Transmission Line Differential Protection

with Fuzzy Signal Processing Support

Waldemar Rebizant1, Krzysztof Solak1
1
Wroclaw University of Technology, Wybrzeze Wyspianskiego 27, 50-370 Wroclaw, Poland
[email protected], [email protected]

Abstract Several approaches may be found in the literature, that

according to the authors, should improve performance of the
This paper presents a new differential protection scheme for line differential relays. The solution presented in [3] is based on
transmission lines with application of fuzzy signal zero-sequence component for detection of current transformer
processing. Traditional differential relays may have saturation. The idea described in [5] makes use of adaptive time-
problems with proper classification of external faults with dependent restraint coefficients that define the shape of
CT saturation. Better protection stabilization for such cases percentage differential curve. Since the two cited solutions do
is obtained with support of fuzzy signal processing. In not guarantee proper operation of the relay for all conditions
proposed solution the input signals as well as the standard either, new protection ideas are still needed to assure improved
percentage characteristic are fuzzified. The performance of protection performance. The newly proposed solution fulfills the
presented fuzzy protection scheme has been tested with the above requirements, with simultaneously maintained sensitivity
signals generated with use of EMTP-ATP program and and operation speed for internal faults calling for prompt
compared to the traditional solution. tripping.

1. Introduction Id

Differential protection is a commonly accepted protection of

single and parallel transmission lines, if only appropriate
communication link connecting all line terminals is available [1,
Operate
2, 3]. The zone of action of differential relay embraces only k2
protected object, which means that differential relay should trip
for internal faults only and restrain for all external disturbances.
In standard solutions the stabilized characteristic (Fig. 1) is
applied and the trajectory of differential/bias currents is tracked
with respect to the relay characteristic to determine whether or k1
Id0 Restrain
not to trip the transmission line [2].
The standard differential relay percentage curve is
determined by four protection settings, [2]: Id0=0.3In, Is2=2In, Is2 Ibias
k1=0.3 and k2=1.5 (specific values used for testing purpose). The
tripping is initiated if: Fig. 1. Stabilized characteristic of the current differential relay

 I d t I op k1 ˜ I bias  I d 0 f or I bias d I s 2   2. Fuzzy protection scheme developed

 I d t I op k 2 ˜ I bias  (k 2  k1 ) ˜ I s 2  I d 0 f or I bias t I s 2  (2) Classical (Boolean) logic based on the concept of truth/
falsity cannot effectively cope with the many ambiguities that

with arise during operation of the power system. Therefore, fuzzy
logic is increasingly being used in decision-making, whereas the
iS  iR criteria signals are described by membership functions. The use
 Id iS  iR 
I bias (3)
2 of fuzzy logic increases the confidence of the decision-making
where: iS and iR being currents measured at line terminals, Ibias – within an area of uncertainty, since the fuzzy logic can deal
amplitude of bias current, Id – amplitude of differential current, better (as compared to Boolean logic) with suspense and
Iop – relay operating current, In – line nominal load current. missing data. In addition, inferencing with multiple objectives in
The majority of external faults are usually not a big problem such systems is a natural way of processing information – it is
for the differential relay. Generally, CT errors due to saturation therefore utterly possible to use numerous criteria in parallel.
during external faults are compensated for by conventional Fig. 2 presents the structure of the new fuzzy protection. The
stabilized characteristic with adequate slope setting. However, main idea of action relies on fuzzification of differential current
when there is a mismatch in CTs’ load or they have non- Id that is further compared with fuzzy setting obtained on the
identical magnetizing characteristics, a possibility still exists basis of the stabilized characteristic (Fig. 1). Additionally, the
that one of the CTs saturates and not the other, which may lead criterion of phase difference is determined, value of which
to unwanted protection reaction [4].

I-162
 1 I min(n) min {I d (n  k )} (7)
Digital k 0 y( N / 4) 1
iS(n) filtering 2
Id(n) P(Id(n)) 1 ( N / 4)1
¦ I d (n  k )
&
calculation
Fuzzification I av (n) (8)
iR(n) 6 N /4 k 0
of criteria
signals Fuzzy Tripping
comparison decision
I max(n) max {I d (n  k )} (9)
k 0 y( N / 4) 1
3 5
Ibias(n), Calculation P(Iop(n))
Fuzzy setting where: N – measuring window length (here N=20).
MF(n) of Iop
An example of how the fuzzification of differential current
4 proceeds is shown in Fig. 3. Based on five samples of magnitude
Phase IP(n) of differential current (Fig. 3a) the adequate values are
comparison calculated according to equations (7), (8) and (9). Next, the
element
triangle membership function is formed as shown in Fig. 3b.
Calculation of Iop (block 3) – the value of operation current is
Fig. 2. Block scheme of the fuzzy adaptive differential calculated according to equations (1) and (2) – based on bias
protection of transmission line current Ibias.
Phase comparison element (block 4) – here the calculated
affects the degree of fuzzification of fuzzy setting. Below the phase difference, (4) or (5), is compared with the operation
various blocks of scheme from Fig. 2 are described in detail. characteristic (see Fig. 4). The adequate threshold values of the
Digital filtering and calculation of criteria signals (block 1) characteristic have been set according to the statistical
– here the main criteria signals (differential current Id (3), bias
current Ibias (3) and phase difference φF) are calculated with use a)
of full cycle Fourier filters. The variable φF can be expressed by 3.5
the formula:
- for asymmetrical faults it is calculated on the basis of 3
negative sequence current since it gives excellent fault
discrimination for such faults [6]: 2.5

i 2S 2
Id [A]

MF 180o  arg
i 2R (4)
1.5

- unfortunately, the negative sequence current can not provide 1

three-phase faults identification. Therefore, for symmetrical
faults the phase difference is calculated on the basis of 0.5
positive sequence current as follows:
0
0.115 0.116 0.117 0.118 0.119 0.12
i1S t [s]
MF 180o  arg
i1R (5) Fuzzification
of differential
b) current
A three-phase fault is detected using overcurrent element
tracking the level of bias currents in all phases. 1
Symmetrical components of the signals can be calculated 0.9
according to the well known matrix formula:
0.8

ªi º 0.7
ª1 1 1 º ª i L1S ( R ) º
« 0S ( R) » 1« »« » 0.6
« i1S ( R ) » 1 a a 2 » «i L 2 S ( R ) » (6)
3«
P(Id)

«i » «¬1 a 2 a »¼ «i L3S ( R ) » 0.5

¬ 2S ( R) ¼ ¬ ¼
0.4
where: a=exp(j2π/3), i0S(R), i1S(R), i2S(R) – zero, positive, negative 0.3
sequence currents at the S and R ends of the line, iL1S(R), iL2S(R), 0.2
iL3S(R) – phase currents at the S and R ends of the line.
Measuring of phase difference is initiated when the 0.1
differential current is greater than or equal to Id0 in any phase. 00
0.5 1 1.5 2 2.5 3 3.5 4 4.5
Fuzzification (block 2) – magnitude of differential current (3) Id [A]
is fuzzified, which means that triangular membership functions Imin Iav Imax
is formed by using minimum Imin, average Iav and maximum Imax
values of differential current (it was assumed that these values Fig. 3. Fuzzification of differential current: a) magnitude of
were calculated for a quarter of fundamental frequency cycle): differential current, b) fuzzy differential current

I-163
information gained through analysis of generated simulation P(Iop)
signals. The output value IP from phase comparison element 1
influences fuzzification of operation current. If the output value P(Idiff)
is close to 1.0 it indicates an external fault. Otherwise (internal
fault cases) the output is close to 0.0. P
Fuzzy setting (block 5) – based on the actual value of
operation current and information from phase comparison block

P(I)
the fuzzy setting is formed as it is illustrated in Fig. 5. The
parameters G1 and G2 determine the fuzzification of membership
function of fuzzy setting and they can be calculated according
to: P1

G1 IP ˜ I bias  0.1 G2 1.5 ˜ IP ˜ I bias  0.3 (10)

0
The values of parameters in (10) are small (G1=0.1 and I [A]
G2=0.3) for IP=0 (this value indicates internal fault) which
Fig. 6. Fuzzy comparison illustration
means that membership function is slightly fuzzy. When IP=1
(this value indicates external fault) both parametres are high and
the membership function of fuzzy setting is quite broad. and P1 - surface area (hatched) lying under a fuzzy setting P(Iop),
Fuzzy comparison (block 6) in this block both membership but within P(Id), [7].
functions fuzzy differential current P(Id) and fuzzy setting P(Iop)
are compared with each other (Fig. 6). The value of fuzzy
comparison is determined by relation (11), where P - the area Pd
³ min[P (I ), P (I
d op )]dI P
(11)
under the membership function of differential current P(Id) ³ P (I ) dI
d
P1

The final decision to trip a protected transmission line is

1 taken when the value of index Pd is greater than threshold 0.7.

0.8 3. Testing of developed fuzzy protection scheme

The idea presented above has been subjected to extensive

IP [pu]

0.6
simulative testing in order to prove its efficiency. The basic
system with HV transmission line under study is shown in Fig.
0.4 7. The overhead transmission line is modeled as transposed one
with distributed parameters frequency dependent JMarti model
[8, 9]. The line of 50km length can be divided into two sections,
0.2 so that internal faults (FL) at almost arbitrary location along the
line can be simulated. External faults are those modeled at
0 busbars (location FBS(R) in Fig. 7).
0 20 40 60 80 100 120 140 160 180 The line is supplied from both sides, where the sending
MF [degrees] equivalent system is assumed to be strong (of high short-circuit
power), while the receiving one is weaker. The power flow can
Fig. 4. Operation characteristic for phase difference
be controlled by variable angle of the receiving source.
The transient response of CTs and the correct models in
1
ATP-EMTP simulation are very important for the evaluation of
0,9 high-speed relaying systems [10]. The 5P30 20VA 1000/1A
0,8 CTs were modeled using the TYPE-96 pseudo-nonlinear
element. In this model there is a possibility to set the residual
0,7
P(Iop(n))

0,6 FBS FBR

FL
0,5
ES ZS CBS CTS ZL CTRCBR ZR ER
0,4
0,3
l=50km
S R S =4GVA
SS=32GVA RS RR R
0,2 G G
Z0/Z1=0.8÷3.0 Z0/Z1=0.8÷3.0
0,1 L/R=0.03 or 0.1 L/R=0.03 or 0.1
0 Z0L=(0.2750+j1.0265) Ω/km C0L=8.5 nF/km
Iop -G Iop Iop +G Z1L=(0.0276+j0.3151) Ω/km C1L=13 nF/km

Fig. 5. Formation of fuzzy setting Fig. 7. Model of the power system with transmission

I-164
flux in the CT core, which is very important for studying CT a)
saturation effects [3]. 1
[0: restraint 1.0: trip]

Detection
Thorough studies have been performed by varying the 400kV
power system parameters, which resulted in over 20000 0.5
different simulation cases. The parameters being changed as
systems strength, fault type, fault resistance, point on wave, 0
residual flux, etc. 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
The figures below present testing results fuzzy scheme
proposed and standard differential relay [2] for internal and b)
external fault with CT saturation. 1

Detection
In Figs. 8-11 an example of three phase external fault at busbar [0: restraint 1.0: trip]
FBS with CTs saturation is presented. The CTs get saturated 0.5
especially in phase L1 at sending end (Fig. 8) and the standard
protection based on the stabilized characteristic with fixed 0
settings maloperates, since the differential-restraining trajectory
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
(phase L1) enters the tripping zone (Fig. 9) and the trip
command is sent to the circuit breakers (Fig. 10a). On the Time (sec)
contrary, proposed algorithm remained fully stable without
Fig. 10. Relay response for L1-L2-L3 external fault (FBR):
issuing false tripping command – it effectively blocks this
a) standard differential protection, b) fuzzy differential
external fault (Fig. 10b).
protection
phase L1 (A)

The relaying schemes were tested for more than 3000

Currents in

5 side R external fault cases obtained from EMTP simulation. The testing
0 results proved that the proposed scheme is immune to external
-5 side S faults (zero percent of incorrect operation). Contrary, the
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 standard protection failed for a few percent of external fault
phase L2 (A)

cases.
Currents in

5
The developed adaptive protection scheme has also been
0 tested for the cases of internal faults, for which unambiguous
-5 tripping command should be issued. A case of L1-L2-L3 internal
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 fault (fault resistance 0Ω) at point FL (fault location 7km from
the sending end) is shown in Figs. 11-13. One can notice that
phase L3 (A)
Currents in

5 when an internal fault occurs the CTs at the sending end deeply
0 saturate (Fig. 11). The trajectory Id-Ibias enters the tripping zone
for all three phases and standard protection properly detects this
-5
case (Fig. 13a). The two algorithms (standard and proposed)
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
detect this internal fault within less than 5ms after fault
Time (sec)
inception.
After analysis of testing results of both protection for over
Fig. 8. Line terminal current waveshapes in case of L1-L2-L3
21000 simulated internal fault cases one can say that average
external fault at busbar (FBR)
time detection of both protections are quite similar.
5
40
phase L1 (A)
Currents in

4.5 phase L1 20 side S

phase L2 0
4 phase L3
Differential current, Idiff [A]

-20 side R
3.5 -40
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
3 40
phase L2 (A)
Currents in

20
2.5 Operate 0
2 -20
-40
1.5 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
40
phase L3 (A)
Currents in

1 Restrain 20
0.5 0
-20
0 -40
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Bias current, Ibias [A]
Time (sec)
Fig. 9. Line protection stabilized characteristic and Id-Ibias Fig. 11. Line terminal current waveshapes in case of L1-L2-
trajectory in case of L1-L2-L3 external fault at busbar (FBR) L3 internal fault at point FL

I-165
 [4] I. Voloh, B. Kasztenny, C.B. Campbell, " Testing line
35 phase L1 current differential relays using real-time digital
phase L2
phase L3 simulators," in Proc. 2001 IEEE/PES Transmission and
30
Differential current, Idiff [A]

Distribution Conference and Exposition, pp. 516 – 521.

[5] Gang W., Baoji Y., Jiali H., Li K.K.: "Implementation of
25
Adaptive Dispersed Phase Current Differential Protection
20 for Transmission Lines", Proceedings of the 5th
Operate International Conference on Advances in Power System
15 Control, Operation and Management, APSCOM 2000,
Hong Kong, October 2000, pp. 64 - 69.
10 [6] Kasztenny B., Voloh I., Udrean E.A.: "Rebirth of the Phase
Restrain Comparison Line Protection Principle", 59th Annual
5 Conference for Protective Relay Engineers, 4-6 April
2006, pp. 193 – 252.
0 [7] Rebizant W., Wiszniewski A., Schiel L.: "Acceleration of
0 2 4 6 8 10 12 14 16 18 20
Bias current, Ibias [A] Transformer Differential Protection with Instantaneous
Criteria", Proceedings of International Conference on
Fig. 12. Line protection stabilized characteristic and Id-Ibias Advanced Power System Automation and Protection, Korea,
trajectory in case of L1-L2-L3 internal fault at point FL 24-27 April 2007, CD-ROM, paper 505.
[8] H.W. Dommel, "Electromagnetic Transients Program.
a)
Reference Manual (EMTP theory book)", Bonneville
1 Power Administration, Portland 1986.
[0: restraint 1.0: trip] [9] J.R. Marti, "Accurate modelling of frequency–dependent
Detection

0.5 transmission lines in electromagnetic transients

simulation", IEEE Trans. Power App. Syst., vol. PAS–101,
0 January 1982, pp. 147–157.
[10] D.A. Tziouvaras, P. McLaren, G. Alexander et. al.:
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 "Mathematical Models for Current, Voltage and Coupling
Capacitor Voltage Transformers", IEEE Transactions on
b) Power Delivery, Vol. 15, No. 1, January 2000, pp. 62 - 72.
1
Detection

[0: restraint 1.0: trip]

0.5

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Time (sec)

Fig. 13. Relay response for L1-L2-L3 internal fault (FL):

a) standard differential protection, b) fuzzy differential
protection

4. Conclusions

The solution for improvement of the line differential

protection under external fault cases is described in this paper.
The results of fuzzy protection performance testing prove that
the proposed algorithm remains fully immune to current
transformer saturation during external faults. The proposed
algorithm is also able to detect internal faults, even those with
severe CT saturation within the same time like traditional
solutions.

5. References

[1] S. Ward, T. Erwin, "Current Differential Line Protection

Setting Considerations", RFL Electronics Inc, 2005.
[2] AREVA, "P54x Application Guide", 2005.
[3] N. Villamagna, P.A. Crossley, "A CT Saturation Detection
Algorithm Using Symmetrical Components for Current
Differential Protection", IEEE Transactions on Power
Delivery, vol. 21, No. 1, pp. 38-45, January 2006.

I-166