New method of power swing blocking for digital distance protection

P.J.Moore A.T.Johns

Indexing terms: Power swings, Distance relay, Reactance transients, Fault detection

Abstract: A new method for preventing digital distance relay operation during power swings is presented based on detection of transients in the measured reactance. The transients are detected by the difference between two reactances calculated using two parallel FIR filters based on coefficients used in the discrete Hartley transform. The transients are present under genuine fault conditions but not during power swings. Fault transient detection therefore enables the relay for faults but leaves it inherently blocked for power swings. Simulation results are presented showing that the fault transient detection method does not lead to increased relay operating time, but allows the relay to differentiate between faults and power swings. The significant advantage of this new method of blocking is that the relay can correctly operate if a genuine fault occurs during the power swing period.

rent magnitude a = angular delay between solutions of fundamental line equation Hh(k) = coefficients of Hartley filter (for k = 0, ..., N-1) K , ( k ) = time reversed coefficients of Hartley filter (for k = 0, ..., N-1)
1 Introduction

Glossary of symbols a, b, c = subscripts denoting phase quantities

0, 1 = subscripts denoting zero and positive phase

sequence quantities R, L = resistance and inductance measured by relay X , X , = reactance and directional reactance measured by relay vr[n],Qn] = discrete time relay voltage and current signals at time index nT i[n]= discrete time current signal time derivative T = sampling interval f , = sampling frequency,f, = l / T p = number of discrete time samples taken per power system frequency cycle Jb = nominal power system frequency K = residual compensation factor, K = 1/3(Z10/Z~,-1) Zl = transmission line impedance D = derived signal being proportional to square of cur@ IEE, 1996 IEE Proceedings online no. 19960055 Paper first received 9th February 1995 and in final revised form 29th August 1995 The authors are with the School of Electronic and Electrical Engineering, University of Bath, Bath BA2 7AY, UK
IEE ProccGenev. Tvansm. Distrib., Vol. 143, No. I , January 1996

Under certain conditions of power system instability unwanted operation of distance relays can occur due to the apparent impedance sensed by the relay falling within the characteristic. This is due to an abnormal transmission line voltage profile. For example, if the voltages at either end of a homogenous transmission line are in anti-phase, then there exists a position of voltage zero at the middle of the line which will be interpreted as a fault by distance protection at either end. Although such conditions are undesirable on a power system, distance relays cannot provide the corrective action needed to restore stability since the phenomenon is not associated with any defect in the protected line. Consequently distance relays are usually blocked from operating during power swings. This is especially important where the instability is recoverable provided critical transmission lines remain in service. The conventional method of blocking relay operation during power swings is to time the movement of the impedance through the zones of the relay. Under genuine, in-zone, fault conditions the impedance measured by the relay will move from its prefault position into the characteristic almost instantaneously. Thus, zones 1, 2, 3 and an outer, power swing detection, zone will pick up simultaneously. However, under power swing conditions the impedance trajectory, being governed by the inertial constants of the generating plant attached to the line, will move slowly, typically taking hundreds of milliseconds to move into the characteristic. In this situation, the outer power swing detection zone will pick up before any other zone. This time delay is used as the basis of the conventional power swing blocking method. Two problems can be experienced with this method. First, the relay may not respond to genuine faults occuring during the power swing period since it is blocked from operation. Secondly, the time delay has to be set with a knowledge of the likely speed of movement of the impedance during the power swing. It is therefore possible that the relay may not correctly block if fast changes in apparent impedance occur. Recently, new techniques for power swing blocking have been reported and include measuring the rate of
19

change of line current [l], measuring the rate of change of apparent resistance [2] and calculating the angle between busbar voltages [3]. The new method to be described here differs from all these techniques in that it detects the condition of a fault, rather than the condition of a power swing. This provides the benefit that no assumptions regarding the nature of the power system are needed. This work is a further development of earlier research in the area of power swing blocking using fault transient detection [4] and is applicable to digital distance relays.

equations for calculating R and

X are given below:

RZ

v[n]i[n - 61- v [ n- 6]i[n] i[n]Z[n - 61 - i[n- 6]Z[n]

%

x = 2xfoL

v [ n- 6]i[n] - ~ [ n ] i[n 61 i[n]Z[n - 61 - i[n - 6]i[n]

where f o is the system frequency and the time derivative is approximated by the central difference formula,

di[n] i[n+ 1 1 - i[n- 1 1 z[n] = -% dt 2T where f, is the sampling frequency.

,

T = -1
fs

(3)

filtering
I

mixing

I
-

+
...__--

--.--.-.1,2x
. -

.
y

9 ......
-1
._

1.
- ,

.- .......... ..... ......

...

impedance

fault

element trip
I

relay trip
. . . .

. . . . . .

...............................

enable flag

Fig. 1 Structure of digital distance relay

Digital distance relay structure

2.1

Overview
RO -9 R, Fig. 2 Relay quadrilateral characteristic and counting strategy

A block diagram of the structure of the digital distance relay used for this work is shown in Fig. 1. The discrete time input signals to the relay, i.e. the three phase voltages and three line currents, are assumed to be sampled at a frequency of 4kHz using a 16 bit analogue to digital converter. Sampling is assumed to not cause aliasing and to occur on all inputs at the same instant of time, i.e. each input channel has a separate sample and hold amplifier. The main filtering function of the relay is performed digitally, for each input signal, using a finite impulse response (FIR) filter based on a sine function and having 20 symmetrical coefficients. The filtered signals are then mixed in order to provide the inputs for three phase elements and three earth elements of zone 1. Each zone 1 element comprises of an impedance calculation stage, where the input discrete time voltage and current signals are converted to discrete time resistance and reactance signals, and a fault evaluation stage, where the impedances are compared against a characteristic in order to set the element trip flag. Any relay element trip will cause a relay trip if the ENABLE flag is set which, in turn, is a function of the power swing blocking feature of the relay to be described in the next Section. For simplicity, this work has considered only the zone 1 elements since the power swing blocking feature is not related to the operation of other zones. Zones 2 and 3 may easily be implemented by providing extra fault evaluation stages reflecting the larger characteristics of these elements, no extra impedance calculation is required.

2.3 Relay characteristic and counting strategy The quadrilateral characteristic of Fig. 2 was used in
all of the relay elements. To fulfil the requirements of transmission protection where fast relay operation is required, a target relay operating time of less than lOms for the majority of faults was chosen. Although the relay digitally filters its inputs, this filtering is not perfect and results in some noise appearing on the impedance estimates. Consequently the single occurence of the impedance falling within the characteristic is not a good indication of a fault. To provide the extra filtering needed to ensure that the relay gives good reach point behaviour, a counting strategy is used as shown in Fig. 2. Each relay element is provided with a counter which is initially set to zero. The counter can increase provided the measured impedance falls within the characteristic and that the directional check described in the next Section is positive. For faults leading to impedance estimates falling within the first 80% of the characteristic, the counter increases at the rate of nine for each sample in this area. When the counter reaches 45, i.e. five coiisecutive samples in this area, the element trips. If the fault is closer to the relay boundary, the counting value is reduced to 4 for measured impedances between 80 and 90% of the characteristic. The counting value is set to 1 for impedances between 90 and 100% of the characteristic. The counting values are reflected negatively for impedances falling beyond the reach point. The counting strategy has been carefully designed in order to provide the extra filtering required for exacting relay operation under all conditions. The counter cannot exceed 90 or go less than zero. After a relay trip occurs, the relay resets 20ms after the element counters return to zero.
IEE Proc.-Gener. Transm. Distrib., Vol. 143, No. 1, January 1996

2.2 Impedance calculation The calculation of measured resistance, R, and reactance, X , in the fault loop is performed using a wellestablished technique [5] which solves the fundamental first-order line equation, i.e. v = Ri + L dildt, for R and L (fault loop inductance) by using two sets of discrete time relaying voltage v(n) and current i(n) separated in time by six samples, which is equivalent to 1.5ms. The
20

2.4 Directionality
In order to directionalise the relay for busbar faults where the voltage collapses to zero, a separate impedance is calculated using the post fault current and the prefault voltage which is stored in memory. This reac, , tance is referred to as the directional reactance, X and is given as ~ [-n p - 6]i[n] - ~ [-n p ] i [ n- 6 1 M 2Tf

x ,

ficient residual current is detected. Note that residual current can be present under normal conditions due to effects such as assymetric loading and nonideally transposed lines, thus the threshold level needs to consider these effects. The presence of phase faults and three phase faults will be detected by the phase elements. The a-b element is provided with the following inputs:
U[n] = U a [ n ] - Ub[n] Z [ n ]= Za[n] - Zb[n] (8) and similarly for the b-c and c-a elements. Biasing the phase elements against the presence of negative sequence current is not carried out since it would prevent relay operation for three phase faults.

Z[n]i[n -6 1 '

i [ n - 6]i[n]'

p = SJ = 8 0

(4)

fo

X , will be positive for forward faults and negative for

reverse faults and is used to directionalise the relay for close-up faults. Hence, in Fig. 2, a directional threshold is established which Xm must exceed in order for the element counter to increase. In eqn. 4, a one-cycle delay for the voltage is achieved by taking values stored in memory 80 samples previously since, with a sampling frequency of 4kHz, one cycle of a 50Hz signal is made in exactly 80 samples. Under dynamic conditions where the power system frequency varies, 80 samples will not correspond exactly to one cycle, how, is used to form only an approximate ever, since X measure of direction, this will not lead to relay maloperation. In fact, errors of up to 25" in the memory voltage do not affect directionality. Delays of greater than one cycle can be used by taking an integer multiple of p in eqn. 4.

position B position A

I I

fault inception

v
A

> \
. ..

/position

.............. ........ .

. t

Fig. 3 Effect of fault on relaying voltage

2.5 Relay elements

The six zone-1 relay elements each have separate impedance calculation and fault evaluation stages. The earth elements need residually compensated line currents and so the signal mixing stage of Fig. l performs the following calculation:

Power swing blocking using fault disturbance detection

3. I The effect of disturbances on measured impedance

The occurrence of a fault in the relay protected zone will result in instantaneous changes in the magnitude and phase of the relay input signals. This is illustrated in Fig. 3 where the relaying voltage is subject to a reduction in amplitude and displacement in phase. Since the relaying algorithm processes a fixed number of input samples at any point in time, the concept of a time window, equal to the length of the relaying algorithm, can be applied to Fig. 3 . The window length, T,s, corresponds to the sum of the group delays of filtering, impedance calculation and fault evaluation used in the relay. The window moves towards the right at every sampling instant and so the right hand edge of the window passes through the most recent voltage sample. In position A, the window is seen to contain only prefault information. More than T,s after the fault, in position C , the window is seen to contain only postfault information. However, in position B the window is seen to traverse the fault point and hence the relay processes both prefault and postfault information. This leads to errors in the impedance measurement process since the waveform present in window B does not represent a single frequency sinusoid, upon which the relay processing is based. This effect is more conveniently described by examining reactances measured by the relay for a typical fault. Fig. 4 shows reactances measured for three faults, in each case the prefault and postfault conditions are identical except that the point on wave of the voltage at the fault inception is different. It is seen that prior to the fault, i.e. equivalent to position A of the window, all three reactances are the same, similarly after the
21

ires[n] =i a [ . ]

+ ib[n] + ic[n] +

(5)

where is,, is the residual current, and

i a [ n ] = i a [ n ] lKlires[n] (6) and similarly for ib, and i,, where K is the residual compensation factor defined by K = 1/3 (Zlo/Zl~-l). Since the modulus of K is taken, residual compensation is achieved using a scalar multiplication in eqn. 6. This is suitable for overhead line applications where the argument of K rarely exceeds 5". For underground cable applications this would not be acceptable and a complex residual compensation technique must be used [6]. Hence, each earth element is provided with the respective phase voltage and residually compensated line current inputs. Since earth faults are always accompanied by zero sequence components in the relay input voltage and current signals, earth elements can be prevented from operation for clear-of-earth faults by testing for the presence of residual current. The magnitude of residual current can be found using the following equation:

D = -2~folI,,,I~ sincu
= i r e s [ n ] i r e s [ n - 61' - i r e s [ n - 6]ires[n]'

(7)

a = 2Tf0,

Js

Proof of this equation is shown in the Appendix. Thus D is compared against a suitable threshold with the effect that earth element operation is prevented if insufIEE Proc-Genes. Transm. Distrih., Vol. 143, No. 1, January 1996

fault, i.e. position C, again the reactances are identical. However, during the fault, i.e. position B, the reactance behaviour is erratic and dependent on the fault inception angle. It is also discovered that changing the FIR filter coefficients used in the relay also affects the behaviour of reactance during the fault. A result similar to Fig. 4 can also be produced by using just one fault, but processing it through the relay three times each using a different FIR filter. This effect - thefuult transient effect - which is also present on the measured resistance, has hitherto been regarded as a nuisance. For example, without careful design of the counting strategy it is possible for the relay to overreach for certain out of zone fault conditions due to the fault transient effect causing the measured impedance to enter the characteristic long enough for the relay to trip. The basis of the work described in this paper is recognition of the fact that the fault transient effect can be used to provide the relay with important information since the fault transient effect is present, to a greater or lesser extent, at the inception of every fault for which the relay must operate. Furthermore, the fault transient effect will not be present under power swing conditions which are caused by changes in generator rotor positions resulting in gradual, rather than instantaneous, changes in relaying voltage and current signals. Hence the potential exists for preventing relay operation under power swings, not, as is the convention, by blocking operation when swings occur, but by enabling operation when faults occur, thus leaving the relay inherently blocked to all power swing effects.

lated reactance. The two reactance outputs can then be subtracted to yield an output which is nonzero only during a fault disturbance. The operation of this technique relies upon the use of two filters which have identical magnitude responses, giving immunity to changes in system frequency or harmonics, but differing phase responses, thus promoting individual behaviour to fault discontinuities. These constraints rule out the use of conventional Fourier type filters based on sine and cosine waveshapes since the magnitude responses are similar, but not identical, which leads to problems with noise susceptibility and system frequency changes. Instead, filters based on coefficients used in the discrete Hartley transform are used.

3.3

The discrete Hartley transform

The discrete Hartley transform was proposed by Bracewell [7] based upon an integral transform presented by Hartley [8]. The Hartley transform can be used to evaluate the Fourier transform and bears the advantage that only real computation is required, as opposed to the Fourier transform which uses complex computation. The N coefficients used in the fundamental component of the discrete Hartley transform, H&), are evaluated from:

for k = 0, 1,. . . , N - 1 (9) and can be considered as coefficients of a FIR filter in the same way that Fourier coefficients constitute 'Fourier filters'. Hartley filter coefficients have the property, not found in Fourier coefficients, that a filter implemented using coefficents Hh(k), and a filter implemented using time reversed coefficients, K h ( k ) ,i.e. H - * ( k ) = H*(N - IC - 1) (10) have identical magnitude responses but differing phase responses. This effect is a consequence of the fact that the direct and inverse forms of the Hartley transform are identical and has recently lead to the use of the Hartley transform in systolic array parallel processing The frequency magnitude response of the Hartley filters is similar to the general form of the Fourier filters frequency magnitude response and is of the form sin(x)lx. For N = 8, the magnitude and argument of the frequency response of the Hartley filters is shown in Table 1 for several spot frequencies.
Table 1: Frequency responses for Hartley filters sampling frequency = 4000Hz)
Frequency
Hh(fl 50Hz

PI.

time, ms

Fig. 4 Effect

o f fault inceptton angIe on measured reactance

3.2 Detection of fault disturbances In order to detect the disturbances described earlier it is necessary to detect only the transient region of the reactance measurement caused by the fault. This was initially attempted by a form of numerical differentiation of the measured reactance [4] which, with reference to Fig. 4, aimed to provide an output during the tranperiod, yet entirely remove the constant reactance present during the prefault and postfault periods. Despite some encouraging results using this approach, its high sensitivity t o noise can cause spurious outputs outside of the fault instant. Superimposed techniques were also considered yet these would require the relay to use synchronous sampling to afford correct fault disturbance detection when changes occur in the power system frequency. The faregoing difficulties were overcome by using an approach based on the calculation of two reactances using identical input signals filtered using two different FIR filters. As discussed earlier, dissimilar FIR filters will react differently to the fault transient period thus leading to differences in the calcuLL

(N= 8,
~

100Hz 1.481L63" 1.4811-126"

1200 Hz 0.662154' 0.662118"

0.754176.5" 0.7541108"

&(fl

3.4

Implementation of fault disturbance

detection process
The detection of fault disturbances using the fault transient effect was implemented using the above techniques. Fig. 5 shows the processing required. Input samples are filtered, in parallel, using the Hartley filters with coefficients described in eqns. 9 and 10. Two reactances are calculated from the filtered signals and subtracted to form a difference, X@@ which is zero except under fault disturbance conditions. Fig. 6 shows the ~ a single phase-to-earth fault at the behaviour of X d to
IEE Pioc -Gener Transm Distiib
Vol 143 No 1 January 1996

relay boundary. To facilitate fault detection using Xdirft this signal is, in turn, squared and then averaged using an eight point moving average filter to yield the signal X2d,v, shown in Fig. 7. It is clear that X2d,v very quickly responds to a fault disturbance. Finally X2dav is compared to a threshold.
enable

defined using the above mixing equations. A threshold of 10Q2 was used for detecting Pdav as shown in Fig. 7.
40

I/

Fig. 5 Fault detection process

time, ms

Fig. 8 Reactance of a to earth element

100 -50 -m

fau It occurrence

E o

prefault ,
058
59 60 61

,,-,-,,
1 4
62 63
I

posttault
l

64

65

55

ul

80

-150

Fig. 6 Re,Yponse of XdirJto an earth fault

:q

-I
48

: : I :

fi

60 54 time, ms Fig. 9 ReJistance of a to earth element 52

56

68

72

76

2500

5 0 0 -. threshold 63 64 time, ms Fig. 7 Resp0n.w of X2dav to un earth fault 58

59

60

61

62

65

66

67

-1 0
60 64 68 time, ms Fig. 10 Directionul reactance o j u to earth element
48 52 56
100

72

76

To ensure fast fault transient effect detection, the number of coefficients used in each Hartley filter, N , was chosen to be 8, which is less than the 20 coefficients used in the relay digital filter. The process shown in Fig. 5 is able to detect fault disturbances arising from all types of fault for which the relay must operate. For this reason, the discrete time input signals are prov ,e [ ~ ] , duced by mixing the relay input signals v,[n], v ~ [ u ] & [ U ] , 4 7 2 1 and ie[n], with the effect that only one fault disturbance detection process is needed for all fault types. Ideally the signals would be mixed to produce a positive phase sequence quantities, however, although possible [ l o ] , this would be computational intensive, and so an alternate, simpler form of mixing was sought. Initially a modal form of mixing, using elements of Karrenbauers modal transformation matrix was tried, however, this was not successful for all faults. The mixing was consequently adjusted to

::
20 0 48
: :
:
i
;

80

52

56

60 64 time, ms

:68
:

I
I-l

76

72

t
t

trip
I

lenobie
Fig. 11 Counter and trip flag for
U

to earth element, and ENABLEflug

v,i,[n] and

= v,[n] - 2ub[n]- 3v,[n]

(11)

imza[n] = i,[n] - 2ib[n] - 3ic[n] (12) Extensive testing of the arrangement for a wide variety of faults, short circuit levels, loading and fault resistances up to 15052 showed that X2davwas always well
IEE Proc.-Gener. Transm. Distrib., Vol. 143, No. 1, January 1996

The threshold prevents unfaulted steady-state values of X2d,y triggering the fault disturbance detection process and could be significantly increased if necessary. In particular, the threshold can be increased to afford greater noise immunity, although the fault detection technique was observed to perform correctly (using the 1052* setting) under likely noise conditions experienced by the relay, e.g. travelling wave noise and harmonics.
23

However, in applications where high levels of noise are anticipated, it is likely that the number of coefficients in the relay's main FIR filter will need to be increased to ensure correct relay operation. It would be prudent, in this situation, to increase the Hartley filter coefficients by a similar ratio. When X2duyexceeds the threshold, an ENABLE flag is set for a period of 20ms. This period is designed to allow correct operation of the fault evaluation process of the relay elements described in Section 2. The 20ms period is reset every time any element counter exceeds zero. The ENABLE flag is then used to allow the relay element trips to cause a relay trip as shown in Fig. 1.

Results

relay voltage and current. Furthermore the ENABLE flag responds to the pole slips, however, due to the assymptotic behaviour of X and R at these positions, there is no risk of relay operation. At times 0.4s, 0.78s, 1.0s and 1.18s, the measured R and X values are in the relay characteristic, this is illustrated for one case in Fig. 14 which shows the passage of impedance through the relay quadrilateral characteristic. The a-b element trip flag is not set during the power swing due to the , , shown in behaviour of the directional reactance, X Fig. 15. It is seen that Xm largely remains negative during the period when R and X are in zone and thus the a-b element counter is inhibited from increasing. This variation in Xm is caused by a combination of the change in system frequency and the change in apparent reactance due to the power swing. Changing the delay , is period over which the voltage used to calculate X evaluated has no effect as shown in Fig. 15 for delays of 1, 2 and 4 cycles. Thus, even without the fault detection enabling of the relay trip flag, the relay remains stable during this particular power swing.

-600

n n
0

enable

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 time, s

260km transmission line

Fig. 13 Relay signals during an instability resulting from 150ms fault at busbar R (relay at S )

three phase fault Fig. 12 Model system used for power swing study

60

8oi .l-!---7
1=0 80s

The performance of the relay under power swing conditions was examined for the model system shown in Fig. 12 which depicts a generating station with two 600MVA sets connected, via transformers, to a 260km, 400kV overhead line which is, in turn, connected to a large power system. The simulation allows a three phase fault to be applied to busbar R which consequently prevents power being exported from the station to the rest of the system. Parameters used in the simulation are given in Appendix 8.2. Fig. 13 shows the measured R and X of the a-b element when only one generating set is connected to busbar S. Initially, the generator exports power at rated load. At time 0.05seconds a fault is applied at busbar R which is cleared after 150ms. During this period R and X measure the total line impedance, although no operation occurs since the relay is set to 80% of this value. Note that the measured R and Xvalues suddenly change at the point of clearance. After the fault is cleared the system becomes unstable resulting in power swing and pole slips. Pole slips result in near assymptotic behaviour in R and X . The ENABLE flag is set by both the fault inception and the fault clearance, since these events result in instantaneous changes in
24

2 00
150
100

50
. c

- 50
-1 00

-1 50 {
-200

1
720
760

a40 WO 920 960 time, ms Fig. 15 Directional reactance using various voltage delays for an instability resulting from I5Oms fault busbar R (relay at S)
IEE Proc -Gener. Transm. D i s t r i b . , Vol. 143, No. 1, Junuary 1996

aoo

If the relay position is now changed to the lv side of the generator transformer, Fig. 16, then R, X and X , are consequently higher and the a-b element trip flag is set when R and X enter the characteristic. However, since a relay trip requires both the element trip flag and the ENABLE flag to be set simultaneously, it is apparent from Fig. 17, that no relay trip will occur since no fault disturbance is detected during the power swing period.
6 0 0 ~ I

disturbance detection process does not increased relay operating times.

lead to

-600
0

n enable fl

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 time, s

-600.-

I n

Fi . 18 Relay signals during an instability resulting from a 2 3 0 m fault at %usbarR with two muchines connected at S (relay at S)

n enable

Fig. 16 Relay signals during an instability resultingfrom a 150mfault at bus bar R (relay on lv side o f generator transformer, meusured impedance referred to hv side of generator transformer)

60

c:
c
?!

-20

$ 40

20 10

10

20

30 40

g 20
0 -20

resistance, Fig. 19 L o w of im edance for t = 0.7 to 0.8s during an instability resulting from 2 3 0 m s ~ u l at t busbar R with two machines connected at S (relay at S )

50

60

70

80

20 30 40 50 60 70 80 90 resistance, L I . Fig. 17 Locus of impedunce for t = 0.7 to 0.8s during an instability resulting from 150m fault at busbar R (relay on lv side of generator transformer, measured impedance referred to hv side of generator transformer)
20 10

10

Figs. 18 and 19 show the effect of a 200ms fault at busbar R with two machines connected and the relay positioned at R. Due to the higher inertia of the two machines, the power swing and pole slip period is far slower. This results in a period of over 0.2s when the a-b element trip flag is set. However, again due to the behaviour of the ENABLE flag, no relay trip will occur.
5
Conclusions

A new approach to blocking digital distance relay operation during power swings has been proposed based on detection of fault disturbances which are present during genuine faults but not during power swings. The relay detects fault disturbances by the use of FIR filters based on the Hartley transform and calculated reactance values. As a result of the disturbance detection, the relay trip flag is enabled during the period when relay operation is likely. This leaves the relay inherently blocked for all events not associated with fault disturbances such as power swings. The extra processing needed for the fault disturbance detection is relatively modest and is less than the processing needed for a single relay element. Results have shown that the fault
IEE Proc.-Gener. Transm. Distrib., Vol. 143, No. I , January 1996

On simulated dynamic system tests, the relay is observed to correctly block under power swing conditions which cause the measured impedance to pass through the characteristic. For some power swing conditions the relay does not require power swing blocking since the directional reactance, calculated by the relay to ensure correct operation during close-up faults, affords a degree of immunity. However, in the general case, power swing blocking is needed and the new approach proved to be reliable. The relay performed correctly even under fast pole-slipping conditions where the generator was slipping every 200ms. A significant advantage of this approach compared to conventional power swing blocking techniques is that the relay will correctly operate if a genuine fault occurs within the protected zone during the power swing period. Furthermore, the only setting associated with the fault disturbance detection process is a relatively insensitive threshold used to determine the presence of fault transient effect reactance. Once this threshold has been established, the power swing blocking feature of the relay is likely to be application independent and not require user settings.
6
Acknowledgments

The authors would like to thank the University of Bath, Power and Energy Systems Group, for the provision of facilities to conduct this research. The authors would also like to extend their gratitude to Thomas Hass, University of Aachen, Germany, who helped with the simulation results while spending an ERASMUS funded study visit to Bath in 1994.
25

eferences

1 ZHANG, Z.Z., and CHEN, D.: An adaptive approach in digital distance protection, IEEE Trans., 1991, PWRD-6, (l), pp. 135142 2 GAO, Z.D., and WANG, G.B.: A new power swing block in distance protection based on a microcomputer principle and performance analysis. IEE Conf. Publ. 348, 1991, pp. 843-847 3 MECHRAOUI, A., and THOMAS, D.W.P.: New blocking principle to overcome the deficiencies of the conventional power swing blocking schemes for distance protection. Proceedings of 29th Universities Power Engineering Conference, September 1994, (Galway), pp. 216-219 4 MOORE, P.J., and JOHNS, A.T.: A new method of power swing blocking for digital distance protection. Proceedings of International Power Engineering Conference, March 1993, (Singapore), pp. 633-637 5 MOORE, P.J., and JOHNS, A.T.: Adaptive digital distance protection. IEE Conf. Publ. 302, 1989, pp. 187-191 6 ZHOU, K.M., MOORE, P.J., JOHNS, A.T., and XU, B.Y.: Accurate residual compensation for a digital distance relay. Proceedings of 24th Universities Power Engineering Conference, 1989, (Belfast), pp. 61-64 7 BRACEWELL, R.N.: Discrete Hartley transform, J. Opt. Soc. Amer., 1983, 73, (12), pp. 1832-1835 8 HARTLEY, R.V.L.: A more symmetrical Fourier analysis applied to transmission problems, Proc. IRE, 1942, 30, pp. 1 4 4 150 9 LONG-WEN, C . , and SHEN-WEN, L.: Systolic arrays for the discrete Hartley transform, IEEE Trans. Signal Process., 1991, 39, (11), pp. 2411-2418 10 LOBOS, T Fast estimation of symmetrical components in real time, IEE Pvoc C, 1992, 139, (1)
-

D = z[n]i[n - k ] - z[n - k]i[n] (15) where k is an arbitrary integer number (k is 6 in eqn. 7), and substituting eqns. 13 and 14 into eqn. 15, gives

D = I , sin(2~fonT)1,2nf0cos(2nfo(n- k ) T )
I , sin(2.irfo(n- k)T)lm2nf0cos(2nfo(n which rearranges to D = 2nfo1k s i n 2 ( 2 n f o n ~s)i n ( 2 n f o k ~ )
-

(16) k)T)

cos2(2xf012~ sin(2nfokir) ) which is equivalent to

+ 2nf&

(17)

D = 2nfO1; s i n ( 2 ~ f o k ~ )

(18)

8.2 Parameters used in power swing study

Line: Voltage = 400kV Z,, = 0.023 +JO.3165;2 per km Z, = 0.388 + jl.Ol9Q per km Length = 2 6 0 h Transformers: Each unit 221400kV, 600MVA Leakage reactance = 0 . 1 2 3 2 ~ ~ Generating plant: Single machine rating 600MVA, 22kV, 50Hz, 2 pole Armature resistance = 0.0031pu Direct axis reactance = 2.8pu Direct axis transient reactance = 0 . 3 6 2 5 ~ ~ Direct axis sub-transient reactance = 0 . 2 2 8 0 ~ ~ Quadrature axis reactance = 2 . 7 2 ~ ~ Quadrature axis transient reactance = quadrature axis sub-transient reactance = 0 . 2 2 0 2 ~ ~ Inertia constant = 4.44s

8. I

Proof of the general form of eqn. 7

The form of this equation is the same as the denominators of the right-hand sides of eqns. 1 and 2. It can easily be proved by reference to a sinusoidal signal:
2[n] = I,

sin(2nfonT)

where T = fs

(13)

and thus the time derivative is given by dnT Hence, forming the expression for D,

d(in) = i[n] = 1,2~fo cos(2nfonT)

(14)

26

IEE ProcGener. Transm. Distsih., Vol. 143, No. I , January 1996