Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

2017 International Conference on Advanced Systems and Electric Technologies (IC_ASET)

A Fault Classification Scheme with High Robustness for


Transmission Lines using Fuzzy-Logic system
Moez Ben Hessine, Sahbi Marrouchi; and Souad Chebbi
LaTICE Laboratory: Dept. of Electrical Engineering
National Higher Engineering School of Tunis, University of Tunis, TUNISIA
benhessinemoez@yahoo.fr, sahbimarrouchi@yahoo.fr, chebbi.souad@gmail.com

The electricity transmission is transported by the extra


Abstract: This paper present a new and accurate scheme high voltage transmission lines (EHV), these lines are
for fault classification of EHV transmission lines using subject to various disturbances (multi-phase faults or single-
fuzzy logic system. The fault classification scheme is phase faults) due with different situations. In order to
developed by using only the post-fault magnitude of maintain the continuity of service providing to customers, a
three phases current and its symmetrical components. protective relaying system is integrated in the electrical
The proposed fault classification technique is able to power system. the protective relaying systems is integrated
classify all faults type can be affect a transmission line in transmission lines to detect quickly the presence of faults
such as the single-phase to ground faults, two-phases and to isolate the faulted part from the rest of power system
faults, two-phases to ground faults and three-phases as soon as possible ensuring thus the power system stability,
faults with high accuracy under wide variety of fault minimizing damage equipment’s and restoring the service
quality.
conditions. The proposed scheme has a good
The Transmission line protection consists of three
performance in high fault conditions such as high fault
major tasks fault detection, classification and fault location
resistances, high fault inception angles and high fault
[1-3]. A fast detection of a fault occurred in transmission
distances to fault from relaying point. Large numbers of line allows fast isolation of the faulty line from service and
test cases are generated to verify the performance of thus offering protection from adverse effects of the fault.
proposed scheme. The simulation studies have been Once we know that a fault has occurred on an electrical
carried out by using Matlab software and Matlab fuzzy- transmission line, the next step is to identify the fault type
logic toolbox. into the different categories based on the phases that are
faulted. Then, the third step is designed to estimate the
Keywords— Fuzzy logic;Fault classifier; Protective relaying distance of the fault in the transmission line.
system; Fault classification; Extra high voltage; Transmission The determination of the fault type is known as
line
fault classification, which is an important aspect of the
protective relaying system for transmission lines. Fault
1 Introduction classification has been a topic of interest for several years
and as a result of this a number of fault classification
Electrical transmission lines are important elements for techniques have been developed by different researchers.
delivering to consumers the powers produced by the Today, Fuzzy Logic and Artificial Neural Networks
electrical energy production stations. A disturbance represent an area of intensive research in different
occurring on transport elements evacuating this production applications of system identification, control systems,
can have severe consequences. Sometimes a national or biomedical application, signal processing and fault
regional blackout can occur due to a fault on one of the diagnosis. Fuzzy logic, which is a mathematical tool based
components of the transport network. Among recent on fuzzy sets theory has rapidly become one of the most
blackout and the affected population we cite as an example successful technologies for developing sophisticated control
the incident of 30 June 2002 in Tunisia: 10 million people, systems today [1-10].
the incident of 3 February 2003 in Algeria: 25 million
people, the incident of 14 August 2003 in united states of 2 Basic of fault protection technique
America and Canada: 50 million people, the incident of 23
September 2003 in Sweden and Denmark: 5 million people, The protective relaying system in transmission line contain
the incident of 28 September 2003 in Italy: 57 million four major units, the first one called fault detection, this unit
people [1]. Thus we understand the particular attention with dedicated to identify the presence of the fault. Then it
which establishes the protections of Transport networks and activates the second unit called fault classification and fault
the means of locating faults. Fast fault location is one of the location units. The Fault Classification unit will interpret
major concerns of electricity companies to ensure the the faults values of the voltage and current measurements.
continuity of service and to reduce the duration of According to these values, it will be capable to identify the
interruptions. fault type if it is single phase to ground (L-g), double phase
(L-L), double phases to ground (L-L-g), or three phases (L-

978-1-5090-6634-6/17/$31.00 ©2017 IEEE 256


2017 International Conference on Advanced Systems and Electric Technologies (IC_ASET)

L-L). On the other hand the Fault location unit will also take logic estimator FL1 reserved for identify the participation of
the faulty values of the voltage and the current ground or not. Fig. 2 shows the fuzzy logic scheme for fault
measurements. classification. The variables treated by the second and the
According to these values, it will be able to estimate the third fuzzy logic estimators (FL2, FL3) responsible for
exact distance to fault from the relaying point. Finally the determining the characteristics of different fault types are
control unit will receive the decisions from the three units in (E1, E2 and E3). These variables are calculated on the basis
which the output of the fault detection unit will trip the of the ratios from magnitudes phase currents, which are
circuit breaker at the fault instance through the control calculated as described in section 3.1.2.
circuit as shown in Fig. 1. However, the outputs of the Fault
classification and the Fault location units will offer an aid to 3.1 Input variables of the proposed fault classifier
identify the fault type and location respectively.
Bus i Current
transformor Circuit breaker
During the fault occurrence in electrical power system, the
conventional control techniques are no longer valid to be
extended to solve the problems caused by faults.
Currents Exceptionally, the symmetrical components method is the
Voltages

most commonly used for the treatment of unbalanced states.


It will be used because it provides considerable simplicity in
Voltage
transformor
Fault Fault Fault location
the analysis of problems of fault classification in
detection unit classification
unit
unit
transmission lines. In this paper, the positive sequence
Control Unit

Fault/no fault Fault type Distanceto fault


components I1, the negative sequence components I2 and the
zero sequence components I0 of the fault currents are
Protection Relay
introduced into the proposed algorithm for fault
classification.
Fig. 1. proposed fault protection technique.
3.1.1 Ground detection criterion
3 Developed Fault Classification scheme
The analysis of the fault current symmetrical components
The fault classification scheme has been developed and gives information about the nature of the fault. The presence
integrated on an extensive base of simulation studies carried of zero sequence components (I0) indicates that a fault
out in the power system simulated using Matlab software. involving ground has occurred. For that, in order to
The proposed technique used only the magnitude of the determine the involvement of ground in fault the value of E0
three phase currents (Ia, Ib, Ic) during the occurrence of a has been considered. It has been observed that the values of
fault, generated by Matlab, and are considered for the E0 are high (greater than 0.1) for faults involving ground and
classification algorithm. low (less than 0.1) for faults not involving ground. The
simulations results confirm that the involvement of ground in
a fault can be easily detected on the basis value of E0. From
the magnitudes of symmetrical components of fault
Fortescue
Transformer currents, the ratio E0 is calculated by Eq. 1.
I1 I2 I0
Calculate bloc
E0 Max abs ( I 0 )
uj
E0 = (1)
x
Max abs ( I 1 )
FL1: Ground Participation

Consequently, the variable E0 is treated by the first fuzzy


logic estimator FL1 in order to identify the participation of
ground or not, the appropriate fuzzy rule base is used for
uj uj

detection of ground fault.


x x

FL2: Ground fault FL3: Isolated fault

L-g faults L-g faults


Fuzzy rule base for detection ground faults:
L-L-g faults L-L-g faults
• if E0 is Low it is not ground faults;
Fig. 2. Fuzzy block diagram of the phase selector. • if E0 is High it is not ground faults.
The variables E0 in the above fuzzy rules should be fuzzy
The adopted fault classification technique consists to variables. The triangular membership function, shown in
identify and separately classify the faults related to ground Fig. 3, has been used to represent this variable.
and the isolated faults to ground. The proposed technique is 1
based on three fuzzy logic estimators (FL1, FL2 and FL3).
Degree of membership

0.8
Low High
FL2 is responsible for the classification of all faults related 0.6

to ground such as the single phase to ground faults (L-g), 0.4

0.2

and the double phase to ground faults (L-L-g). FL3 is 0

designed to classify the isolated faults such as double phase 0 0.2 0.4
E0
0.6 0.8 1

faults (L-L) and three phase faults (L-L-L). This action is Fig. 3. Fuzzy variables and membership functions for E0.
initiated by the value of E0 to be treated by a first fuzzy

257
2017 International Conference on Advanced Systems and Electric Technologies (IC_ASET)

3.1.2 Phase angle criteria • if E1 ∈ MFghigh et E2 ∈ MFgMed et E3 ∈ MFgLow then :


Fault type is: a-g;
The characteristicof the different fault types (related and
isolated faults to ground) are determined in terms of E1, E2 • if E1 ∈ MFgLow et E2 ∈ MFgHigh et E3 ∈ MFgMed then :
and E3, which are calculated asdescribed below. First, the Fault type is: b-g;
post-fault magnitudes currents as mentioned above, the • if E1 ∈ MFgMed et E2 ∈ MFgLow et E3 ∈ MFgHigh then :
ratios R ab , R bc and R ca , are calculated as follows: Fault type is: c-g;
• if E1 ∈ MFgLow et E2 ∈ MFgHigh et E3 ∈ MFgLow then :
Max I a Max I b Max I c
R ab = ; R bc = ; R ca = (2) Fault type is: a-b-g;
Max I b Max I c Max I a
• if E1 ∈ MFgLow et E2 ∈ MFgLow et E3 ∈ MFgHigh then :
Fault type is: b-c-g;
Knowing that the magnitudes of these quantities may
possibly exceed the extreme -1 and 1 that the fuzzy system • if E1 ∈ MFgHigh et E2 ∈ MFgLow et E3 ∈ MFgLow then :
terminal is able to process, and in order to make these Fault type is: c-a-g;
calculated ratios compatible with tolerable margins of study, if E1 ∈ et E2 ∈ MFph et E3 ∈ MFph then :
Low High High
• MFph
a normalization technique of these ratios R ab , R bc and R ca Fault type is: a-b;
will be of great importance. For that reason, the normalized
if E1 ∈ et E2 ∈ MFph et E3 ∈ MFph then :
Low Low High
• MFph
values characterizing the inputs of the fuzzy logic system
are calculated as follows: Fault type is: b-c;
if E1 ∈ et E2 ∈ MFph et E3 ∈ MFph then :
High Low Low
• MFph
n R ab Fault type is: c-a;
R ab = ;
if E1 ∈ et E2 ∈ MFph et E3 ∈ MFph then :
Max ( R ab , R , R ca ) Low Low Low
bc • MFph
R bc (3) Fault type is: a-b-g.
R nbc = ;
Max ( R ab , R bc , R ca )
n R ca 1
a-g b-g c-g a-b-g b-c-g c-a-g a-b b-c c-a a-b-c

R ca =
Max ( R ab , R bc , R ca ) 0.8
Degree of membership

0.6

0.4

Finally, the differences of these normalized values are 0.2

determined as follows: 0

0 5 10 15
Outputs

n Fig. 4. Fuzzy variables in the consequent parts and related decimal


E1 = R ab − R nbc ; E 2 = R nbc − R ca
n n
; E 3 = R ca n
− R ab (4) numbers to represent fault type.
1

As already mentioned, the characteristic of different fault 0.5


0
Low-g Med-g High-g

types are determined in variable E1, E2 and E3. For that, the
Degree of membership

-1 -0.5 0 0.5 1
E1
fuzzy rule matrix for fault classification is developed on the 1
0.5 Low-g Med-g High-g
basis of these variables E1, E2 and E3. 0
-1 -0.5 0 0.5 1
E2
1
3.3 Rules matrix 0.5 Low-g Med-g High-g
0
-1 -0.5 0 0.5 1
E3
As previously described in the previous section, the
characteristics of different fault types are determined in Fig. 5. Fuzzy variables and membership functions for E1, E2 and E3 for
terms E1, E2 and E3represented by fuzzy variables. Thus to ground fault.
classify the different fault type, the fuzzy rule base is 1

developed and based on the variations of these terms. In this 0.5


0
Low-phase Medium-phase High-phase

context, the fuzzy rules for the proposed fault classification -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Degree of membership

Input1
of all fault types may affect the transmission lines are 1

developed according the following diagram: 0.5

0
Low-phase Medium-phase High-phase

To represent the variables in consequent parts in the form of -1 -0.8 -0.6 -0.4 -0.2 0
Input 2
0.2 0.4 0.6 0.8 1

fuzzy variables, each fault types has been considered equal 1

to one decimal number, which is demonstrated in Fig. 4. In 0.5


0
Low-phase Medium-phase High-phase

this figure if the crisp output of the fuzzy logic estimator -1 -0.8 -0.6 -0.4 -0.2 0
Input 3
0.2 0.4 0.6 0.8 1

lies between [1.5, 2.5], represents "a-g" fault has been


occurred. As shown in Figs. 4 to 6, triangular fuzzy Fig. 6. Fuzzy variables and membership functions for E1, E2 and E3 for
phase fault.
membership functions have been used for all of input and
output quantities.
Fuzzy rule base for FL1 and FL2:

258
2017 International Conference on Advanced Systems and Electric Technologies (IC_ASET)

4 Fault classifier performance


Table I: Simulation results in case of
4.1 Model system phase to ground faults
Fault conditions Outputs variables
The considered power system is shown in Fig. 7. The power
Lf Rf FIA Desired Actual
system model is composed of a synchronous generator (pu) (Ω) (°) (°) Output Output
connected to infinite bus using a transmission line which is 0.2 5 15 10 2.0011
presented by distributed parameters [1-3].This system was 0.4 25 25 40 2.0032
simulated with Matlab for fault analysis and producing test 0.6 50 35 60 2 2.0019
cases for examining the performance of fuzzy logic-based 0.8 150 45 120 1.9901

(a-g)
fault classifier. 0.95 200 90 360 1.9806
0.2 5 15 10 3.0012
• Line length=100km; 0.4 25 25 40 3.0170
• Transmission line impedance: 0.6 50 35 60 3 2.9910
positive sequence impedance = 0.0275+j0.422 0.8 150 45 120 3.0034

(b-g)
o
Ω/km 0.95 200 90 360 3.0071
o zero sequence impedance = 0.275+j1.169 Ω/km 0.2 5 15 10 3.9901
0.4 25 25 40 4.0177
o Positive sequence capacitance = 9.483 nF/km 4
0.6 50 35 60 4.1019
o zero sequence capacitance = 6.711nF/km
0.8 150 45 120 4.0051

(c-g)
Fault 0.95 200 90 360 4.0127
VS∠0 VR∠ δ L
Bus i Bus j
CT Table II: Simulation results in case of
CB
double phase and three phase faults
Synchr. Gen
615 MVA
Transformer VT Transmission Line Infinite Bus
400Kv
Fault conditions Outputs variables
Relay
Lf Rf FIA Desired Actual
(pu) (Ω) (°) (°) Output Output
CT : Current Transformer Lf .ZLine
F
(1 − Lf ).ZLine 0.2 5 15 10 10.9827
VT : Voltage Transformer
CB : Circuit Breaker 0.4 25 25 40 10.9912
(a-b)

δL : Load angle (1 − L f ).Ycc (1 − Lf ).Ycc


10
L f .Ycc L f .Ycc

Lf : Fault location 2 2 2 2
0.6 50 35 60 10.0019
Transmission Line Model
0.8 150 45 120 9.9941
Fig. 7. The power system model. 0.95 200 90 360 9.9869
0.2 5 15 10 11.0099
0.4 25 25 40 11.0798
4.2 Simulation and testing
(b-c)

0.6 50 35 60 11 10.9945
0.8 150 45 120 10.9701
All possible types of faults (a-g, b-g, c-g, a-b, b-c, c-a, a-b-c, 0.95 200 90 360 11.1009
b-c-g, c-a-g, a-b-c) have been simulated with a different 0.2 5 15 10 12.0460
fault scenarios. The test cases are produced for different 0.4 25 25 40 12.0149
(c-a)

0.6 50 35 60 12 12.0666
distance to fault (Lf), different fault resistance (Rf), different
0.8 150 45 120 12.1300
load angle (δ) and different fault inception angle (FIA). In 0.95 200 90 360 12.0734
this paper, the proposed fault classification algorithm is 0.2 5 15 10 14.0091
tested under the following fault scenarios: 0.4 25 25 40 13.9988
0.6 50 35 60 14 14.0197
(a-b-c)

• Five distance to fault (Lf): 0.2L, 0.4L, 0.6L, 0.8L 0.8 150 45 120 14.1009
0.95 200 90 360 14.1211
and 0.95L from the relaying point, where L is the
total length of the line.
• Five fault resistance (Rf): 5, 25, 50, 150 and 200Ω. In order to test the performance of the proposed fault
• Five load angle (δ): 15°, 25°, 35°, 45° and 90°. classification scheme, a large number of fault simulation
studies, were examined. The proposed fuzzy logic-based
• Five fault inception angle (FIA): 10°, 40°, 60°,
fault classification scheme, classified all the faulted
120° and 360°.
generated test cases correctly for all the mentioned values of
fault location, fault resistance, load angle and fault inception
Simulation studies are achieved with different combinations
angle. It was observed that classification performance of the
that should be studied by considering the above-mentioned
proposed method is more accurate in a wide variety of
values for fault location, fault resistance, load delta, and
system pre-fault conditions. As an example, from Fig. 4, if
fault inception angle. For each of these combinations, all ten
the crisp output of the fuzzy logic system estimators lies
types of faults (i. e., a-g, b-g, c-g, a-b, b-c, c-a, a-b-g, b-c-g,
between [5.5, 6.5], an "a-b-g" fault will be indicated.
c-a-g, a-b-c) should be applied, resulting a different studied
The above results reveals that the proposed approach is
fault simulation. The performance of the proposed fault
computationally simple in comparison to other conventional
classifier is checked for all of these test cases shown in
approaches and yields classification in less than half power
Tables I and II.

259
2017 International Conference on Advanced Systems and Electric Technologies (IC_ASET)

frequency cycle. Moreover it provides a high accuracy of Fig. 9. Output of fault classification during b-c fault at 0.1p.u with fault
resistance of 10Ω, FIA=30° and δ=45°.
fault’s identification.
In the other hand, we have simulated the outputs for each 5 Conclusion
fuzzy logic estimator based on fault classifications for
certain faults. We have taken as examples the single phase A fuzzy logic based algorithm for fault classification in
to ground fault "a-g" and the double phase fault "b-c" under extra high voltage transmission line has been presented. The
these fault conditions (Lf=0.1p.u, Rf=10Ω, FIA=30° and proposed algorithm requires the consideration only the three
δ=45°). phase current measurement at one end of transmission line.
The Discrete Fourier Transform (DFT) is used to determine
Fig. 8 shows the outputs of fuzzy logic estimators (FL1, FL2
the magnitude of the three phase currents and the positive,
and FL3) based fault classification for "a-g" fault. Based on
negative and zero sequence components of fundamental
these simulation it can be seen that the outputs is low (0) up
frequency of these currents. The time taken by this method
to 70 ms of time shows that there is no fault situation. After is about one cycle for a 50 Hz system. Three fuzzy logic
70 ms it started increasing and reaches high "1" by for FL1 estimators are used for accurate identification of the faulted
and "2" for FL2 and "0" for FL3 at 79 ms of time. We base phase. The first one reserved for determine the participation
on the fuzzy logic rules the fault occurred is "a-g" fault. So of ground or not, the second use for identify the faults
fault detection time in this case is 7 ms. related to ground and the third for determine the isolated
Similarly, Fig. 9 shows the output of fault phase faults. The proposed fuzzy logic-based fault classification
identification and fault classification for double phase Fault scheme identified all of the test cases correctly. The
"b-c". After 70 ms of time the output of FL2 started performance of the proposed method is not affected by a
increasing to high "11" shows that the fault is a-b fault. wide variety of pre-fault system loading level, fault
Fault classification time is within half cycle time in this resistance and fault distances from the relaying point.
case. Proposed fuzzy logic based protection relay accurately
detect the fault for all fault cases. References
Simulation results prove the ability of the fault classifier to
produce a correct response in all test cases. In addition, the [1] Moez, B.H., Houda, J. and Souad, C. (2015) ‘Neural
results show the stability of outputs of the fuzzy logic and Network Approach to Fault Location for High Speed
the fast convergence of the output variables to the desired Protective Relaying of Transmission Lines’,
Computational Intelligence Applications in Modeling
values for different fault scenarios. This confirms the and Control, Studies in Computational Intelligence,
effectiveness of the proposed fault classification scheme. Vol. 575, pp.283–314.
The results show that the fuzzy logic is able to generalize [2] Moez, B.H., Houda, J. and Souad, C. (2014) ‘Fault
the situation from the provided templates accurately detection and classification approaches in transmission
indicates the fault type and can be used for online fault lines using artificial neural networks’, Proceedings of
the 17th International Mediterranean Electrotechnical
classification in transmission lines. Conference (Melecon), Beirut, Lebanon, pp. 520-524.
[3] Moez, B.H. and Souad, B.S. (2014), ‘Accurate Fault
X: 79
Y: 0.99 Classifier and Locator for EHV Transmission Lines
1 based on Artificial Neural Networks’, Mathematical
0.5
X: 70
Y: 0
Problems in Engineering, ID 240565, Vol 2014.
Output of fault phase identification

0
40 50 60 70 80 90 [4] Chuansheng, X., Chenchen, Z., Dapeng, D, and
Output-FL1 Pengyuan, Z. (2014) ‘Electric vehicle industry
development environment evaluation in China based on
and classification

2
X: 79
BP neural network’, Int. J. of Simulation and Process
Modelling, Vol.9, No.4, pp.234 –239.
X: 70
1 Y: 0 Y: 1.98
0
40 50 60 70 80 90 [5] Shahram, T. and Elham, M. (2015) ‘Using discrete
Output-FL2 event visual simulation to teach process modelling in
1 MBA operations management courses’, Int. J. of
0
Simulation and Process Modelling, Vol.10, No.1,
pp.45– 64.
-1
0 20 40 60 80 100 [6] Galina, M. and Vitaly, B. (2014) ‘integrated planning
Output-FL3
and scheduling built on cluster analysis and simulation
optimisation’, Int. J. of Simulation and Process
Fig. 8. Output of fault classification during a-g fault at 0.1p.u with fault Modelling, Vol.9, No.1/2, pp.81–91.
resistance of 10Ω, FIA=30° and δ=45°. [7] Letizia, N., Alessandro, C., Carlos, A., and Rafael D.
(2014) ‘Hybrid approach for container terminals
1
performances evaluation and analysis’, Int. J. of
0 Simulation and Process Modelling, Vol.9, No.1/2,
pp.104 – 112.
Output of fault phase identification

-1
0 20 40 60
Output-FL1
80 100
[8] Dong, X., Kong, W., and Cui, T. (2009) ‘Fault
and classification

1 classification and faulted-phase selection based on the


0 initial current travelling wave’, IEEE Trans. Power
-1
Delivery, Vol. 24, No.2, pp.552-559.
0 20 40 60
Output-FL2
80 100
[9] Gohokar, V.N. and Khedkar, M.K. (2005)‘Faults
20 locations in automated distribution system’, Electr.
10 X: 70
Power Syst. Res, Vol. 75, pp, 51–55.
0
Y: 0 X: 78
Y: 10.94 [10] Florez, J.M., Melendez, J.G. and Carrillo, C. (2008)
40 50 60 70 80 90 ‘Comparison of impedance based fault location
Output-FL3

260
2017 International Conference on Advanced Systems and Electric Technologies (IC_ASET)

methods for power distribution systems’, Electr. Power


Syst. Res, Vol. 78, pp.657–666.

261

You might also like