voltages and currents at the three ends of the line. Ideally, the
sum of these signals should be zero under healthy conditions but significantly finite
under internal fault conditions. Thus, for a three-­terminal line such as that shown in
Figure 11.1 we have
(
 
k
 
k
 
k = 0 healthy conditions
SX + SY + SZ k = 0, 1, 2 (11.7)
‍ ¤ 0 internal fault conditions ‍
For a two-­terminal line, the signal Sz is equal to zero and the above equation reduces
to
(
 
k
 
k = 0 healthy conditions
SX + SY k = 0, 1, 2 (11.8)
‍ ¤ 0 internal fault conditions ‍
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


11.3.2.1 Two-terminal lines
Assuming a distributed-­parameter, transposed three-­phase transmission line, the
modal voltage and current relationships in the frequency domain can be described
by (8.30):
  
    2     d2 I k !      2  
 
d2 V k ! k
=  ! V ! ,
k
= k ! I k , k = 0, 1, 2
‍ dx2 dx2 ‍
(11.9)
As previously explained in Chapter 8, the above equation shows that wave propa-
gation in a three-­phase line can thus be considered in terms of three independent
Digital line differential protection 293

Figure 11.5  asic two-­terminal line configuration giving current-­reference


B
directions

 
components, each possessing itsown 
modal propagation constant ‍ k (!)‍and asso-
k
ciated modal surge impedance ‍Z0 (!)‍, where k = 0, 1, and 2.
The solutions of the voltage and current differential equations shown in 11.9
take the form:
      
 k x  k x
‍ V k
! = K 1 e + K2 e ‍ (11.10)
and
   h  
k
  i
k
 
k
I k ! = K1 e x  K2 e x /Z0
‍ ‍
where K1 and K2 are arbitrary constants and x is an arbitrary length of line.
In the above equation, the frequency variable ω has been dropped from the
propagation constant γ(k) in order to simplify the notation.
The arbitrary constants K1 and K2 can be determined from a knowledge of the
boundary conditions for a particular system. Consider the two-­terminal line shown
in Figure 11.5, which has the following boundary conditions [6]:
(
0 at end X
x= (11.11)
‍ l at end Y ‍
where l is the length of the line.
By substituting (11.11) into (11.10), we obtain the voltage and current at end X:
    
k
VX ! = K1 + K2
     k   k = 0, 1, 2 (11.12)
k
‍ I X ! = K 1  K 2 /Z0 ! ‍
and those relating to end Y will be
      
k k k
VY ! = K1 e   l + K2 e  l
     k = 0, 1, 2 (11.13)
k k k k
‍ IY ! = K1 e l + K2 e l /Z0 ‍
By substituting the values of K1 and K2 obtained from (11.12) into (11.13), we finally
obtain:
           h          i
k k k k   k
VY ! + Z0 ! IY ! = e l Vxk ! + Z0 ! Ixk ! k = 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, (11.14) becomes:
294 Digital protection for power systems
h  
     k          i
k k k k
VY !  Z0 ! IY ! = el VX ! + Z0 ! IY ! k = 0, 1, 2
‍ ‍ (11.15)

or
   
k k
‍ SX + SY = 0 k = 0, 1, 2‍ (11.16)
 
k
where SX(k) is equal to the right-­hand side of (11.15) and ‍SY is
‍ equal to the left-­hand
side negated.

11.3.2.2 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 (11.14)
can be derived in terms of the current and voltage at the line ends. When the result-
ing equations are combined together (and applying the fact that the sum of currents
at the tee point T is zero), we obtain:
     
k k k
‍ SX + SY + SZ = 0 k = 0, 1, 2‍ (11.17)
where
          
k k k k
SX = AX VX ! + BX Z0 ! IX !
 
k     
k   k   k  
SY = AY  BY VY ! + AY + BY Z0 ! IY !
 
k   k   
   k   k  
‍ SZ = AZ  BZ VZ ! + AZ + BZ Z0 ! IZ ! ‍
   
k l k l
AX = e X  1, BX = e2

X +1 
     
  k lX +lY   k lX lY
AY = e    
, BY = e    
   
  k lX +lZ   k lX lZ
‍ AZ = e , BZ = e ‍
lX, lY, lZ are the lengths of sections XT, YT and ZT, respectively.

11.3.2.3 Trip-decision logic
It is seen from the foregoing that, in theory, the signals at the ends of a line (whether
of the two- or three-­terminals type) should sum to zero under all healthy conditions.
However, because of quantisation, transducer errors, etc., it becomes necessary to
apply a small threshold Th to the sum signal of (11.17). Thus in practice the simplest
relay tripping decision would be
     
k k k
‍ |SX + SY + SZ | > Th k = 0, 1, 2‍ (11.18)
The above equation
  applies to a three-­terminal application. In the case of a two-­
k
terminal line ‍SZ ‍becomes zero.
Digital line differential protection 295

It will be noted that the tripping decision described by (11.18) is fairly simple
compared with that used in the previously described current differential schemes. In
consequence, a relatively simple trip-­decision logic process can be used to provide
the necessary degree of dependability and sensitivity. The simplest criterion that
provides an adequate performance for most protective applications involves the ini-
tiation of a trip signal if two consecutive samples of the signal summation of (11.18)
exceed the fixed threshold setting Th.

11.4 Application of wavelet to protection of tapped


transmission line

In this section, a wavelet-­based digital protection scheme for a tapped transmis-


sion line is discussed [8–11]. The wavelet packet transform (WPT) is used in the
development of this protection scheme. As discussed in Chapter 2, the WPT pro-
vides more detailed and accurate signal decomposition than the discrete wavelet
transform, which helps in achieving proper discriminating between the forward and
reverse faults to which extra-­high voltage tapped lines are subjected. The main idea
of this scheme is based on deriving an add (ADD) signal using WPT coefficients that
are proportional to the fault current at each terminal.

11.4.1 Selection of mother wavelet and scale


The key factors in analysing a transient signal using the wavelet transform are (1)
selecting the mother wavelet and (2) deciding the number of multiple decomposi-
tion steps.
Regarding the selection of mother wavelet, different criteria are used to deter-
mine the optimal mother wavelet. Among these criteria considered are the maxi-
mum description length and summation of detailed coefficients over a one cycle. In
the development of this scheme, the absolute values of the decomposition coefficient
are summated over one-­cycle window for each fault position in the process of select-
ing the optimal mother wavelet. It has been found that for tapped transmission line
applications, mother wavelets Daubechies 4 (db4) and Symlets 5 are the best con-
sidering different types of faults with a fault resistance up to 200 Ω. Additionally,
it has been found that using db4 mother wavelet helps in ensuring the speed of the
protection scheme, as well as improving the ability of harmonics elimination. It has
also been found that db4 mother wavelet matches to high degree the decomposed
signals and, consequently, it is selected for this protection scheme.
Regarding the number of multiple decomposition steps, it will be noted that
the optimal number of levels of resolution indicates the minimum number of scale
changings during the multiresolution analysis without violating the synthesis of the
decomposed signal. The number of levels provides information regarding the signal
frequency bands. The condition for the minimum number of scaling has an effect on
the computations required to decompose the signal and can also provide the ability
to realize the multiresolution analysis. The number of frequency sub-­bands increases
with the number of levels of resolution, which, in turn, provides better and accurate
296 Digital protection for power systems

representation of frequency components. The downside is the volume of calculation


increases as the number of levels of resolution increases, which affects the speed of
protection. The number of decomposition steps is usually decided by comparing the
scale of sampling frequency with that of the frequency component of the desired
signal. Based on the above discussion, the third level of resolution is selected as the
optimal for the digital protection algorithm under consideration.

11.4.2 Basic description of the protection scheme


The WPT-­based protection scheme of tapped transmission line is designed based on
deriving an ADD signal at each bus that is proportional to the fault.
Let us apply the WPT-­based protection scheme on the three-­terminal (or tapped)
line shown in Figure 11.1. On the occurrence of a fault on this line, the measured
current signals ia, ib, ic at buses X, Y, and Z are decomposed up to the third level using
the db4 mother wavelet. The decomposition process results in the WPT decomposi-
tion tree (see Chapter 2 for the detailed discussion).
This followed by calculating the SUM IBus parameter at buses X, Y, and Z using
(11.19), whereby the different third-­level wavelet coefficients (W30–W37) for each
current signal at each bus are added to each other. It will be noted that the reason for
using wavelet coefficients up to W37 is because they provide sufficient information
regarding disturbances:
7 !X,Y,Z
 X,Y,Z Pn P 
SUM IBus j a,b,c = W3m j (11.19)
j=1 m=0
‍ a,b,c ‍

where a, b, and c refer to the three phases. X, Y, and Z are the three buses of the line.
j represents the most recent sample. n is the number of samples in the window and
m is element number of the third level.
The fault is considered as an internal fault, if the SUM IBus exceeds a certain
value of local threshold (Lthr), otherwise, it is an external fault. Confirmation of
this condition important to prevent mal-­operation due to wide variations of system
configuration and operating conditions. The value of Lthr is set to 150 to prevent the
mal-­operation during normal switching operation.
The performance of the protection algorithm is further reenforced by the adding
the values of all SUM IBus of each bus of the corresponding phase as follows:

‍ ADDa,b,c = SUM IBusXa,b,c + SUM IBusYa,b,c + SUM IBusZa,b,c ‍ (11.20)

A communication facility is needed (as shown in Figure 11.2) so that the three buses
can communicate with each other. The protection relay establishes its final decision
whether to trip or not based on the received SUM IBus signal from the remote ends.
The relay is set to initiate a trip command only when the ADD signal exceeds a posi-
tive threshold (Pthr) value. Figure 11.6 shows the flow diagram of the WPT-­based
protection scheme for tapped transmission line.
Digital line differential protection 297

Figure 11.6  low diagram of the wavelet packet transform-­based protection


F
scheme for tapped transmission line (after [8])
298 Digital protection for power systems

11.5 Summary

This chapter is devoted to discuss the digital line differential protection, which
include two-­terminal lines as well as multi-­terminal and tapped lines.
Current-­based differential schemes are discussed in section 11.2, which include
the basic principles of line current differential protection for two-­terminal and multi-­
terminal lines, FM current differential protective scheme, and modal current-­based
protection scheme.
To overcome the problems associated with differential schemes, the composite
voltage- and current-­based scheme is introduced in section 11.3. This included the
basic operating principles and the formation of terminal signals.
Application of WPT to protection of tapped transmission line is covered in sec-
tion 11.4. The selection of mother wavelet and scale that suits this application is
highlighted. This is followed by covering the basic description of the developed
protection scheme.

References

[1] AIEE Working Group of the Line Relay Protection Sub-­Committee.


‘Protection of multiterminal and tapped lines’. AIEE Fall Meeting, Paper
CP60-­1274; 1960.
[2] IEEE study Committee report on protection aspects of multi-­terminal lines
[IEEE Report 79, TH0056-­2-­PWR]. 1979.
[3] Kitagawa M., Andow F., Yamaura M. ‘Newly developed FM current-­
differential carrier relaying system and its field experience’. IEEE PES Winter
Meeting; Paper F78; 1978. pp. 291–7.
[4] Yamaura M., Masui M., Okita Y. ‘FM current differential carrier relaying,
developments in power system protection’. IEE Conference Publication.
1980;185:156–60.
[5] Aggarwal R.K., Johns A.T. ‘The development of a new high speed three-­
terminal line protection scheme’. IEEE/PES Summer Meeting; Paper 85 SM;
1985. pp. 3200–7.
[6] Aggarwal R.K., Johns A.T. ‘A differential line protection scheme for power
systems based on composite voltage and current measurements’. IEEE/PES
Winter Meeting; Paper 89WM 053-­0 PWRD; 1989.
[7] Aggarwal R.K., Johns A.T. ‘New approach to teed feeder protection using
composite current and voltage signal comparison’. Developments in Power
System Protection, IEE Conference Publication. 1989;302:125–29.
[8] Bhalja B., Maheshwari R.P. Wavelet-­based Relaying Scheme for Tapped
Transmission Line [online]. 2009. Available from https://www.tandfonline.​
com/doi/pdf/10.1080/15325000902817267?needAccess=true [Accessed 29
Jan 2019].
[9] Hamid E.Y., Kawasaki Z.-­I. ‘Wavelet-­based data compression of power
system disturbances using the minimum description length criterion’. IEEE
Transactions on Power Delivery. 2002;17(2):460–6.
Digital line differential protection 299

[10] Saleh S.A., Rahman M.A. ‘Testing of a wavelet-­packet-­transform-­based dif-


ferential protection for resistance-­grounded three-­phase transformers’. IEEE
Transactions on Industry Applications. 2010;46(3):1109–17.
[11] Jung H., Park Y., Han M., Lee C., Park H., Shin M. ‘Novel technique for fault
location estimation on parallel transmission lines using wavelet’. International
Journal of Electrical Power & Energy Systems. 2007;29(1):76–82.
This page intentionally left blank

You might also like