G

JD

a

ARRAA

KGGJ

1

tsuucttonpfteso

matcgi

0d

Electric Power Systems Research 80 (2010) 1074–1081

Contents lists available at ScienceDirect

Electric Power Systems Research

journa l homepage: www.e lsev ier .com/ locate /epsr

round potential rise of multi-grounded neutral and shield wires in joint systems

anak Acharya1, Wilsun Xu ∗,2

epartment of Electrical and Computer Engineering, 2nd Floor ECERF, University of Alberta, Edmonton, Canada T6G 2V4

r t i c l e i n f o

rticle history:eceived 2 November 2009

a b s t r a c t

Power line faults create the ground potential rise (GPR) on both the neutral and shield conductors whenthe transmission lines (TL) and distribution lines (DL) are built on the same structures. The durations

eceived in revised form 20 January 2010ccepted 22 January 2010vailable online 20 February 2010

eywords:round potential riserounding

and magnitudes of resulting GPRs are unique for DL faults and TL faults because the corresponding faultcurrents are significantly different in terms of their magnitudes and durations. This paper analyzes andcompares the safety impacts of TL faults and DL faults in the joint structures. Approximate formulas areestablished to describe the GPR characteristics. Computer simulation results are provided to illustratethe effects of different parameters on GPRs in various configurations.

© 2010 Elsevier B.V. All rights reserved.

oint system

. Introduction

The overhead distribution lines (DL) are often built under theransmission lines (TL) on the same towers. The safety impact ofuch configurations under short-circuit conditions are difficult tonderstand because the DL-created GPR and TL-created GPR havenique characteristics. The shield wire of TL establishes the directontact with the conductive towers which serve as grounding ofhe shield wire. On the other hand, the neutral wire on the sameower is provided with a separate dedicated grounding assembly,r is bonded with the shield wire. The conductors used for theeutral wire and shield wire are not the same. Generally steel isreferred for the shield wire and Aluminum Conductor Steel Rein-orced (ACSR) for the neutral wire. Also the physical positions ofhese conductors on the same structure lead to varying degree oflectromagnetic coupling with phase conductors under fault. Con-equently, the resulting GPRs are affected even for the same amountf fault currents in TL and DL.

In the past, a lot of studies have been done for single circuitulti-grounded configurations and a great deal of literature is

vailable [1–10], but limited work has studied the composite sys-

ems comprised of multiple multi-grounded conductors. Mostlyomputer-based methods were preferred for the studies of multi-rounded systems. Major shortcomings of such methods includenability to provide intuitive understanding on the interaction

∗ Corresponding author. Tel.: +1 780 492 5965; fax: +1 780 492 1811.E-mail address: wxu@ualberta.ca (W. Xu).

1 Student Member, IEEE.2 Fellow, IEEE.

378-7796/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.epsr.2010.01.014

and effects of various factors. Alternatively, analytical methodscan be developed to compensate such shortcomings. Computermodels were used to estimate the GPR of the multi-grounded neu-tral (MGN) in [2–4] and power flow studies were performed in[5–7]. Analytical approaches for the GPR analysis were proposedin [8–10]. The GPR assessment of multi-grounded communicationcable bonded with power line’s neutral wire was performed in [9].The principles of [9] can be applied to the joint transmission and dis-tribution system with multi-grounded conductors. In our previouswork, analytical methods were proposed to provide understand-ings of the characteristics of MGN lines [11]. The work done in thispaper complements the previous studies as it is extended to thejoint T&D systems.

It is generally agreed that the TL fault currents are larger, but arecleared faster than the DL faults. As a result, field engineers assumethat the TL caused GPRs are less severe than the DL caused GPRs.As will be shown later, this assumption is wrong. Also effectivenessof bonding the neutral wire with shield wire for lowering the GPRlevel is not fully understood yet. This paper reveals a number offactors that play a crucial role in determining the GPR.

The main objective of this paper is to address the above-mentioned concerns by using the analytical formulas and computersimulation. An EMPT-based computer tool was used for the sim-ulation. Since the power industry has accepted the associatedtechniques and models, they are not described here. Detailed mod-els can be found in [12]. Sensitivity studies are performed to

illustrate the effect of a number of parameters on the GPR. Thereminder of the paper is organized as follows. Section 2 presentsthe problem and the system under investigation. Mechanism of GPRgeneration is presented in Section 3. Results are shown in Section4 and the conclusions in Section 5.

J. Acharya, W. Xu / Electric Power Systems Research 80 (2010) 1074–1081 1075

Fig. 1. (a) Physical layout of TL and DL conductors, (b) Schematic diagram.

Table 1Conductor positions on the tower.

Conductor Horizontalposition (m)

Verticalheight (m)

Mid-spanheight (m)

T-Line Ph#A −2.15 16.01 15.06T-Line Ph#B 2.15 16.01 15.06T-Line Ph#C −2.15 12.91 11.96D-Line Ph#A −0.58 9.97 8.61D-Line Ph#B 0.64 9.97 8.61

2

tddgTtzlp

C7

TL

Table 3Base case data and their variation range.

Parameter Typical value Base case and sensitivity

Groundingresistance, RG

5–25 �Base case: 15 �Sensitivity: 7–25 �

Substationgrounding res.

– TL-sub: 0.15 �DL-sub: 0.15 �

Grounding interval Neutral: 40–75 m Base case: 75 mShield: 60–100 m Sensitivity: 75–600 m.

Parallel exposure 200–3000 mBase case: 4 km

D-Line Ph#C 1.85 9.97 8.61D-Line Neutral −0.25 7.89 7.08T-Line Shield 0.15 19.41 19.17

. Study system

Fig. 1 shows a general layout of the study system where theransmission conductors are positioned above the distribution con-uctors on the same tower. The TL and DL run in parallel for a certainistance only. The DL’s neutral wire is insulated from the tower androunded with dedicated ground rods (not shown in figure). TheL’s shield wire, however, is not insulated from tower’s body. Sohe tower itself serves as a ground rod. Table 1 provides the hori-ontal and vertical positions of the conductors, Table 2 shows theine impedances calculated using the EMTP models [12], and Table 3

rovides the system data for base case and sensitivity studies.

The system information was provided by the electrical utility inanada. Based on the available data, the line length was chosen to be.5 km so that the effect of varying parallel length can be examined.

able 2ine impedance data (�/km).

Mutual impedances betweenD-line phase wire and neutral wire zDN = 0.0583 + j0.4734D-line phase wire and shield wire zDS = 0.0576 + j0.3409T-line phase wire and neutral wire zTN = 0.0579 + j0.3567T-line phase wire and shield wire zTS = 0.0573 + j0.4030Neutral wire and shield wire zNS = 0.0577 + j0.3292

The self-impedances ofNeutral wire zNN = 0.3966 + j0.9119Shield wire zSS = 3.5638 + j0.9518

Sensitivity: 1–5 km

Line length – ∼7.5 kmShield wire 5/16′′ steel Rdc = 3.5067 �/km

The following six basic configurations are identified and studied sothat comparisons can be made in terms of their safety benefits:

(1) D-line with neutral wire (MGN)(2) T-line with shield wire (MGS)(3) T&D lines with shield, but without neutral wire(4) T&D lines with neutral, but without shield wire(5) T&D lines with neutral and shield isolated(6) T&D lines with neutral and shield bonded

The GPRs are presented in the form of volts per kA of fault cur-rent. This offers two advantages. First, it establishes the basis forcomparison of different configurations irrespective of the fault cur-rent magnitudes. Second, the numbers can be indicative to any faultcurrents (as they are uncertain). Caution should be exercised whilecomparing TL caused GPR/kA and DL caused GPR/kA values becausethe fault currents of TL and DL can be quite different. In such cases,the actual magnitudes of GPR should be considered. The acronymsNGPR and SGPR are used to denote the GPRs developed in neutralwire and in shield wire, respectively.

3. Mechanism of GPR generation

Ref. [11] presents the mechanism of GPR generation, but forthe line-to-ground fault only. The basic principles are describedin this section. The faults involving a single ground conductor willbe investigated first and then more complex schemes will be dealtwith. A generic term ‘ground conductor’ refers to either a neutralwire or a shield wire. The GPR can develop in the ground conduc-tor under two distinct conditions: (1) unfaulted ground conductor– when the fault does not involve this conductor, and (2) faultedground conductor – when the fault involves this conductor.

3.1. GPR of the unfaulted ground conductor

An unfaulted ground conductor experiences the GPR when thenearby phase conductor is at fault. This occurs when phase conduc-tor falls on the ground without making any contact with the groundconductor. The fault current flowing in the phase conductor inducesa voltage on the multi-grounded conductor (Fig. 2a). The inducedvoltage (EG) is distributed along the fault-exposure length and canbe determined as

EG = egl = zmutuall · IF (1)

where zmutual is the mutual impedance between phase and ground

conductors and l is the length of the exposure with fault current.The voltage (eg) of one segment can be modelled as an equivalentcurrent source connected between the two grounding points basedon the principle of Thevenin to Norton circuit conversion (Fig. 2b).Note that the downstream portion of the ground wire does not

1076 J. Acharya, W. Xu / Electric Power Systems Research 80 (2010) 1074–1081

Fasfi

eSo

I

w

atiato

G

G

wenta

a

Z

wTot

(also grounded). A fault (line-to-ground or line-to-wire) occurs atthe location of node 2. The fault current will induce currents inthe ground wire and in the parallel wire. Consider one segmentof the ground wire to explain the GPR developed in it (Fig. 4b). In

ig. 2. Illustration of GPR mechanism in the multi-grounded conductor: (a) fault inparallel multi-grounded conductor, (b) conversion of voltage source into current

ource, (c) equivalent model with current injection into the ground wire, and (d)nal equivalent circuit of ground wire.

xpose to the fault current, so there is no induced voltage or current.ince current IG is independent of the length of grounding span, allther segments have the same current.

G = egl

zggl= zmutual

zggIF (2)

here zgg is the self-impedance of the ground conductor.Fig. 2b can be further transformed into that of Fig. 2c without

ffecting the nodal voltages and segment currents. Assuming thathe distance between node 1 and node 2 is large enough so that thenfluence of current source (IG) at one node (e.g. node 1) does notffect the voltage of the other current-injected node (e.g. node 2),he neutral network can be modelled as shown in Fig. 2d. The GPRsf the node 1 and node 2 are given by

PR1 = IG(Zeq-1//RG) ≈ IGZeq-1 (3)

PR2 = −IG(Zeq-2u//Zeq-2d//RG) ≈ −IG(Zeq-2u//Zeq-2d) (4)

here Zeq-1 is the equivalent impedance, Zeq-2u and Zeq-2d are thequivalent impedances seen upstream and downstream from theode 2, respectively. The minus sign is placed in (4) to signify thathe GPR1 and GPR2 are of opposite polarity due to directions ofssociated currents.

The equivalent impedance of the multi-grounded ladder [11] ispproximated as

eq =√

s · zgg · RG (5)

here s is the grounding interval and RG is the grounding resistance.he impedances Zeq-1, Zeq-2u and Zeq-2d are equal due to symmetryf the ladder. The ratio GPR1 to GPR2 is approximately 2. Therefore,he maximum GPR is located at node 1.

Fig. 3. The injection of current sources during the line-to-wire fault: (a) line-to-wire(or contact fault), (b) equivalent circuit at node 1, and (c) equivalent circuit at node2.

3.2. GPR of the faulted ground conductor

The representative cases of the faulted ground conductor are:insulation failure of phase conductor causing short circuit withshied wire via tower, or shield wire falling on the phase wire, orphase conductor falling on neutral wire, etc. In this category, theground conductor contacts the phase conductor. These faults arecategorised as the contact faults in this paper and are differentfrom ground faults. Fig. 3 illustrates the case of contact fault. Themechanism of GPR generation is similar to that of unfaulted groundconductor case, but the ground conductor will carry a portion offault current entering from the fault point in addition to the currentproduced by induction (Fig. 3c). Thus the GPR at the fault locationwill be much different from that of line-to-ground fault. The GPRof the islanded end (node 1) will be the same.

The GPR at fault location (node 2) is given by

GPR2 = (IF − IG)(Zeq-2u//Zeq-2d) (6)

The current involved in (4) is IG only, but it is modified in (6) as(IF − IG) due to short circuit of phase and ground conductors.

3.3. Effect of the other parallel ground conductor

Fig. 4 shows a phase wire, a ground wire and a parallel wire

Fig. 4. The effect of parallel ground conductor: (a) a faulted phase conductor and twoparallel ground conductors, and (b) induced voltages and their equivalent currents.

Systems Research 80 (2010) 1074–1081 1077

apwb2

G

w

I

wcomdf

G

s

s

3

rs

I

wmwiIdilr

3

rn

I

wtnsttb

3

foH(dl

Fig. 5. The effect of the sources on both ends of T-line: (a) T-line with two sourcesand parallel D-line, (b) current sources in the MGS during the T-fault, and (c) currentsources in the MGN during the T-fault.

lent impedance of the bonded network and the injected currents atspecific nodes.

J. Acharya, W. Xu / Electric Power

ddition to the fault-current-induced voltage (eg), the current ofarallel conductor will induce another voltage (ep) in the groundire, but with opposite polarity. As a result, the GPR at node 2 will

e affected. Considering the line-to-ground fault, the GPR at nodeis given as

PR2 = (IG − IGp)(Zeq-2u//Zeq-2d) (7)

here

Gp = zgp

zggIP and IP = zmutual

zppIF (8)

here IGp is the current induced in the ground conductor due tourrent of another parallel conductor (IP) and zpp is the impedancef the parallel conductor. Comparing (7) and (4), the first term isodified by IGp. If the phase conductor contacts the ground con-

uctor directly or indirectly during the fault, the resulting GPR atault location (node 2) will be

PR2 = [IF − (IG − IGp)](Zeq-2u//Zeq-2d) (9)

Again, comparing (9) with (6), the current IG is modified byubtracting IGp.

The above principle can be applied to examine the effect TL’shield wire on the GPR of DL’s neutral wire and vice versa.

.3.1. Effect of shield wire on GPR of the neutral wireIn this case, IG represents the current induced by the fault cur-

ent on neutral wire and IGp represents the current induced by thehied wire’s current. Then

Gp = zNS

zNN× zDS

zSSIFD = 0.031IFD (10)

here zNS is the neutral-to-shield mutual impedance, zDS is theutual impedance between the distribution phase wire and shieldire and IFD is the fault current on distribution line. The current

nduced by the shield current (IGp) is relatively small compared toG (=0.48IFD), i.e. 6.5% of IG. Therefore the presence of the shield wireoes not significantly affect the GPR of the neutral wire. However,

t should be emphasized that if there were faults on transmissionine, the fault current will be significantly higher than IFD and theesulting neutral GPR will increase accordingly.

.3.2. Effect of neutral wire on GPR of the shield wireIn this case, IG represents the current induced by the fault cur-

ent on shield wire and IGp represents the current induced by theeutral wire’s current. Then

Gp = zNS

zSS× zTN

zNNIFT = 0.033IFT (11)

here zNS is the neutral-to-shield mutual impedance and IFT ishe fault current on transmission line. The current induced by theeutral current (IGp) is 30% of IG (where IG = 0.11IFT), which is con-iderable. Therefore the presence of the neutral wire actually helpso lower the GPR of the shield wire. If the faults occur on distribu-ion line, the GPR would not be higher than those originally causedy the transmission faults.

.4. Transmission line supplied from both ends

Fig. 5 shows the double circuit T&D lines where T-line is suppliedrom two opposite ends. During a D-line fault, there is no effect

n NGPR whether T-line is supplied from one end or both ends.owever, for the T-line faults, the presence of downstream supply

right source) causes injection of additional current sources in theownstream section of the neutral wire corresponding to the fault

ocation. The procedure describe earlier is applicable to this as well.

Fig. 6. Bonding of neutral and shield wire in the parallel section.

After the injected currents and their locations are identified, themathematical procedure described earlier can be directly appliedwith appropriate values. The results are provided in Section 4.

3.5. Bonding of D-line neutral and T-line shield

The bonding of neutral and shield wires essentially combinesthe grounded node of neutral and grounded node of shield intoa single node as shown in Fig. 6. As a result, self-impedances ofthese wires will be in parallel, so are their grounding resistances.The currents associated with the shield wire and those associatedwith neutral wire will combine in the bonded section. The GPR isfunction of equivalent impedance and amount of injected current.Therefore, the effect of bonding can be examined through equiva-

Fig. 7. Bonding schemes, (a) separately grounded, (b) grounded together.

1078 J. Acharya, W. Xu / Electric Power System

3

Fignt

F

Z

a

Z

aMooMiahtiaid

3

sttDrcat

•••

aSn

I

Fig. 8. The currents injected in the bonded neutral and shield network.

.5.1. Effect on equivalent impedanceFig. 7 shows two schemes of bonding for a grounding span. In

ig. 7a, the neutral wire and shield wire having their own ground-ng assembly are bonded at each grounded nodes, so there are 2rounding resistances RGN and RGS at each location. In Fig. 7b, theeutral wire and shield wire have the common grounding resis-ance RG.

The equivalent impedance of the bonded circuits for scheme ofig. 7a is

eq-2g ≈√

s · (zNN//zSS) × (RGN//RGS) (12)

nd for scheme of Fig. 7b is

eq-1g ≈√

s · (zNN//zSS) × RG (13)

The equivalent impedances were computed using (12) and (13)nd were compared with the equivalent impedances of MGN andGS individually. The equivalent impedance given by (12) was 33%

f the MGS alone and 65% of the MGN alone (given by (1)). On thether hand, the equivalent impedance given by (13) was 47% ofGS alone and 90% of MGN alone. It implies that larger reduction

n equivalent impedances can be achieved for MGS than for MGNs a result of bonding. This is true because the neutral conductoras smaller impedance and adding a more resistive shield conduc-or in parallel will have relatively smaller impact on the equivalentmpedance. However, the neutral wire helps to reduce the equiv-lent impedance from the shield wire’s perspective. As the GPRs directly proportional to the equivalent impedance, the GPR willecrease compared to before bonding.

.5.2. Effect on current injectionConsider the circuit of Fig. 5 again. For simplicity, ignore the

ource on the right side of the transmission line. Therefore, onlyhe left hand side source will contribute to the fault current andhe induced currents are on the left side of the fault location only.ue to the bonding, the Fig. 5b and c can be combined together,

esulting in a circuit shown in Fig. 8. So the subscripts L and R in theurrents referencing the left source and right source, respectively,re dropped. Bonding of the neutral wire and shield wire also affectshe amount of current due to following currents:

The fault current at the fault locationThe induced neutral current and the induced shield currentThe current induced in the shield wire by neutral current and thecurrent induced in the shield wire by neutral current

From Fig. 8, it can be seen that the parallel sections of neutralnd shield wires (N2–N4 or S2–S4) are combined. Nodes N2 and

2 become a single node (N2S2) and nodes N4 and S4 into a singleode (N4S4) and are of great interest.

For node N2S2, the current sources are

N2S2 = INF − ISN − INS

s Research 80 (2010) 1074–1081

For node N4S4, the current sources are

IN4S4 = ISN + INS − INF − ISF + (0 or IF)

The last term will be zero if the fault is line-to-ground and IF ifthe fault is line-to-wire. Thus the GPR of the nodes that are withinthe parallel section will be affected. The current injected in the nodeoutside the parallel zone is not affected, (e.g. Node S1). The impactof modified current injection in these will affect the GPR of theother nodes as well. It is very difficult to establish simple, but accu-rate, formula for this situation. More accurate results are shown inSection 4, here the focus is to show the peak GPR are affected.

3.6. Comparison of TL-fault and DL-fault caused GPRs

The magnitudes of the GPR caused by TL faults are higher thanthose caused by the DL faults because the fault current of DL is gen-erally much lower that of TL. The safety impact of GPR on human isdetermined by the magnitude and duration of the GPR. According tothe theory of electrocution established by Dalziel [13], the thresh-old current leading to electrocution is inversely proportional to thesquare root of the current duration. Further considering the fact thetouch and step voltages are in proportion to the GPR, we can thuscompare the safety impact associated with the GPR created by theTL faults and DL faults. A TL-fault poses higher risk if

GPRT-fault

GPRD-fault>

√tD

tT= � (14)

where tD and tT are the fault clearing times of the DL and TL, respec-tively, and � is defined as the breaker trip factor (BTF). The breakersused in distribution systems are much slower than the breakersused in transmission systems. For example, DL and TL have 30-cyclebreakers and 5-cycle breakers, respectively. Then

� =√

305

= 2.45

Thus the TL-faults will be more risky if GPRT-fault is 2.45 timesgreater than the GPRD-fault. Using the equations derived in Section3 and the system data given in Section 2, the highest GPR producedin the neutral wire by the faults on D-line and T-line are:

NGPRD-fault = 0.50IFD (line to ground fault, islanded end)NGPRT-fault = 0.30IFT (line to shield fault, fault location)

where the IFD and IFT are the fault currents of D-line and T-line,respectively. Then

NGPRT-fault

NGPRD-fault= 0.3IFT

0.5IFD> 2.45 or

IFT

IFD> 4 (15)

i.e. the TL-fault poses a higher risk if IFT > 4IFD. The highest shieldGPR produced by the D-line and T-line faults are:

SGPRD-fault = 0.26IFD for a line-to-neutral fault and bondedSGPRT-fault = 0.98IFT for a line-to-shield fault, unbonded

This results that a TL fault poses a higher risk if

IFT

IFD> 0.65 (16)

Although the breaker trip factor (�) depends on the selection ofbreakers, most TL faults will have a fault current IFT that satisfies(16), if not both (15) and (16). Thus we can conclude that the TL faultis more severe than the DL fault even though a TL fault is clearedfaster.

J. Acharya, W. Xu / Electric Power Systems Research 80 (2010) 1074–1081 1079

Table 4GPRs for line-to-ground fault and line-to-wire fault.

Ground potential rise Line-to-ground fault Line-to-wire fault

Faultlocation

Islandedend

Faultlocation

Islandedend

NGPR (D-line fault) (V/kA) 250 500 300 500SGPR (T-line fault) (V/kA) 110 220 980 220

Table 5Summary of NGPR results obtained analytically.

Configurations D-line fault(assume, IFD = 1 kA)

T-line fault(assume, IFT = 5 kA)

V/kA V V/kA V

D-line only 500 500 – –D-line + T-line but no shield 500 500 190 950D-line + T-line with MGS 500a 500 180 900

4

tfl(itmffwle

ci

rt3wwFeaNtmidT

TS

Fig. 9. NGPR profiles for the T-line-to-shield fault (except for D-line only).

Fig. 10. SGPR profiles for the T-line-to-shield fault.

Fig. 11, the NGPR has two equal peaks at the ends of the parallelsection when the fault lies downstream of the parallel zone. Onthe other hand, the SGPR profile in Fig. 12 exhibits extremely high

D-line + T-line andMGN–MGS bonded

500a 500 300 1500

a No change because max NGPR occurs outside the parallel zone.

. Results

Table 4 shows the GPR results for line-to-ground fault and line-o-wire fault. For the distribution line, the NGPR at the fault locationor the line-to-neutral fault is marginally higher that that for theine-to-ground fault. This is because a large part of the fault currentIFD) flows into the neutral wire, leading to less current dissipatingnto the earth through (RG). Consequently the NGPR is less althoughhe neutral wire comes in contact with phase wire. For the trans-

ission line, however, the SGPR for the line-to-shield fault is byar larger as the shield wire does not carry a significant part of theault current. In summary, the faults involving the multi-groundedire (neutral or shield) are worse. It is important to note that the

ine-to-wire fault has virtually no impact on the GPR of the islandednd (node 1) of the multi-grounded wire (Fig. 3 and Table 4).

As mentioned earlier, the maximum GPR will occur at theurrent injection nodes. Tables 5 and 6 show the summary of max-mum NGPR and SGPR for the T-line and D-line configurations.

The reminder of this section shows the computer simulationesults. Figs. 9 and 10 depict the GPR profiles for different configura-ions under the line-to-shield fault. The fault occurs on T-line about.5 km from the substation. For the single circuits, same distanceas considered for the fault from the source. The T-line fault currentas approx 12.6 kA and the D-line fault current was about 2.5 kA.

ig. 9 shows that the maximum NGPR occurs on the upstreamnd of the parallel section. But the maximum SGPR always occurst the fault locations. The T-line faults significantly increase theGPR compared to D-faults and the situation becomes worst when

he neutral and shield wire are bonded. On the other hand, the

aximum SGPR of the single T-line remains the same even after

ntroduction of D-line (unbonded case) as shown in Fig. 10. Itecreases significantly when the neutral and shield are bonded.hese results agree with the analytical results.

able 6ummary of SGPR results obtained analytically.

Configurations D-line fault(assume, IFD = 1 kA)

T-line fault(assume, IFT = 5 kA)

V/kA V V/kA V

T-line only – – 980 4900T-line + D-line but no neutral 95 95 980 4900T-line + D-line with MGN 54 54 983 4915T-line + D-line and MGN–MGS

bonded266 266 300 1500

Fig. 11. NGPR profiles for different fault locations (fault on T-line).

The GPR profiles are further illustrated in Figs. 11 and 12 forvarious fault points along the T-line. The GPR/kA index is used inthese figures since only the T-line faults are compared. As seen in

peaks at the fault locations. The main reason is magnitude of faultcurrent.

Fig. 12. SGPR profiles for different fault locations (fault on T-line).

1080 J. Acharya, W. Xu / Electric Power Systems Research 80 (2010) 1074–1081

Fig. 13. NGPR profiles for different fault locations (fault on T-line, N–S bonded).

F

iuaziFzfd

4

ps

stibAeh

Fig. 16. Comparison of NGPR for D-line faults and T-line faults.

ig. 14. SGPR profiles for different fault locations (fault on T-line, N–S bonded).

For the bonded circuits, the NGPR and SGPR profiles are shownn Figs. 13 and 14, which are different from Figs. 11 and 12 fornbonded cases. Fig. 13 shows that the NGPR becomes the highestt the fault location provided that the fault occurs in the parallelone. If the fault moves outside this zone (downstream), the max-mum NGPR occurs at the upstream end of the parallel section. Inig. 14, the SGPR is the highest for the fault outside the parallelone and is similar to that of Fig. 12 (unbonded). However, if theault occurs in the parallel zone the maximum SGPR will decreaserastically (compare Figs. 12 and 14).

.1. The maximum GPR magnitudes

This case involves T-line fault (to shield) in the middle of thearallel section. It is clear from Fig. 15 that the NGPR increasesignificantly when the fault occurs on T-line.

The results (Fig. 15) also reveal that the bonding of neutral andhield increases the NGPR. However, the maximum SGPR decreaseso the level of NGPR. This is for the following reason. The NGPRs produced by the induced current when unbonded. But after

onding, the T-line fault current is directly involved in the NGPR.lthough the equivalent impedance is reduced due to bonding, thisffect is dominated by the amount of T-fault current. On the otherand, the decrease in SGPR is the reflection of reduction in equiva-

Fig. 15. Max GPR magnitudes for different configurations (fault on T-line).

Fig. 17. Comparison of SGPR for D-line and T-line faults.

lent impedance. The T-fault current involved in SGPR is almost thesame before and after bonding.

The maximum NGPR for D-line and T-line faults are shown inFig. 16. The NGPRs caused by the T-line faults are much higher thanthat caused by the D-line faults. For T-line faults, the bonding con-figuration gives the worst NGPR. The maximum NGPR for the D-linefaults is not affected whether the shield is present or not. There isno effect of bonding on the maximum NGPR because the maximumoccurs outside the parallel zone of T&D lines. The maximum SGPRfor D-line and T-line faults are shown in Fig. 17. The SGPRs caused bythe D-line faults are negligible compared to those caused by T-linefaults. Unlike the NGPR, the SGPR will be reduced due to bondingfor the T-line faults.

The following conclusions can be drawn from the above figures:

• The maximum NGPR will increase significantly as a result of faulton T-line.

• The shield wire on top of the T-line does not have any noticeableeffects on NGPR. However, if the shield and neutral are bonded,the NGPR will increase considerably. Note that the effect of bond-ing applies in the parallel zone only.

• For the T-line faults, the maximum SGPR does not increase by theD-line configurations (Fig. 15). Similarly, for the D-line faults, themaximum NGPR will not be increased by the T-line configurations.

4.2. Effect of T&D parallel exposure length

The parallel length of T-line and D-line was varied from 1 to5 km. Fig. 18 shows that the NGPR increases initially with the length

of exposure when the fault occurs on T-line (middle of parallel sec-tion). However, it does not increase further when the parallel lengthis more than 3 km. Fig. 19 shows that the SGPR is independent ofthe exposure length when the fault occurs on T-line.

Fig. 18. Max NGPR for parallel exposure lengths (T-line fault in the parallel section).

J. Acharya, W. Xu / Electric Power System

Fig. 19. Max SGPR for parallel lengths (T-line fault in the parallel section).

Fig. 20. Max NGPR for parallel lengths (fault on T-line d/s of the parallel section).

F

uTzN(rmi

5

so

[

[

[

[

ig. 21. Max SGPR for parallel lengths (fault on T-line d/s of the parallel section).

It can be seen from Fig. 18 that the NGPR for the bonded config-ration is much higher than that for the unbonded configuration.his is particularly true if the fault location is inside the parallelone (where bonding exists). But if the fault occurs outside, theGPR will be less for bonded case compared to the unbonded case

Fig. 20). Similarly the effect of bonding on the maximum SGPR iselevant when the fault is inside the parallel zone (Fig. 19) and theaximum SGPR does not change due to bonding if the fault location

s outside the parallel zone where bonding does not exist (Fig. 21).

. Conclusions

The GPR characteristics of multi-grounded neutral (MGN) andhield (MGS) in the joint systems are illustrated. Main conclusionsf this study are summarized as follows:

s Research 80 (2010) 1074–1081 1081

• The line-to-wire faults are generally more severe than the line-to-ground faults. The effects of these faults are identical on theparallel line.

• Bonding of neutral and shield does not improve GPR. It will createhigher level of GPR in the parallel circuit.

• The GPR caused by T-faults can still be dangerous than that causedby D-faults even if T-faults clear faster.

• The GPR will increase for a certain range of parallel exposurelength.

References

[1] M.V. Lat, Determining temporary overvoltage levels for application of metaloxide surge arresters on multi-grounded distribution systems, IEEE Transac-tions on Power Delivery 5 (2) (1990) 936–946.

[2] Y. Rajotte, J. Fortin, G. Raymond, Impedance of multi-grounded neutrals onrural distribution systems, IEEE Transactions on Power Delivery 10 (3) (1995)1453–1459.

[3] Y. Rajotte, J. Fortin, B. Cyr, Lightning overvoltages on LV networks fed by MVlines with a multi-grounded neutral, in: CIRED Conference, 1999.

[4] Y. Rajotte, J. Fortin, B. Cyr, G. Raymond, Characterization of the groundimpedance of rural MV lines on Hydro-Quebec’s system, in: CIRED Conference,1997.

[5] K. Oka, J. Yoshinaga, S. Koizumi, S. Uemura, Y. Ariga, Study of neutral groundingfor 22 kV distribution systems, IEEE PES T&D Conference and Exhibition (3)(2002).

[6] K. Oka, S. Koizumi, K. Oishi, T. Yokota, S. Uemura, Analysis of a neutral groundingmethod for a three-phase four-wire 11.4 kV distribution system, IEEE PES T&DConference and Exhibition (2) (2002).

[7] T.H. Chen, W.C. Yang, Analysis of multi-grounded four-wire distribution sys-tems considering the neutral grounding, IEEE Transactions on Power Delivery16 (4) (2001) 710–716.

[8] L. Levey, Calculation of ground fault currents using an equivalent circuit and asimplified ladder network, IEEE Transactions on Power Apparatus and SystemsPAS-101 (8) (1982) 2491–2497.

[9] L. Levey, A new method of analysis for multi-grounded cable shields induc-tively and conductively coupled to multi-grounded neutral power lines,IEEE Transactions on Electromagnetic Compatibility EMC-29 (2) (1987) 116–125.

10] L. Levey, Computation of fault currents and voltages along a multi-groundedneutral power line having multiple phase conductors, IEEE Transactions onPower Delivery 6 (4) (1991) 1541–1548.

11] J. Acharya, Y. Wang, W. Xu, Temporary overvoltage and ground potential risecharacteristics of distribution feeders with multi-grounded neutral, IEEE Trans-actions on Power Delivery, in press, 9 pp.

12] H.W. Dommel, EMTP Theory Book, Microtran Power System Analysis Corpora-tion, British Columbia, Canada, 1996.

13] C.F. Dalziel, A study of the hazards of impulse currents, AIEE Transac-tions on Power Apparatus and Systems 72 (2 (Part III)) (1953) 1032–1043.

Janak Acharya (S’05) received the B.Sc. degree in electrical engineering from Tribhu-van University, Nepal, in 2002. He obtained the M.Sc. degree in electrical engineeringfrom the University of Saskatchewan, Saskatoon, Canada, in 2005. Currently, he isworking toward the Ph.D. degree at the University of Alberta, Edmonton, Canada.His research interests are power quality and reliability.

Wilsun Xu (F’05) received the Ph.D. degree from the University of British Columbia,Vancouver, Canada, in 1989. He was an engineer with BC Hydro, BC, Canada, from1990 to 1996. Dr. Xu is presently a NSERC Industrial Research Chair and a Professorwith the University of Alberta, Edmonton, Canada. His main research interests arepower quality and harmonics.