Proceedings of the 14th International Middle East Power Systems Conference (MEPCON10), Cairo University, Egypt, December 19-21,

2010, Paper ID 202.

Fault Identification of Overhead Transmission Lines Terminated with Underground Cables

A. Elmitwally, S. Mahmoud, and M. H. Abdel-Rahman
Elect. Eng. Dept., Mansoura University, Mansoura, 35516, Egypt, [email protected]

Abstract --- In this paper, a scheme for fault detection and

identification of three phase overhead transmission lines ended with underground cables is proposed. Fault detection technique is based on mean square value of the difference between incoming and out going three phase currents of each section. Theses differences are compared against threshold setting values. Faulty phase identification is based on the analysis of three phase currents at one end of transmission line. The transient currents are processed by Discrete Wavelet Transform multi-resolution analysis. In order to find out the faulty phase, the absolute sum of the Wavelet Transform coefficients of the third level detail d3 of each phase current, corresponding to the frequency band 25-12.5 kHz, is computed over one cycle of power frequency. It is used as input to a rule-base system to identify the fault type. Many case studies are provided to validate the proposed algorithm.

Keywords: Transmission lines, Underground cables, Discrete

wavelet transform, Faults, Identification.
I.

INTRODUCTION

Modern fault classification techniques can be subdivided into several methods. One of the main methods depends on detecting the waveform of currents, voltages, or both of them. This technique depends on fast sampling and monitoring these waveforms prior to and/ or after fault occurrence [1, 2]. In the past, some papers have been reported in the area of waveform distortion, most of them deal with the effect of fault generated transients on the existing protection relays mainly using Fourier analysis [3]. However, Fourier analysis only provides the frequency information and usually losses the time information. Although short time Fourier transform overcomes the time location property, it does not provide multiple resolution in frequency and time, which is a potential feature for analysis signals containing both high and low frequency component together. Wavelet algorithms [4] process data at different scales so that it may provide multiple resolutions in frequency and time, which will mainly, used in this study to classify the faults. Because the original signal or function can be represented in terms of a wavelet expansion using coefficients in a linear combination of the wavelet functions. Wavelet techniques have been successfully applied to many signal and image processing areas, and more recently, several papers have demonstrated the potential applications of wavelets in power system analysis [5,6].

Fault classification and faulted-phase selection play a critical role in the protection for a transmission line. Identifying the faulted-phase is to satisfy single-pole tripping and autoreclosing requirements [7]. Identifying the type of fault, e.g., single-phase grounding fault, phase-to-phase fault, etc supports the digital relays in order to select different algorithm elements to deal with different fault situations. In this paper, a scheme for fault detection and identification of three phase overhead transmission lines ended with underground cables is proposed. Fault detection technique is based on mean square value of the difference between incoming and out going three phase currents of each section. Theses differences are compared against threshold setting values. Faulty phase identification is based on the analysis of three phase currents at one end of transmission line. The transient currents are processed by Discrete Wavelet Transform (DWT) multi-resolution analysis using 6-level DWT. In order to find out the faulty phase, the absolute sum of the DWT coefficients of the third level detail d3 of each phase current S(Cd3), corresponding to the frequency band 2512.5 kHz, is computed over one cycle of power frequency. S(Cd3) is given to a rule-base system to identify the fault type. Many case studies are provided to validate the proposed algorithm. II. SIMULATED RADIAL SYSTEM The The system under study is shown in Fig. 1 and it is represented as 220 kV. It consists of three sections, one is transmission line and the others are cables. The ATP/EMTP program is used in this work for modeling the system as below [8].

Fig. 1 Simulated power system

 AC. Source The source is modeled as a 220 kV sinusoidal A.C voltage source.  Overhead transmission line The line consists of a 200 km overhead transmission line. The overhead line is similar to the typical Egyptian 220 kV power line. All relevant data about these lines were offered by

452

the Ministry of Electricity and Energy of Egypt and are given in Fig. 2, and Table 1.

Table 2 Radial cable parameters

R1 R2 R3 R4 R5 c c r1- r2 c , r1 , s , r2 0 mm 20 mm 40 mm 43 mm 45 mm 1.742e-8 .m 2.84e-8 .m 2.7 1

III. FAULT DETECTION A. Algorithm design

Fig. 2 Overhead line configuration

Table 1 Overhead transmission line data

Number of conductors per phase Diameter of a single conductor Spacing between conductors in the bundle Conductor resistivity Earth resistivity Span 2 27 mm 30 cm 3.147*10-8 .m 80 .m 360 m

current(A)

 Underground cable A 220 kV underground cable of 1500 m length is used. The Bergeron model of the cable is also used. Configurations and installations of the cable line are shown in Fig.3. The physical and insulation constants of each cable phase are identical and its data is given in Table. 2. Power system simulation is carried out using the ATP/EMTP. The facilities of ATP/DRAW graphic interface and LCC (Line Cable Constants) program are used in order to implement the overhead line transmission line and cables. Bergeron model is used as a quite suitable model for more exact computer simulations
2nd insulation (r2 , r2) Sheath (s , s) 1st insulation (r2 , r2) Core (s , s)

Fig. 4 shows the phase currents measured for single-phase fault case occurring at fault distance of 50 km in the transmission line section depicted in Fig. 1. The waveforms are measured at points m1, m2, m3 and m4. Applying the current difference for each transmission section in the system shown in Fig.1 can provide a unit protection and therefore the fault event can be detected and the faulty section can be identified. According to this transmission system, Fig.5 shows the flow chart of the fault detection algorithm. Its structure is as follows: At normal conditions: 1-The 3-phase currents waveforms are measured at points m1, m2, m3 and m4 of the considered system shown in Fig.1.
5000 current(A) i1a 0 i1b i1c

-5000

0.02

0.04

0.06 Time (sec)

0.08

0.1

0.12

(a) Phase currents at point m1

5000 i2a 0 i2b i2c

-5000

0.02

0.04

0.06 Time (sec)

0.08

0.1

0.12

(b) Phase currents at point m2

2000 current(A) i3a 0 i3b i3c

-2000

0.02

0.04

0.06 Time (sec)

0.08

0.1

0.12

Ground

(c) Phase currents at point m3

2000 current(A)

1m

i4a 0 i4b i4c

-2000

0.02

0.04

0.06 Time (sec)

0.08

0.1

0.12

(d) Phase currents at point m4

0.28 m 0.28 m

Fig.3 Structure and installation of three phase single core cable

Fig. 4 Single phase- to- ground (A-G) fault at 50 km

2- Define the fault detector current settings. It is carried out as follows: a- Calculate the difference between the incoming and outgoing current signal for each section.

453

b- Calculate the mean square values of the above current difference signals x1abc, x2abc, x3abc where xn (a,b,c) are mean square values of the current difference signals of three phase for each section c- Estimate the normal current difference threshold of the nth section as: xn threa=R x max(xn(a,b,c)) where max (xn(a,b,c)) is the maximum value of the current difference in the three phases of section n. R is a selected scaling up factor. To process an unknown case: 3- The mean square values of the current difference signals xn(a,b,c), are dermineetd as 2.b above. 4- The maximum phase mean square value of the currant difference signals in each section, max((xn(a,b,c)), is compared against the current threshold xnthres 5- If max (xn(a,b,c)) is greater than xn thres then the section n is faulty. Otherwise, it is under normal condition.
Start

For section three, x3a = 261.80, x3b = 214.60 and x3c = 215.788.

The maximum mean value considering the results of section one is x1a = 1288.71, considering section two is x2a = 6519058.239 while considering section three is x3a = 261.80. By comparing x1, x2 and x3 by xn threa =10000, it is found that x1a < x1 threa, x2a > x2 threa and x3a < x3 threa. Thus, section two is the faulty one. B. Evaluation of the proposed algorithm In this section, more evaluation is carried out where the same fault scenarios, (phase-to-ground, phase-to-phase-toground, and three -phase faults) respectively, are tested at different fault point along the transmission system with different fault resistances 0, 10, 50, 100 and 1000 . The simulated faults at different points on line as follows: 1. Faults on the left cable (section one) Fig.6 illustrates the mean squares measured for the three sections considering the (A-G) fault F1 occurring at 40 ms on the left cable. The figure illustrates the values against different fault resistances. Correctly, the algorithm indicates the faulty section as well as the fault phase. The performance is that x1a is greater than the setting while the other mean squares are less than the setting. Such performance is approved up to fault resistance of 900 because the increase of fault resistance results in the decreasing of fault current so more difficulty in fault detection.
10
10

Current at the measuring nodes

Define the fault detector current setting Input 3 currents Waveforms at the selected measuring points Calculating difference signals between phase currents of each section

mean square value

Get mean square value of the difference signals between phase currents of each section x(a,b,c) .

c a1 10
8

c b1 c c1 c n1

10

Select the maximum of mean square value of the difference signal between Phase currents of each section (xmax)

10

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

(a) Measurements for the left side cable

Normal Case If xmax > difference current settings
10
5

c a2 c b2 mean square value c c2 10

4

c n2

This section is faulty

Fig.5 Flowchart for the fault detection technique

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

(b) Measurements for the overhead line

When the proposed fault detection algorithm is applied on the measured signals depicted in Fig.4 that is for single line to ground fault (AG) in T.L at 50 km. Calculating the mean square values of the difference between the incoming and outgoing phase currents in each section, it is found as follow: For section one, x1a = 1288.71, x1b = 353.45 and x1c = 352.94. For section two, x2a = 6519058.23, x2b = 8976.06 and x2c = 10241.62.

10

c a3 c b3 mean square value c c3 10

3

c n3

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

(c) Measurements for cable at right side

Fig.6 Single- phase- to- ground fault (A-G) on the right side cable

454

10

2. Faults along transmission line (section two) For (A-G) fault cases on the overhead line, Fig.7 illustrates the performance of the proposed algorithm. Correctly, the algorithm points out the faulty section 2 and the faulty phase a as x2a is greater than the setting and the other mean squares are under the setting. This performance is confirmed up to fault resistance greater than 1000 .
10 mean square value
4

c a2 c b2 mean square value c c2 10

4

c n2

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

a1
10

c b1 10
3

(b) Measurements for the overhead line

10 c a3 mean square value 10
8

c c1 c
n1

c b3 c c3 c n3

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

10

(a) Measurements for the left side cable

10 10
7

10

c a2 mean square value

6

c b2 c c2 c n2

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

10

(c) Measurements for the right side cable Fig.8 Single-phase fault (A-G) at left cable

10

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

IV. FAULT IDENTIFICATION A. Algorithm design

(b) Measurements for the overhead line

10
4

c a3 mean square value c b3 c c3 10

3

c n3

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

(c) Measurements for the right side cable. Fig.7. Single- phase-to- ground fault (A-G) on Transmission Line

3. Faults on right cable (section three) For (A-G) fault cases on the right cable, Fig.8 illustrates the performance. x3a is greater than the threshold setting declaring that the faulty phase a and the faulty section is section three.
10
4

In section 3, the fault is detected and the faulty section can be identified using the mean squares of the current differences of each transmission section. However, more study to identify the fault type is carried out by investigating the transients generated due to fault occurrence. Therefore, the DWT is used in order to extract the fault transients and then a proposed algorithm is used to provide a correct decision on the faulty phase. As aforementioned, the transient currents are extracted by 6-level (DWT) multi-resolution analysis. The choice of mother wavelet is equally important in detecting and localizing different types of fault transients. A commonly used mother wavelet suitable for protection applications is Daubechies wavelet [9], and, for the present work, the db(24) wavelet is used. The current signals are sampled at frequency of 200 kHz, i.e. 4000 sample per cycle. The zero sequence current is calculated as: (1) io= (ia+ ib+ ic)/3 The zero sequence current is also analyzed at 6-level DWT. In order to find out the faulty phase, the absolute sum of the DWT detail d3 coefficients sd3 of each phase corresponding to the frequency band 25-12.5 kHz is computed over the first post-fault one cycle period of power frequency. The faulty phase (s) is identified as follows: 1. The phase with maximum sd3 () is recognized as faulty phase or one of the faulty phases depending on the fault type. 2. The ratio R of to the sd3 of the other two phases is calculated.

c a1 c b1 mean square value 10

3

c c1 c n1

10

10

100

200

300

400 500 600 fault resistance in ohm

700

800

900

1000

(a) Measurements for the left side cable

455

3. 4.

The phase with a value of R > 0.8 is also recognized as faulty phase. Otherwise, it is a healthy phase. To identify the earth fault, the (sd3) of the zero sequence current io (sd3)0 is compared to a selected threshold value . If (sd3)0 > , then the case referrers to a ground fault. Otherwise, the case points out to phase fault.

Start

Current at the measuring nodes and calculate zero sequence current

DWT multi-resolution analysis

d3io d3ic d3ib d3ia

Absolute sum of DWT coefficient (sd3)

(sd3)o (sd3)c (sd3)b (sd3)a

The fault type is identified as fallows: a. (i) If the number of faulty phases is 3, then the fault is 3phase fault (ii) If the number of faulty phases is only one and (sd3)0 < then the fault is symmetrical 3-phase fault. (iii) If the number of faulty phases is three and (sd3)0 > then the fault is symmetrical 3-phase-to-ground fault. b. If the number of faulty phases is only one and (sd3)0 > , then the fault is single phase- to- ground fault. c. If number of faulty phases is 2, then the fault is phase-tophase fault (i) If (sd3)0> , then the fault is ( phase-tophase to ground ) fault (ii) If (sd3)0< , then the fault is (phase-to- phase) non ground fault The sampling rate can be reduced to 100 or 50 kHz, but the used coefficients will be detail d2 or d1, respectively. More reduction in the sampling frequency can be considered, however; the absolute sum detector has less gain [9]. The proposed algorithm to identify the faulty phase(s) can be represented in Flowchart as shown in Fig.9. B. Algorithm performance When the proposed fault identification algorithm is applied on the measured signals depicted in Fig.10. This figure shows the waveforms for single-phase fault (A-G) at L = 10 km. Accordingly, the transients are localized using DWT as shown in Fig.10 b in which they are during one power cycle after the fault occurrence. Calculating the absolute sum detail d3 coefficient of each phase as in Fig.10 c, it is found that: For phase (a,b,c), (sd3)a = 1321.36, (sd3)b=729.714, R2= (sd3)c=731.75. The maximum sd3 is (sd3)a= 1321.36. Then it is found that, R1=(sd3)b/(sd3)max= 0.5522 and algorithm correctly indicates a-g fault. (sd3)c/(sd3)max= 0.5538 and (sd3)o= 477.14. Accordingly, the
8000 6000 4000 2000 0
300 200 100 0 -100 -200
current(A) 1500 1000 500 0 -500 -1000 -1500 0 0.02

Maximum of [(sd3)a, (sd3)b,(sd3)c]

The phase of maximum value sd3 is faulty one

Get the ratio between absolute sum of every phase and the maximum one

Faulty phase selection

R > 0.8
yes

No

No

If number of faulty phases=1 yes 3 symmetrical fault

The corresponding phase is faulty

Non-symetrical fault

yes

L-L-G fault

If (sd3)0>
No

1- to ground fault

Double line fault(L-L) fault

Fig.9 Proposed fault type identification technique

iRa iRb iRc

0.04

0.06 Time (sec)

0.08

0.1

0.12

(a) Three phase currents waveforms

detail level 3 d3ia d3
ib

d3ic d3io

0.01

0.02

0.03

0.04

0.05 Time(s)

0.06

0.07

0.08

0.09

0.1

(b) The detail d3 coefficient of each phase

SI SIa1 SIb1 SIc1 SIo1

0.01

0.02

0.03

0.04

0.05 Time (sec)

0.06

0.07

0.08

0.09

0.1

(c) The absolute sum detail d3 coefficient Fig.10 Single-phase fault (A-G), L= 10 km, rf = 0 .

456

Similarly, Fig.11 shows the results for A-G ground fault at L= 180 km, rf = 0 . The faulty phase has the highest absolute sum when it is compared with the others. Fig.11 shows detail d3 and sd3 obtained after wavelet decomposition for singlephase-to-ground fault. It can be observed that when fault occurs, it is found that the faulty phase (a) has the highest absolute sum when it is compared with the others. Thus (sd3)a = (sd3) max, then maximum absolute sum is phase (a), and R1, R2 have small ratio where R1= (sd3) b / (sd3) max, R2 = (sd3) c / (sd3) max ,(sd3) 0 > , then this fault is a-g.
3000 2000 current(A) 1000 0 -1000 -2000 -3000 0 0.02 0.04 0.06 Time (sec) 0.08 0.1 0.12 iRa iRb iRc

The proposed fault detection algorithm is evaluated using double- phase-to- ground fault (A-B-G) in T.L at 100 km. Absolute sum of detail d3 DWT coefficients is depicted in Fig.13. It is found that: for phases (a,b,c), (sd3)a = 1515.86 , (sd3)b= 1485.09, (sd3)c=377.4 and the maximum sd3 is (sd3)a= 1321.36. R1=(sd3)b/(sd3)max = 0.97 and g fault.
SI 8000 SIa1 6000 4000 2000 0 SIb1 SIc1 SIo1

R2=

(sd3)c/(sd3)max = 0.249 and (sd3)o = 119.27, then this fault ab-

0.01

0.02

0.03

0.04

0.05 Time (sec)

0.06

0.07

0.08

0.09

0.1

(a) Three phase currents waveforms

detail level 3 300 200 100 0 -100 -200 -300 0 0.01 0.02 0.03 0.04 0.05 Time(s) 0.06 0.07 0.08 0.09 0.1 d3ia d3ib d3ic d3io

Fig.13 The absolute sum detail d3 coefficients for ground fault (A-B-G), Lf=100 km, rf = 0 .

double- phase-to

(b) The detail d3 coefficient of each phase

SI 8000 SIa1 6000 4000 2000 0 SIb1 SIc1 SIo1

Fig.14 shows absolute sum of detail d3 DWT coefficients for three- phase-to-ground fault type. In this figure, (sd3)a = (sd3) max, and R1, R2 have small ratio where R1= (sd3) b / (sd3) max, R2 = (sd3) c / (sd3) max ,(sd3) 0 < , then this fault is abcg fault. All above simulated cases ensured that the proposed technique has the ability to identify all fault types.
SI 10000 8000 6000 4000 2000 0 SIa1 SIb1 SIc1 SIo1

0.01

0.02

0.03

0.04

0.05 Time (sec)

0.06

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04

0.05 Time (sec)

0.06

0.07

0.08

0.09

0.1

(c) The absolute sum detail d3 coefficient Fig. 11 Single- phase fault (A-G), L= 180 km, rf = 0 .

Fig.14 The absolute sum detail d3 coefficients for three- phase-toground fault, Lf= 100 km, rf = 0 .

C. Response for different fault resistances

Different cases of phase b to ground fault with different fault resistance (0, 10, 50, and 1000) are considered. The results are presented in Figs. 15 to 19. The proposed method is valid in detecting high resistance ground faults on a transmission system.
SI 4000 SIa1 3000
SIa1 SIb1 SIc1 SIo1

Fig.12 shows absolute sum of detail d3 DWT coefficients for double phase fault type. In this figure, (sd3)a = (sd3)b = (sd3) max

then maximum absolute sum are phase (a,b), and R2 have

small ratio where R2 = (sd3)c / (sd3)max ,(sd3)0 < , then this fault is a-b fault type.
SI 8000 6000 4000 2000 0

SIb1 SIc1 SIo1

2000 1000 0

0.01

0.02

0.03

0.04

0.05 Time (sec)

0.06

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04

0.05 Time (sec)

0.06

0.07

0.08

0.09

0.1

Fig.12 The absolute sum detail d3 coefficients for double-phase fault in Transmission Line (A-B) at Lf=100 km, rf= 0

Fig.15 The absolute sum detail d3 coefficients for single- phase fault (B-G), Lf = 60 km, rf = 0 .

457

SI 3000 2500 2000 1500 SIa1 SIb1 SIo1 SIc1

VI. REFERENCES

[1] IEEE Working Group on Estimating the Lightning Performance of Transmission Lines, ''IEEE Working Group Report Estimating Lighting Performance of Transmission Lines II-Updates to Analytical Models'' , 1000 IEEE Trans. Power Delivery, Vol. PWRD-8, no. 3,pp. 1254-1267,1993. 500 [2] M.Kezunovic and B. perunicic,''Automated Transmission Line Fault 0 Analysis Using Synchronised Sampling At Two Ends'', IEEE Trans. Power 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Time (sec) Sys., V.11, NO.1, Feb. 1996. Fig.16 The absolute sum detail d3 coefficients for single- phase-to- [3] A.T .Johns, Z.Q.B, Y .H.Song, R.K.Aggarwal'', Spectrum Analysis of FaultInduced Transients For, The Developments of Protection Equipment'', IEEE ground fault (B-G), Lf=60 km, rf=10 SI Proc. APSCOM, Singapore, 1993. 2000 [4] Graps'', An Introduction To Wavelets'', IEEE Computer Science SIa1 Engineering 2, V.2,pp. 50-61, 1995. 1500 SIb1 SIc1 [5] S. J. Huang, C. T. Hsieh,'' High Impedance Fault Detection Utilizing A 1000 SIo1 Wavelet Transform Approach'', IEEE Transactions On Power Delivery, V. 14, NO .4, PP. 1401-1410,Oct.1999. 500 [6] D.C. Robeston, O. I. Camps, J. S. Mayer, W. B. Gish,'' Wavelet 0 Electromagnetic Power Transients'', IEEE Transactions On Power Delivery, 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Time (sec) V. 11, NO. 2, PP. 1050-1058, April 1996. Fig.17 The absolute sum detail d3 coefficients for single- phase-to- [7] N.Elkalashy, M.Lehtonen, H.Darwish, M.Izzularab and A-M.Taalab "DWTBased Investigation of Phase Currents for Detecting High Impedance Faults ground fault (B-G), Lf= 60 km, rf = 50 SI Due to Leaning Trees in Unearthed MV Networks" IEEE/PES General 1500 Meeting, Tampa, Florida, USA, June 24-28, 2007. SIa1 [8] Vanhumlen, "Alternative Transient Program", Rule book, EMTP Center, SIb1 1000 SIc1 Belgium, 1991. SIo1 [9] Raghuveer M .rao and Ajit S. Bopardikar, ''Wavelet transforms; 500 introduction to theory and applications, '' ADDISON-WESLEY, 2000.
0 0 0.01 0.02 0.03 0.04 0.05 Time (sec) 0.06 0.07 0.08 0.09 0.1

Fig.18 The absolute sum detail d3 coefficients for single- phase-toground fault(B-G), Lf=60km, rf=100
SI 500 400 300 200 100 0 0 0.01 0.02 0.03 0.04 0.05 Time (sec) 0.06 0.07 0.08 0.09 0.1 SIa1 SIb1 SIc1 SIo1

Fig.19 The absolute sum detail d3 coefficients for single phase fault (phase B), Lf=60 km, rf = 1000

V. CONCLUSION A scheme for fault detection and identification of three phase overhead transmission lines ended with underground cables is proposed. Fault detection technique is based on mean square value of the difference between incoming and out going three phase currents of each section. Theses differences are compared against threshold setting values. Faulty phase identification is based on the analysis of three phase currents at one end of transmission line. The transient currents are processed by Discrete Wavelet Transform (DWT) multiresolution analysis using 6-level DWT. In order to find out the faulty phase, the absolute sum of the DWT coefficients of the third level detail d3 of each phase current sd3, corresponding to the frequency band 25-12.5 kHz, is computed over one cycle of power frequency. sd3 is given to a rule-base system to identify the fault type. The proposed scheme has been able to detect and classify all the 11 transmission line fault types for different fault locations and wide range of fault resistance.

458