1110 IEEE Transactions on Power Delivery, Vol. 14,No.

3, July 1999

Transient Short Circuit Current Calculation using Decoupled Networks

P. J. LagacC
L. A. Dessaint J. Mahseredjian A. Chartrand
M. Lavoie
Ecole de Technologie Superieure Hydro-Quebec (iREQ) Hydro-Quebec
1 100 rue Notre-Dame Ouest 1800Boul. Lionel-Boulet Complexe Desjardms, CP 10000
MontrCal, QuCbec, Canada H3C 1K3 Varennes, Quebec, Canada J3X 1S1 Montrdal, Qudbec, Canada H5B 1H7
ABSTRACT Classical fault analysis methods for power sys- algorithm can handle single or multiple fault cases by apply-
tems are based on symmetrical components to represent vari- ing superposition in the final calculation of vrn.
ous fault configurations. The complete electrical network is With increased speed and memory capacity of modern
decomposed using a complex transformation matrix into computers, research efforts in the area of short-circuit analysis
decoupled zero, positive and negative sequence circuits. have continued beyond the conventional system modeling
Faulted power system steady-state voltages and currents are approach. Equivalent nonlinear impedance characteristics
computed using a nodal analysis solution method. have been developed to determine the effect of varistors on
This paper presents an extension of the above methodol- the short-circuit current fundamental component [3]. The
ogy using discretized RLC elements for the computation of effect of unbalanced nonlinear impedance on the steady-state
the actual transient response of a network for various fault fault current has been evaluated by implementing an iterative
types. The decoupled networks are obtained using a real trans- solution procedure with shunt current injection in the
formation matrix. sequence networks [4]. The combination of time and fie-
quency domain techniques has been developed to include
Keywords: Short-circuit currents, transients, fault analysis combinations of constant impedance, current and power load
models [5]. Conventional short-circuit methodology has been
combined with power-flow solutions allowing the consider-
1. INTRODUCTION ation of steady-state unbalanced system conditions [6].
Recently artificial neural networks have been used for detec-
A short-circuit program is a useful and basic power system ting fault induced high frequency noise [7].
analysis tool. Traditional short-circuit calculations are per- The present approach is to estimate the maximum asym-
formed using symmetrical components [1][2]. A three-phase metrical short-circuit current from the symmetrical short-cir-
network of arbitrary topology is decomposed into decoupled cuit current using the Xm method and assuming exponential
zero, positive and negative sequence networks using a com- decay behavior of fault current contributions typical for first
plex transformation matrix. A common approach is to write order circuits [SI. For higher order circuits found in series
sequence network equations through nodal analysis formula- compensated networks, exponential decay behavior alone is
tion: Y,V,=Im for the sequence counter m = 0,1,2. yrnis not sufficient to represent transients and an alternate method
the complex admittance matrix, V, is the complex vector of must be used. In a continuous effort to extend the range of
unknown node voltages and I, is the complex vector of applicability of the commonly used short-circuit software, this
known nodal current injections. The pre-fault sequence volt- paper proposes to implement a new approach based on numer-
ages and currents are found from a typical load-flow solution. ical integration and related to EMTP [9] type simulations.
The sequence network equations are used to compute the fault
impedances (Zo, Z, and Z2) at fault location required with
pre-fault voltages in the solution of fault current equations. 2. NETWORK EQUATIONS
The sequence network equations are reused in the computa-
tion of the faulted network steady-state solution. This type of Since symmetrical components involve complex coeffi-
cients, their use is not well suited for time-domain computa-
PE-307-pwRD-O-08-1998 A paper recommended and approved by tions where all quantities are real. It is chosen to create
the IEEE Transmission and Distribution Committee of the IEEE Power decoupled circuits using a , p and 0 components which are
Engineering Society for publication in the IEEE Transactions on Power
Delivery. Manuscript submitted December 29,1997; made available for
related to phase quantities as follows [ 101:
printingAugust 14, 1998. Vabc = T V O ~ P (1)
which is :

0885-8977/99/$10.00 0 1998 IEEE

 1111
Transformation from phase quantities is given by: any network node.
VOaP = T' vabc (3)
Bold characters are used to distinguish vectors and matrices.
3. FAULT EQUATIONS
Similar equations are written for relating currents i h and
ioab. This transformation is used in both fiequency domain The three decoupled fault circuits seen at a fault bus are
and time-domain to create 3 decoupled networks from a 3- presented P Fig. 2. The subscript F is used to denote the fault
phase system. terminal. V,F(t) is the Thevenin voltage contributed by the
To achieve a time-domain nodal solution, real parameters steady-state solution, R,F is the Thevenin resistance,
are needed as network data and a discretized model equivalent i,W(t-At) is contribution from history current sources in
must be evaluated for all RLC components of the network. A the transient response, VmF(t) is the fault branch voltage and
generic discrete equivalent for the RLC branch at time t is i,F(t) is the fault current. If there is no fault current injected
shown in Fig. 1. The current source is a history term evaluated in a given circuit then the corresponding history
at the previous time-point (t - At ). It is chosen to program i,w(t -At) is zero.
trapezoidal integration for stability and precision and to use
halved time-step (At / 2 ) Euler-backward for calculating ini-
tial conditions for all state variables and to avoid numerical
oscillations at switching instants [ll].
It is recalled that the equivalent resistance equation
,R = R + 2L / At + At / (2C) is the same for both integration
methods, but the history current equation in trapezoidal inte-
gration [9]:
ih(t-At)= (A\
\-/

2 v ~ (-t At) - v(t - At) +

is different from the Euler-backward
At
ih(t--)=2 (5)

ipF(t)

Figure 2: Decoupled fault circuits seen at the fault bus

These circuits can be used to compute fault currents for

- arbitrary fault conditions, some typical cases are presented
Figure 1: A discretiid RLC branch component below. It is noticed that in addition to standard fault configura-
tions the undertaken methodology allows formulating equa-
When all network components are given the representation tions for series-fault conditions such as the one-conductor-
of Fig. 1, a set of 3 decoupled nodal analysis equations can be open or two-conductors-open cases.
written:
Ymvm = i, m = O,a,P (6) 3.a Sinele-line-to-mound fault
The right hand side current vectors are constructed from his-
C O 0
tory term currents sources. A history current source contrib- b o 0
utes a positive value to the node it enters and a negative value
to the node it excites. The ground node is eliminated. This set
of equations is solved at each simulation time-point for v, .
The contribution fiom steady-state conditions is found Figure 3: Single-line-to-ground fault
from? standard load-flow solution saved in phasor vectors
VO, VOland Vo . The real part ,Of a phasor vector evaluated at The phase conditions for this type of fault (see Fig. 3) are:
each tune-point t is given by Vm(t) and can be added to its ibF(t) = 0 icF(t) = 0 (7)
time-domain solution from v, to obtain the final solution at
and using the transformation matrix Tf :
 3.d One-conductor-oDen fault

The circuit configuration for the one-conductor-open con-

dition is shown in Fig. 4. The fault current equations are
derived by applying superposition of decoupled fault circuits
looking into the left side (L) and then into the right side (R).

CLO
ik
-- if
OCR

bLo ik - -:i obR

aLo i,L- *R
- ='a oaR
Figure 4: The one-conductor-open fault case
3.b Three-Dhase-to-eround fault
The phase conditions are given by:
If two identical fault branches are added to phases b and c vk(t) = vE(t) v)(t)= VF(t) (26)
in Fig. 3 for a balanced network, it appears that:
i#(t)+ibF(t)+icF(t)= 0 iOF(t)= 0 (15) ik(t)=O ia(t)=O
R ib(t)=-iE(t)
L ik(t)=-if(t) (27)
The Oap component relations are found using the transfor-
i#(t) = J E i , F ( t ) (16)
mationmatrix T' :
v$(t) = 1/ & V o F ( t ) + m v@(t) (17) (28)
According to Fig. 2: JZvbF(t)-v~(t)=JZviF(t)-ViF(t) vfjF(t)= vlF(t)
) ~ ~ F ( ~ ) + R , F ~ , ~ F ( ~ - A ~ ) - R , F ~ , F ( & ) (18)
v a ~ ( t=
If equations and (18) are replaced into (17) and combined ibF(t) = -ioF(t)
R i k ( t ) = -i&(t) iL (t) = -iRPF (t) (29)
PF
with ( 12) and (16), it is found that:
ibF(t) = - J Z i b ( t ) (30)
When equations (10) and (18) are rewritten for left (L) and
right @) quantities and combined with equations for 0 and a
circuits fiom (28) to (30), the 0 current component becomes
isolated
ibF(t) = (3 1)

In a similar manner, the p circuit equations are taken from

(28) and (29) and combined with the left (L) and right (R) ver-
sions of equation (24) to give:
 1113

rent sources in i, .
significant speed-up is achieved'by noticing that
ed networks can be solved on completely indepen-
dent processors.

4. IMPLEMENTATION -%.
5. APPLICATION CASES

The method presented in this paper has been implemented The first test case is the IEEE-14 bus system with added
in a production level package. It constitutes a direct upgrade 40% series compensation on the line connecting buses 7 and
and simply a new option for a standard short-circuit analysis 8. Fig. 5 shows the fault current resulting fiom a three-phase-
program used and maintained at Hydro-Qukbec. The needed to-ground fault at bus 7. The solution with the new short-cir-
input network data remains the same, since the zero, positive cuit program is superposed over the solution found fiom the
and negative sequences are the same as the 0 , a and p com- EMTP (DCG-EPRI version 3). Both waveforms are almost
ponents respectively. Transmission lines remain modeled as identical. Constant impedance loads and 60 Hz pi-sections for
fundamental fiequency pi-sections. For a balanced transmis- transmission lines are used in both programs.
sion line the a and p mode impedances are identical. The
new option allows data fields for fault occurrence time in
addition to fault location and fault type. It is allowed to apply
multiple-simultaneous and multiple-non-simultaneous faults.
The overall solution process can be summarized through the
following algorithmic steps:
v,,
1. create the vector tables
2. find the needed resistances R,F for all fault locations
3. compute fault currents using fault equations
4. use sparse matrix methods to solve equation (6)
-L
5 . update history currents and increment time b , ,

20
I

' 0 tims) ' 0 80 ' io0

6. go back to step 3 until the maximum simulation time Figure 5: Three-phase fault current, IEEE-14 system
There are no At delays in this procedure. The vectors v, are
The second test case is taken from a Hydro-QuBbec 800
found from a standard load-flow software with PQ-load mod-
els. Both To and vp are zero for balanced network condi- busbar and a 1000 branch network study. Simulation was done
tions, but the presented methodology remains functional if with load modeled as constant impedance, transformers as
equivalent T models, lines as equivalent pi-section models
unbalanced conditions are considered. The Thevenin resis-
and DC links as constant current sources. Generators were
tances R,F are found by injecting a unity current source at modeled using a fixed voltage source behind a linear transient
the fault location and calculating the corresponding voltage reactance. Further detailled simulations are also possible since
from equation (6). An intermediate step between steps 3 and 4 various models are available for representing nonlinear com-
is the updating of i, (in equation (6)) with fault and history ponents of the network [9] The line-to-ground fault at Mon-
current sources. History updating for trapezoidal integration is tagnais substation is shown in Fig. 6, both transient and
replaced by two consecutive At I 2 Euler-backward updates at steady-state solutions found in the same short-circuit program
each fault initiation and termination. are superposed.
It should further be considered that balanced networks
leads to decoupled 0 a p circuits and for these cases it is not
always necessary to use all the possible values ( 0 ,a and p )
of m . The fault equation (14), for example, requires only the
o and a circuits. For other cases, such as the line-to-line
fault, only one decoupled circuit contributes to the fault cur-
rent. Such conditions require only 213 or 113 of the equation
system solved in standard EMTP-type transient analysis meth-
ods with coupled three-phase representation. V V
In series-fault cases, such as the one-conductor-open fault, I
, I
only a single step can be programmed for both left and right 0
I 1
20
I I
40
I I
60
I I
80 100
circuit contributions by injecting two opposite sign fault cur- t (ms)
Figure 6: Line-to-ground fauit at Montagnais substation
1114
This illustrates the limitations of a traditional steady-state ing Conventional Power Flow and Fault Analysis Soft-
fault current calculation. Table 1 shows the frequency of the ware, IEEE Trans. Power Delivery, vol. 11, no. 1,
oscillation calculated fiom the FFT of Fig. 5 and Fig.4, and as February 1996, pp. 422-427.
estimated fiom series RLC circuits. Series-capacitance of
Hydro-Quebec case takes into account 40%line compensation
and equivalent series inductance obtained from the X/R
method for the fault at Montagnais substation.
[7] Z. Q.Bo, R. K. Aggarwal, A. T. J
A New Approach to Ph r ,H. Y.Li, and Y. H.
Selection using Fault
enerated High Frequency Noise and Neural Networks,
IEEE Trans. Power Delivery, vol. 12, no. 1, January 1997,
Table 1 Natural response frequency pp. 106-115.
[SI AIEE Committee Report : ANSI Application Guide for ac
High-Voltage Circuit Breaker Rated on a Symmetrical
IEEE 14 Current Basis, ANSI/IEEE C37.010-1979
20 22 [9] H. W. Dommel : Electromagnetic Transients Program
Reference Manual, EMTP Theory Book, Bonneville
Power Administration, Portland, USA, August 1986.
6. CONCLUSIONS
: Circuit Analysis of AC Power Systems,
This papercontributes a new method for extending the Wiley & Sons, New York 1943.
range of applicability of short-circuit analysis to transient [113J. Lin and J.R. Marti : Implementation of the CDA Proce-
analysis, while maintaining the standard front-end of the dure in the EMTP, IEEE Trans. Power System, vol. 5,
short-circuit program. It proposes a fast and parallel solution no. 2, May 1990, pp. 394-402.
approach through decoupled circuits combined in various
fault conditions for transient short-circuit current calculations. Pierre-Jean Lagack (M84) received the B.Sc.A.,
M.Sc.A., and Ph.D. in Electrical Engineering from Ecole Pol-
ytechnique de Montreal (Canada) in 1982, 1985 and 1988
7. ACKNOWLEDGMENTS respectively. He is presently a professor in Power Systems at
b o l e de Technologie Superieure.
The authors would like to acknowledge the Natural Sci- Louis A. Dessaint (MSS-SM91) was born in Paris,
ences and Engineering Research Council of Canada, and the France, in 1953. He received the B.Ing., M.Sc.A. and Ph.D. in
Government of Quebec for their financial assistance. Electrical Engineering from Ecole Polytechnique de Montreal
(Canada) in 1978, 1980 and 1985 respectively. He is presently
REFERENCES a professor of Electrical Engineering at Ecole de Technologie
Suweure, Montreal. His main research interests are the
[ 11 P.M. Anderson : Analysis of Faulted Power Systems, The microprocessor-based control of electric drives and robotics.
Iowa State University Press, 1973 Dr. Dessaint is a member of the Order of Engineers of
[2] C. L. Fortesque : Method of Symmetrical Coordinates Quebec.
Applied to the Solution of Polyphase Networks, Proc. MicheiLavoie (M) was born in Montreal, Canada, in
Conf. Amer. Inst. Elec. Eng., 1918. 1943. He received the B.Sc.A. degree in Physics Engineering
[3] D. L. Goldworthy : A Linearized Model for MOV-Pro- and the M.Ing. degree in Electrical Engineering both from
tected Series Capacitors, IEEE PES Trans. Power Sys- Ecole Polytechnique de Montreal (Canada) in 1971 and 1980.
tem T-PWRS Nov 87, pp.953-958. He is presently of professor in Power Systems at Ecole de
[4] J. Mahseredjian, P. J. Lagace, S . Lefebvre and A. Char- Technologie Superieure. He is a member of the Order of Engi-
neers of Quebec.
trand Superposition Technique for MOV-Protected
Jean Mahseredjian (M) received the B.A.Sc., M.A.Sc.
Series Capacitors in Short-Circuit Calculations, IEEE and Ph.D. in Electrical Engineering fiom Ecole Polytechnique
Trans. Power Delivery, vol. 10, no. 3, July 1995, pp. de Montreal (Canada) in 1982,1985 and 1990 respectively. At
1394-1400. present he is a researcher at Institut de Recherche dHydro-
[5] S . M. Halpin, L. L. Grisby, C. A. Gross and R.M. Nelms Quebec and an associate-professor at Ecole Polytechnique de
An Improved Fault Analysis for Unbalanced Multi- Montreal.
Phase Power Distribution Systems, IEEE Trans. Power Adrien Chartrand (M) received the B.Sc.A. in Electrical
Delivery, vol. 9, no. 3, July 1994, pp. 1332-1338. Engineering from Ecole Polytechnique de Montrkal (Canada)
[6] D. Reichelt, E. Ecknauer, and H. Glavitsch : Estimation in 1974. He joined Hydro-Quebec in 1974, and was involved
of Steady-state Unbalanced System Conditions Combin- in load-flow, short-circuit and stability studies. His activities
are currently in technical support and planning.