high imp fault simulation_powconf2010

1

Abstract-- A new model of High Impedance Fault (HIF) in

electrical power distribution feeders is introduced in this paper.

Proposed model is based on Emanuel arc model and contains

some features, such as nonlinearity characteristics of HIF and all

frequency components up to 12 kHz. Fast Fourier Transform

(FFT) and Principal Component Analysis (PCA) are used on real

test data to extract and reduce the features. Also Bonferroni

confidence interval is applied for comparing simulated models

with measured currents. This model contains 6 arc models like

Emanuel arc models. Presented model is simulated with

ATP/EMTP and probability switch is used for simulating random

state of HIF feature. 5 models are developed for several

amplitudes of HIF current. All models are accurate for simulation

of HIF current.

Index Terms-- Bonferroni method, High impedance fault

model, Power distribution System, Principal component analysis.

I. INTRODUCTION

igh impedance faults (HIFs), occur at primary network

level in electric power distribution systems when an

overhead conductor breaks or touches a high impedance

surface such as asphalt road, sand, cement or a tree. Because

of high impedance at the fault point, current doesn't cause an

excessive change therefore it is generally difficult for

conventional over-current relay to detect it. When this type of

fault happens, energized high voltage conductor may fall

within reach of personnel. Also, arcing often accompanies

these faults, which poses a fire hazard. Therefore, from both

public safety and operational reliability viewpoints, detection

of HIFs is critically important [1].

Arcing and nonlinear characteristics of HIFs cause a stochastic

nonlinear current, which has certain characteristics in transient

and steady state sections that make them identifiable.

Many researchers propose various detection methods. Some

detection methods are based on experimental data [2-6], and

other methods are based on simulated HIF models [7-15].

An accurate modeling method for HIF is essential for the

development of reliable detecting algorithms. The HIF

model’s data must contain the complex characteristics of HIF

such as nonlinearity, asymmetry and the low frequency of HIF

currents.

A. R. Sedighi is with the Department of Electrical & computer

Engineering, Yazd University, Yazd, IRAN (e-mail: [email protected]).

M. R. Haghifam is with Faculty of Electrical & Computer Engineering

Tarbiat Modares University, Tehran, IRAN (e-mail:

[email protected]).

Many models are introduced with respect to some HIF features

as explained in the next section.

In order to obtain a good model for HIF in this paper, the 40

experimental data were collected on a 20 kV distribution

network with sampling rate of 24.670 kHz.

The proposed model is applied based on previous experiments

and some new features that were extracted from collected data.

Some HIF models are explained in the next section, in the

third section a background of proposed method of HIF models

are introduced. HIF data collection is explained in the forth

section and the fifth section shows HIF model and simulation

method. Results are shown in the sixth section.

II. REVIEW OF HIF MODELS

HIF models can be dividing in two groups. The first group

is based on Emanuel model that was introduced in 1990, [8].

The other HIF models can be placed in the second group.

Emanuel model is based on laboratory measurements and

theoretical components. As shown in Fig. 1, the arc is modeled

using two DC sources, connected as anti- paralleled by two

diodes.

Fig. 1. The Emanuel arc model

In 1993, for consideration of nonlinearity in earth

impedance, resistance and inductance of Emanuel model were

exchanged with two nonlinear resistances [9]. Fig. 2 illustrates

that model.

Based on the arc theory, a realistic model of HIF embracing

nonlinear impedance, time-varying voltage sources and

transient analysis of control system (TACS), controlled switch,

as shown in Fig. 3, was introduced in 1998 [10].

A simplified Emanuel model was introduced in 2003. As

shown in Fig. 4 the model has two unequal resistances that

represent asymmetric fault currents [12].

In 2004, a simplified 2-diode HIF model was introduced in

[13], as shown in Fig. 5. This model consists of a nonlinear

Simulation of High Impedance Ground Fault

In Electrical Power Distribution Systems A. R. Sedighi, Member, IEEE, and M. R. Haghifam, Senior Member, IEEE

H

2010 International Conference on Power System Technology

978-1-4244-5940-7/10/$26.00©2010 IEEE

2

resistor, two diodes and two dc sources that change amplitudes

randomly every half cycle.

Fig. 2. The introduced HIF model in 1993

Fig. 3. The HIF model based on arc essences


Another HIF model was introduced in 2005, as shown in

Fig. 6. It contains two diodes and polarizing ramp voltages to

control arc ignition instants. The arc model consists of linear

resistor (representing the ground path resistance), the nonlinear

time varying resistor r(t) (representing the dynamic arc) as well

as DC and AC sources. The sources ensure asymmetry of the

arc current and voltage (DC sources) and variable arc ignition

and quenching point (AC sources) [14].

Fig. 5. The HIF model based on two diodes and two variable sources


As shown in above figures, all models are based on

Emanuel arc model and researchers have tried to complete it

and come up to a better model for HIF. Also in this paper a

new HIF model is introduced by using the Emanuel arc model.

Some other types of HIF models are introduced in [11,15,16].

III. BACKGROUND OF THE PROPOSED METHOD

In this section the Principal component analysis and

Bonferroni method are used as mathematical methods and the

proposed model are addressed.

A. Principal component analysis

Principal component analysis (PCA) is a simple linear

transformation technique that compresses high-dimensional

data with minimum loss of data information. Principal

component analysis (PCA, also called K–L transformation) is

one of the most widely used dimension-reduction techniques in

most practical cases. PCA finds the linear subspace that best

represents data without using information of class labels,

which is usually called unsupervised dimension reduction

method. In PCA a vector is first decomposed into a linear

combination of orthogonal basis functions in which the

combination coefficients are uncorrelated, and then the

dimension of the feature vector is reduced. More details are

explained in [17-18].

3

B. Bonferroni method

The Bonferroni method is a simple method that allows

many comparison statements to be made (or confidence

intervals to be constructed) while still assuring an overall

confidence coefficient is maintained.

This method applies to an ANOVA situation when the

analyst has picked out a particular set of pair wise comparisons

or contrasts or linear combinations in advance. This set is not

infinite, as in the Scheffé case, but may exceed the set of pair

wise comparisons specified in the Tukey procedure.

The Bonferroni method is valid for equal and unequal

sample sizes. We restrict ourselves to only linear combinations

or comparisons of treatment level means (pair wise

comparisons and contrasts are special cases of linear

combinations). More details are explained in [19].

IV. HIF DATA COLLECTION

In this research HIF current data was gathered from tests on

a real network [2-4]. For HIF data collection a radial 20kV

feeder in Qeshm Island, Iran, was chosen for high-impedance

fault tests. Feeder length was 19.5 km and HIF locations were

approximately 8.5km from the source end. The feeder was

energized from another 20 kV feeder through two distribution

transformers (20/0.4 kV, 100kVA) connected back-to-back.

The high- and low-voltage connections of transformers were D

and Y, respectively. The high-voltage sides are connected to

feeders and the low-voltage sides are connected together

through the low-voltage switch. Three phase voltages and

currents were recorded using Hall-effect current transformers

(CT), a resistive voltage divider (PT), power analyzer and

computer. Sampling rate of recorded data was 24.670 kHz and

total recorded time was 15 s for each test. A schematic of the

connections and the site are shown in Figs. 7-8.

Fig. 7. Schematic of instrument connection

Fig. 8. The experiment site

For a HIF test a conductor was connected to one phase of

the feeder and for each test it was dropped to the ground as

illustrate in Figs. 9 and 10. The fault studies were conducted

on seven types of surfaces (wet and dry asphalt, cement and

soil, and dry tree) at two locations, approximately 8209 and

8446m from the site. Three tests were conducted for each type

of surface at each location for a total of 42 data sets. For the

reason that two data set were not recorded correctly, 40 data

sets are used in this work. Faulted phase current signal on

various conditions are shown in Fig. 11.

Fig. 9. Connection of a conductor to one phase.

Fig. 10. Dropping of conductor to earth.

V. HIF MODEL AND SIMULATION METHOD

HIFs have complex characteristics and features. Some of

them are listed in the following. These features are based on

research and considerations that were presented in previous

papers [11,20].

- Nonlinearity phenomena of ground impedance affect

HIF current curve. Fault current grows to its

maximum value in about 50~60 ms. The most

noticeable fact is that the initial current is only about

60% of the final value and it grows to the final value

in about three to four cycles. This reducing in initial

current is due to a smaller effective initial contact

between the conductor and the ground. When the

contact area is small, as via a small arc, the density of

the current at the arc/ground interface, and hence the

voltage gradients will be large. This will result in

localized arcing and ionization. The arc will then

penetrate the ground between the earth particles and

thus enlarging the effective contact with the ground.

That is, when a conductor arcs to the ground surface,

the arc will not stay terminated at exactly the surface

4

of the ground, but will cause ionization within the

earth and so penetrates the ground, thus enlarging the

effective area of the equivalent electrode.

- The earth resistance is another character of HIF

nonlinearity of the current, which is affected by the

quality of the earth, genus and its humidity.

- The voltage and current of HIF have the same phase,

so resistance is inherent in HIF.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

Fig. 11. HIF current curves on different surfaces. (a) Dry asphalt. (b) Wet

asphalt. (c) Dry cement. (d) Wet cement. (e) Dry soil. (f) Wet soil. (g) Dry

wood.

- The voltage-current characteristic curve of HIF is

nonlinear.

- Fault current has different waveforms for positive and

negative half cycle.

This paper presents a modeling method for representing the

above mentioned characteristic of HIFs. Also this model

supports all frequency components of the recorded HIF current

from various experimental data on a distribution system that

was explained in the previous section. The introduced model

simulates the first eight cycles of HIF current.

In many previous researches, the models were based on

Emanuel arc model and researchers changed the model

parameters and came up with HIF current. This paper pays a

very special attention to HIF. When a broken conductor fell on

earth, some arcs occurred as illustrated in Fig. 12. These arcs

are the bases for the proposed model.

Fig. 12. Arcs in HIF test on cement surface.

Based on Emanuel model the proposed model used several

arc models in parallel where their combination produced HIF

current. Fig. 13 shows this new HIF model.

Fig. 13. New HIF model based on several Emanuel arc model

Random state of HIF is shown using STATISTIC switch

in EMTP. The state of on and off one arc is shown in Fig.13 in

the sixth arc model. Uniform state with 0.01s deviation, is used

for distribution function of statistic switches. Arc parameters

in the introduced model are determined based on recorded

current and voltage data. As shown in Fig. 14 voltage and

current have the same phase after fault occurred and earth

resistance can be derived from it.

5

Fig. 14. HIF current and voltage

Vp and Vn , DC voltage source, can be derived from v-i

curve of HIF as introduced in Fig. 15, based on explanation in

[8].

Fig. (15): v-i curve of HIF

By using switch time regulation and other parameters the

current amplitude in one simulation can be controlled from

60% of the final value to final value after 50~60 ms.

The simulated current and recorded current were compared

as explained in the following:

1. The first eight cycle of recorded current with sampling

rate 224.67 kHz are used.

2. Using FFT, all frequency components (2048) are

extracted.

3. Based on main frequency component all amplitudes are

normalized.

4. The number of extracted features at this state are 4096

(amplitude and phase).

5. The features are reduced to 38 using PCA, saving

98.3% of the energy.

6. For mean of each 38 features, Bonferroni intervals were

calculated with α= 0.0001.

7. For simulated current, steps 1~5 are repeated.

8. If all 38 simulated current features in step 7, fall with in

the Bonferroni interval in step 6, then the simulated

current is like HIF current. Otherwise some parameters of

HIF model will be modified and go to step 7.

With respect to above program five model that simulate

HIF current with different amplitude are introduced.

VI. RESULTS

With respect to previous section explanation, 5 separated

models for HIF current are introduced with several amplitudes

(9~90 A) for various surfaces. In each HIF model six arc

models are used. All parameters for each arc model are

obtained based on recorded HIF current. STATISTIC

switches with uniform distribution are used for access to

random state of HIF current.

Several sample of five simulated HIF current are illustrated

in Fig. 16

(1)

(2)

(3)

(4)

(5)

Fig. (16): simulated HIF currents

Fig. 17 shows several result for model #1.

0.00 0.05 0.10 0.15 0.20 0.25[s]

-50.0

-37.5

-25.0

-12.5

0.0

12.5

25.0

37.5

50.0

[A]

Simulated HIF Current

0.00 0.05 0.10 0.15 0.20 0.25[s]

-60

-34

-8

18

44

70

[A]


0.00 0.05 0.10 0.15 0.20 0.25[s]

-90

-60

-30

0

30

60

90

[A]

Simulated HIF current

0.00 0.05 0.10 0.15 0.20 0.25[s]

-8

-5

-2

1

4

7

10

[A]


0.00 0.05 0.10 0.15 0.20 0.25[s]

-30

-20

-10

0

10

20

30

[A]


6

(a)

(b)

(c)

(d)

(e)

Fig. 17. Several result for model #1.

Fig.18 shows voltage and current of the first state in a

curve, as shown voltage and current have the same phase after

fault occurred.

Fig. 18. Simulated HIF voltage and current of the first state

v-i curve of simulated HIF for the first state is shown in

Fig. 19

All parameters for 5 presented models and their arcs are

shown in table 1.

Fig. 19. Simulated HIF v-i curve of the first state

TABLE 1: ARCS PARAMETERS OF HIFS MODELS

As explained, each arc is connect to the model with a

statistical switch that has mean time as show in table 1 with

uniform distributed function and 0.01 second duration.

VII. CONCLUSION

In this paper a novel HIF model is presented. This new

model is based on Emanuel arc model. Several arc models are

used together until simulated HIF currents are similar to real

recorded HIF currents. FFT is used for feature extraction and

PCA for their dimension reduction. Simulated and recorded

currents are compared with Bonferroni interval. EMTP is used

for simulation of HIF model and statistical switches are used

for producing random HIF current.

Vn Vp Rn Rp t(ON/O

ff)

paramet

er

4500 4000 1050 1000 0.06 Arc1

sta

te

1

8100 8000 3000 2900 0.07/

0.11 Arc2

7600 7500 3550 3500 0.09 Arc3

10500 10000 3750 3700 0.1 Arc4

1300 1000 4010 4000 0.08 Arc5

3500 3000 2850 2800 0.13 Arc6

1900 1800 905 900 0.06 Arc1

sta

te

2

3500 3000 2850 2800 0.08/

0.13 Arc2

4500 4000 2550 2500 0.12 Arc3

11000 10000 2150 2100 0.14 Arc4

2500 2000 805 800 0.1 Arc5

3550 3050 2900 2850 0.15 Arc6

1000 900 505 500 0.06/

0.18 Arc1

sta

te

3

2700 2500 2050 2000 0.08 Arc2

1050 1900 2550 2050 0.12 Arc3

11500 11000 1900 1800 0.14 Arc4

1050 1000 1050 1000 0.1 Arc5

1900 1800 905 900 0.19 Arc6

900 750 300 280 0.06 Arc1

sta

te

4 2800 1800 2500 2000 0.1 Arc2

4000 3000 2700 2500 0.12 Arc3

11500 11300 2750 2700 0.14/

0.16 Arc4

1050 950 805 800 0.08 Arc5

11450 11250 2700 2650 0.18 Arc6

10000 9000 1505 1500 0.06 Arc1

sta

te

5 9000 8000 9600 9000 0.08 Arc2

11050 11000 7000 6500 0.12 Arc3

12500 12000 6500 5800 0.14/

0.16 Arc4

2050 2000 11100 10000 0.1 Arc5

11500 11000 6200 5300 0.16 Arc6

0.00 0.05 0.10 0.15 0.20 0.25[s] -30

-20

-10

0

10

20

30

[A]

Simulated HIF Current *

0.00 0.05 0.10 0.15 0.20 0.25[s] -30

-20

-10

0

10

20

30

[A]


0.00 0.05 0.10 0.15 0.20 0.25[s] -30

-20

-10

0

10

20

30

[A]


0.00 0.05 0.10 0.15 0.20 0.25[s]

-30

-20

-10

0

10

20

30

[A]


0.00 0.05 0.10 0.15 0.20 0.25[s] -30

-20

-10

0

10

20

30

[A]


0.00 0.05 0.10 0.15 0.20 0.25[s]

-30

-20

-10

0

10

20

30

[A]

-20

-15

-10

-5

0

5

10

15

20

[kV]

v-i curve of Simulated HIF

7

VIII. REFERENCES

[1] M. Aucoin, B.D. Russell, C.L. Benner, “High impedance fault

detection for industrial power systems “ Industry Applications

Society Annual Meeting, 1989., Conference Record of the 1989

IEEE, vol.2, pp.1788 –1792,1989.

[2] A.-R. Sedighi, M.-R.Haghifam, O.P.Malik, M.-H. Ghassemian,

“ High Impedance Fault Detection Based on Wavelet Transform

and Statistical Pattern Recognition” IEEE Transactions on

Power Delivery, Volume 20, pp.2414 – 2421, Oct. 2005.

[3] A.-R. Sedighi, M.-R. Haghifam, O.P. Malik, ”Soft Computing

Applications in High Impedance Fault Detection in Distribution

Systems” Electric Power Systems Research, Volume 76,

pp.136-144, September 2005.

[4] M.-R. Haghifam, A.-R. Sedighi, O.P. Malik, “Development of a Fuzzy

Inference System Based on Genetic Algorithm for High Impedance

Fault Detection” Generation, Transmission and Distribution, IEE

Proceedings-Volume 153, pp.359 – 367, May 2006.

[5] Shyh-Jier Huang; Cheng-Tao Hsieh “High-impedance fault

detection utilizing a Morlet wavelet transform approach” Power

Delivery, IEEE Transactions on, Vol. 14, pp. 1401 –1410, Oct.

1999,.

[6] A. Lazkano, J. Ruiz, E. Aramendi,L.A. Leturiondo, “A new

approach to high impedance fault detection using wavelet packet

analysis” Harmonics and Quality of Power, 2000. Proceedings.

Ninth International Conference on ,Vol.3 pp. 1005 –1010,

2000.

[7] M. Aucoin, “ Status of high impedance fault detection” IEEE T-

PAS-104, No.3, pp.: 638-643,1985.

[8] A.E. Emanuel, D. Cyganski, J.A. Orr, S. Shiller, E.M.

Gulachenski, “High impedance fault arcing on sandy soil in 15

kV distribution feeders: contributions to the evaluation of the

low frequency spectrum” Power Delivery, IEEE Transactions

on , Vol.5, pp. 676 –686, April 1990.

[9] A.M. Sharaf, L.A. Snider, K. Debnath, “A neural network based

back error propagation relay algorithm for distribution system

high impedance fault detection” Advances in Power System

Control, Operation and Management, 1993. APSCOM-93., 2nd

International Conference on , pp. 613 –620,1993.

[10] David Chan Tat Wai; Xia Yibin “A novel technique for high

impedance fault identification” Power Delivery, IEEE

Transactions on, Vol.13 , pp. 738 –744, July 1998.

[11] S.R. Nam, J.K. Park, Y.C. Kang, T.H. Kim,”A modeling method

of a high impedance fault in a distribution system using two

series time-varying resistances in EMTP” Power Engineering

Society Summer Meeting, 2001. IEEE, Vol.2, pp.1175 – 1180,

July 2001.

[12] T.M. Lai, L.A. Snider, E. Lo. “Wavelet Transform Based Relay

Algorithm for the Detection of Stochastic High Impedance

Faults” International Conference on Power System Transient, in

New Orland, pp. 1-6,IPTS 2003.

[13] Y. Sheng, S.M. Rovnyak, “Decision Tree-Based Methodology

for High Impedance Fault Detection” Power Delivery, IEEE

Transactions on, Vol.19, Issue: 2, pp.533 – 536, April 2004.

[14] M. MICHALIK, W. REBIZANT, M. LUKOWICZ, S.-J. LEE,

KANG: "Wavelet Transform Approach to High Impedance Fault

Detection in MV Networks", Proceedings of the 2005 IEEE

PowerTech Conference, St. Petersburg, Russia, CD-ROM, paper

73, June 2005.

[15] Marek Michalik; et al, “ High-Impedance Fault Detction in

Distribution Networks With Use of Wavelet-Based algorithm”

IEEE Transactions on Power Delivery, vol. 21, No. 4, pp. 1793-

1802, Oct. 2006.

[16] T. Cui and Xinzhou Dong and Zhiqian Bo and S. Richards, “

Integrated scheme for high impedance fault detection in MV

distribution system” Transmission and Distribution Conference

and Exposition: Latin America, 2008 IEEE/PES,13-15 pp.1 – 6,

2008. [17] J.T. Tou, R.C. Gonzalez, Pattern Recognition Principles , 4th

Edition, Massachusetts, Addision-Wesley Publishing Company,

1981.

[18] M. Nadler, E.P. Smith, Pattern Recognition Engineering ,

JOHN WILEY & SONS, INC., 1992.

[19] NIST/SEMATECH e-Handbook of Statistical Methods,

http://www.itl.nist.gov/div898/handbook/, date.

[20] D. I. Jeerings, J. R. Linders, "Ground Resistance--Revisited,"

IEEE Transactions on Power Delivery, vol. PWRD-4, No. 2,

pp. 949-956, Apr. 1989.

Ali-Reza Sedighi (M’09) was born in Anarak, Iran, on

September 15, 1968. He received the B.S., degree in

electrical engineering from Isfahan University of

Technology, Isfahan, Iran in 1990 and M.Sc., degree

and Ph.D. in electrical engineering from Tarbiat

Modarres University, Tehran, Iran, in 1994, 2004

respectively all in power engineering. Currently, he is

an Associate Professor in the Power System

Engineering Group at Department of Electrical &

computer Engineering, Yazd University, Yazd, IRAN.

His main research interests are Electric Distribution Systems, Signal

Processing in Electrical Power Systems and Power System Intelligent Control.

Mahmood-Reza Haghifam (M’95–SM’06) was

born in Iran in 1967. He received the BS, MSc and

PhD degrees in electrical engineering in 1989, 1992

and 1995. He joined Tarbiat Modares University as

assistant Prof. in 1965. He is now a Full Professor in

Power Systems at the Tarbiat Modarres University

(TMU), Tehran, Iran He is a Senior Member of the

IEEE (and IEEE Iran Section Industrial relationship officer). Also he is a

research Fellow of Alexander Von Humboldt in Germany. He has been

awarded by DAAD and AvH in 2001, 2006 and 2009 for research stays in

German universities. He was as visiting Prof. in university of Calgary,

Canada in 2003. His main research interests are Power System Restructuring,

Power System Reliability, and Electric Distribution System.

high imp fault simulation_powconf2010

Documents