1

Assessment of Errors Introduced by Common

Assumptions Made in Power System Studies
Jaime Peralta, Student Member, IEEE, Francisco de León, Senior Member, IEEE, and
Jean Mahseredjian, Senior Member, IEEE
calculations, but only in the zero-sequence component.
Abstract—Power system studies using commercial or in-house Transient stability studies are typically (traditionally)
developed software are performed making several assumptions. performed using phasors and assuming balanced networks
This paper analyzes the errors introduced by the most common and loads. Only the positive-sequence representation of the
modeling simplifications made for load-flow, short-circuit, and
transient stability studies. Load-flow studies for transmission system is considered for symmetrical three-phase faults.
systems almost invariably rely on the positive-sequence The main contribution of the paper is to show that
approach. While this may be acceptable for most transmission significant errors can be introduced by the traditional
system conditions, errors of up to 15% can be found by not assumptions made in power system studies. Contrary to
considering the unbalanced nature of power systems common belief that the errors are in the order of 1% to 5% on
components. Errors larger than 10% are found in short-circuit the conservative side, this paper shows that the errors can be
studies by using symmetrical components. Common assumptions
also introduce differences in transient stability analysis. The as large as 15%, 10% and 14% for load-flow, short-circuit and
results for the studied test case show differences of up to 14% in transient stability studies, respectively. Moreover, the errors
the determination of the maximum clearing time. With the are not always on the conservative side.
current development status of computing hardware and Another contribution of the paper is to show that more
available algorithms, more accurate modeling is currently precise modeling methodologies are efficient and well within
available even for large systems. the capabilities of modern computers available to anyone.
Index Terms—Electromagnetic Couplings, Positive-Sequence, The problem with precision of currently used methods has
Multiphase Systems, Transmission and Distribution Networks, been recognized by CIGRE and a new Working Group
Unbalanced Load-Flow, Short-Circuit, Transient stability. C4.503 has been created in 2011. The primary scope of this
WG is the critical assessment of existing numerical modeling
I. INTRODUCTION and computation techniques for power system studies from

P OWER system studies are usually performed making a steady-state to switching transients. It includes an assessment
number of assumptions on the characteristics of of errors produced by traditional analysis methods, such as
transmission lines. For example, power flow studies in balanced networks, usage of sequence networks and
transmission systems are carried out using the positive- fundamental frequency based solutions. EMT-type methods
sequence equivalent network. This implies that the system is will be used as reference.
not only electromagnetically balanced (or continuously
transposed), but also that the loads are balanced. II. ACCURATE MODELS
The load unbalance is commonly considered in load-flow This section describes the characteristics of the detailed
studies for distribution systems by performing three-phase model for the different studies to be used as the basis for
simulations. However, continuous transpositions of lines and comparison. The EMTP-type (Electromagnetic Transients
cables are frequently assumed. The mutual coupling effects Program) “accurate” modeling presented in this paper is based
between parallel circuits are rarely (if ever) considered. on the EMTP-RV [4]-[5] software and methods, whereas the
Short-circuit studies are normally based on the standardized “traditional” studies are performed with the software packages
methods: IEC [1] or ANSI/IEEE [2], [3]. The calculations are PSS/E [6], PSAF and CYMDIST [7] for unbalanced
performed in the “sequence domain” using Fortescue’s distribution systems. EMTP-type methods are circuit based
symmetrical components transformation. This transformation and provide a time domain solution. Phasor solutions are used
only applies to electromagnetically balanced lines. However, in the traditional methods.
because of the many virtues of symmetrical components, the
A. Load Flow
electric power industry heavily relies on them. Mutual
couplings between parallel lines are considered in some fault Induced voltages in parallel circuits, transmission lines or
between phases, result from electromagnetic field coupling.
The electric field induction is mainly a function of the
J. Peralta is with the Ecole Polytechnique de Montreal, Montreal, Canada
(e-mail: [Link]@[Link]). proximity of parallel lines, the line phase spacing and the
F. de Leon is with Polytechnic Institute of NYU, Six Metrotech Center, voltage magnitude of the transmission line [8]. Magnetic field
Brooklyn, NY, 11201, (e-mail: fdeleon@[Link]). induction is a function of the geometrical configuration of the
J. Mahseredjian is with the Ecole Polytechnique de Montreal, Montreal,
Canada (e-mail: [Link]@[Link]). lines, the separation between parallel circuits, and the loading

978-1-4577-1002-5/11/$26.00 ©2011 IEEE

 2

conditions. Both effects are implicit in the calculation of conventional transient stability simulators, these equations are
mutual impedances and admittances that form part of the line integrated with the network equations together with the loads
primitive impedance and admittance matrices [9]. that can be represented as constant or voltage dependent
The adopted accurate model for load-flow studies is based impedances.
on considering the following characteristics of transmission Constant impedance loads are directly included as part of
and distribution systems: actual position of conductors in the the algebraic network equations. Voltage dependent loads are
line towers, coupling between parallel lines, transposed or represented as current injections at the corresponding node in
non-transposed lines, explicit modeling of neutral conductors the system network equation. Dynamic loads are typically
and grounding wires, multiphase (1 or more) modeling, load represented as induction or synchronous motors and are
unbalances and different load models. treated similarly to synchronous machines [10].
The accurate model used in this paper considers generator
B. Short-circuit
load unbalances by introducing a zero sequence component
Traditional short-circuit computation packages are based on for the stator equations typically neglected in positive-
symmetrical components for balanced networks where faults sequence solvers. In steady-state, the generator stator matrix
can be balanced or unbalanced. In some cases the loads are is a cyclic phase domain matrix. In time-domain, the full
neglected and the pre-fault voltages are assumed to be 1.0 pu. system set of equations includes machine damper windings
Machines are modeled using ideal sources behind the and zero-sequence effects in addition to the capability to
transient, subtransient or synchronous reactance. model saturation in d- and q-axis.
A more accurate description of physics is obtained by
solving the electric circuit in steady-state using the actual III. ERRORS INTRODUCED IN LOAD-FLOW STUDIES
multiphase representation with no simplifications in the
Traditional positive-sequence load-flow methods are based
network topology. The following features are included in the
on the assumption of ‘continuously-transposed’ transmission
selected accurate model: balanced and unbalanced loads using
lines. Positive-sequence impedances are computed assuming
constant impedance or constant current models, shunt
that the three phases are built in perfect symmetry and with
magnetizing and charging currents, line electromagnetic
ideal line transposition in order to balance the electromagnetic
couplings, actual pre- and post-fault network voltages and
couplings.
currents for all phases.
Neglected from both the approximated and the accurate A. Transposition and Mutual Couplings
types of simulations are the effects of the magnetic flux This section presents the differences found in node voltages
penetration in motors and generators. The system impedances when accurately modeling actual line transposition against the
are considered constant (time and frequency invariant). balanced positive-sequence (Pos. Seq.) representation. All
Additionally, since a phasor solution is obtained, the couplings between phases and nearby lines are considered in
asymmetry of short-circuit currents is not considered. Note, the accurate model.
however, that since a phasor solution is used to compute The studied four cases correspond to actual high voltage
short-circuit, the asymmetry caused by decaying dc transmission lines (or line sections) excluding all types of
components due to the time of switching are not considered. compensation (series or shunt) in order to assess the error
With time-domain simulations it is feasible to study the introduced by the electromagnetic unbalance effect only. The
actual short-circuit currents and obtain higher precision selected loading varies from 0.8 to 1.7 times their surge
including the dc components. In that case the fault currents impedance loading (SIL).
can show the effect of the changing reactance from
subtransient to steady-state value in rotating machines. The Double circuit 220 kV line
fault current asymmetries are also found in time-domain The load flow in a typical double circuit line of 60 km and
simulations. Another advantage in time-domain simulations, 220 kV is calculated when supplying a balanced load of 420
not explored in the paper, is the possibility of including MW and 120 Mvar. The computed receiving end node
nonlinearities, frequency dependent impedances and load voltages and corresponding errors are shown in Table I. For
models capable of representing the dynamics of asynchronous the accurate model the line is transposed producing three
machines. sections of equal lengths. The largest errors found in voltage
C. Transient Stability magnitude and angle are 2.5% and 1.1% (120° base)
respectively. The error in power losses for this line is 1.1%.
Synchronous machine models suitable for transient stability
studies are described in references [10] and [11]. All these
Single circuit 345 kV line
models are derived from Park’s transformation in which the
This case corresponds to a 150 km, single circuit and 345
machine electrical behavior can be represented by equivalent
kV line including three transpositions, similar to the previous
circuits in the rotor direct and quadrature axes. Generator
case, for the accurate representation. The results of voltage
equations are algebraic equations represented by impedances
errors are listed in Table II for a normal load of 240 MW and
in d- and q-axes and assuming a balanced load which allows
90 Mvar, and in Table III for a heavy load of 360 MW and
neglecting the zero sequence components. In most
150 Mvar. The voltage magnitude error is very small (0.1%)
 3

under normal loading, which confirms the accuracy of the Three 735 kV parallel lines
assumptions in positive-sequence solvers. However, the error This test case includes three 735 kV single-circuit parallel
in losses is 1.8%, which is not negligible. lines of 230 km feeding a load of 1,650 MW and 300 Mvar.
From Table III, it can be observed that heavy loading The three lines are modeled as mutually coupled and
slightly increases the voltage magnitude error to 0.3%. The transposed at three locations in the accurate representation,
power loss error is increased to 2.1%. If this 345 kV line is while the positive sequence model assumes three perfectly
not transposed, the largest voltage magnitude errors become balanced and decoupled lines. The three lines have a
1.8% and 4.1 % for the normal and heavy loaded cases separation of 38 m between each other. The results presented
respectively. in Table V show that the largest error in voltage magnitude is
TABLE I. 3.3%. The error in power losses is 0.7%.
Node Voltages – 60 km 2 x 220 kV Transposed Line In order to isolate the impact of electromagnetic couplings
Voltage Mag Error Angle Error between parallel lines, an additional case was simulated to test
phase (pu) (%) (deg) (%) the error introduced when there is only one 735 kV line. The
results presented in Table VI show that the voltage magnitude
C1 - A 0.958 2.5% -6.32 1.1%
error in phase C is 1.6%. Taking the difference with the
C1 - B 0.920 -1.5% -127.62 0.1%
voltage error phase C of circuit 3 in Table V we get that the
C1 - C 0.932 -0.2% 112.24 -0.2%
effect of having parallel lines is approximately 1.7%, which
C2 - A 0.958 2.5% -6.32 1.1% means that approximately half of the voltage error comes
C2 - B 0.921 -1.5% -127.63 0.1% from the effect of line unbalance and half from parallel line
C2 - C 0.933 -0.2% 112.24 -0.2% couplings for this particular case.
Pos. Seq. 0.935 - -7.55 -
TABLE V.
TABLE II. Node Voltages – Three Parallel 735 kV Transposed Lines
Node Voltages – 150 km, 345 kV Transposed Line, Normal Load
Voltage Mag Error Angle Error
Voltage Mag Error Angle Error phase (pu) (%) (deg) (%)
phase (pu) (%) (deg) (%) L1 - A 0.940 -1.8% -14.04 0.0%
L1 - A 0.956 -0.1% -6.68 -0.1% L1 - B 0.951 -0.6% -134.65 0.4%
L1 - B 0.956 0.0% -126.56 0.0% L1 - C 0.980 2.5% 107.07 1.1%
L1 - C 0.958 0.1% 113.49 0.1% L2 - A 0.937 -2.1% -13.87 0.2%
Pos. Seq. 0.957 - -6.59 - L2 - B 0.953 -0.4% -134.75 0.5%
TABLE III. L2 - C 0.986 3.0% 107.41 1.4%
Node Voltages – 150 km, 345 kV Transposed Line, Heavy Load L3 - A 0.935 -2.3% -14.07 0.0%
L3 - B 0.954 -0.3% -134.93 0.6%
Voltage Mag Error Angle Error
L3 - C 0.988 3.3% 107.43 1.4%
phase (pu) (%) (deg) (%)
Pos. Seq. 0.957 - -14.07 -
L1 - A 0.904 -0.3% -10.47 -0.1%
L1 - B 0.906 0.0% -130.36 0.0% Fig. 1 shows the behavior of the largest error introduced in
L1 - C 0.908 0.1% 109.76 0.1% the node voltage magnitude at the end of two 735 kV parallel
Pos. Seq. 0.907 - -10.37 - lines for different line separation distances. The impact starts
Double circuit non-transposed 440 kV line to be negligible (lower than 0.5%) for separation distances
This case considers a 60 km, double circuit and 440 kV line larger than 100 m. This error represents only the impact of
that is not transposed in the accurate representation. The load- parallel circuits and does not include the error caused by the
flow was simulated when supplying a balanced load of 600 fact that the lines are not perfectly balanced.
MW and 150 Mvar. The largest error in voltage magnitude is TABLE VI.
Node Voltages –735 kV Transposed Line
14.7% (see Table IV). The error found in power losses is
1.2%. Voltage Mag Error Angle Error
phase (pu) (%) (deg) (%)
TABLE IV.
Node Voltages – 60 km 2 x 440 kV Non-transposed Line A 0.938 -1.9% -14.84 -0.7%
B 0.960 0.4% -134.35 0.2%
Voltage Mag Error Angle Error C 0.972 1.6% 106.96 1.0%
phase (pu) (%) (deg) (%) Pos. Seq. 0.957 - -14.07 -
C1 - A 0.826 -11.5% -13.48 0.5%
C1 - B 0.929 -0.5% -139.58 4.2% In general, important differences can be observed in the
C1 - C 1.071 14.7% 112.11 5.7% voltage magnitudes and angles when the traditional positive-
C2 - A 0.826 -11.5% -13.48 0.5% sequence representation is used instead of accurate line
C2 - B 0.929 -0.5% -139.58 4.2% models. Depending on the line configuration, loading, length
C2 - C 1.071 14.7% 112.11 5.7% and transposition, these voltage errors may vary from 0.3% to
Pos. Seq. 0.933 - -13.98 - 3.3%. There are cases when the error could be as high as 15%
when lines are non-transposed. Although no pattern of
 4

behavior was found for the error in the power losses, the distribution systems (±10%). Voltage angle error also follows
positive-sequence assumption presents errors from 0.7% to an exponential pattern as long as the load power factor is kept
2.1%. constant as shown in table VII.
Voltage Magnitude Error
735 kV - Parallel 25.2 kV radial feeder
1.2%
A second line section operating at 25.2 kV was also
1.0% modeled using the line configuration as provided in [12].
Phase and neutral conductors used are 266 ACSR and 1/0
0.8%
ACSR respectively. The same load-flow test was performed
0.6% for this 25.2 kV line configuration and the results are shown
in Tables IX and X.
0.4%
TABLE IX.
0.2% 25.2 kV - Largest Voltage Error as a Function of Line Length

0.0% Length Mag Error Angle Error

0 5 100 150 200 250 300 350 400 450 500
(km) (pu) (%) (deg) (%)
Separation between lines [m]
5 0.975 -0.4% -1.45 0.0%
Fig. 1. Voltage magnitude error versus line separation – 735 kV line
10 0.947 -0.9% -3.00 0.0%
B. Distribution Lines Unbalance and Transposition 15 0.915 -1.5% -4.68 -0.1%
This section analyzes the errors introduced on node voltages
TABLE X.
in non-transposed and line-unbalanced distribution systems 25.2 kV, 5 km - Largest Voltage Error as a function of Line Loading
when perfect transposition or phase spacing symmetry is
assumed. The study uses two typical line section feeders Load Mag Error Angle Error
operating at 12.47 kV and 25.2 kV. (MVA) (pu) (%) (deg) (%)
9 0.975 -0.4% -1.445 0.0%
12.47 kV radial feeder 15 0.956 -0.7% -2.46 0.0%
A typical 12.47 kV line section was modeled accurately 26 0.911 -1.4% -3.63 0.0%
using the line configuration provided in [12]. Phase and
neutral conductors used are 336.4 ACSR and 4/0 ACSR Similarly to the 12.47 kV line, the voltage errors increase
respectively. exponentially with the line section length and loading.
Different balanced loads were placed at the end of the line However, for the same range of per unit voltage drop, the
section to assess the impact of loading variation in voltage voltage magnitude errors are smaller for the 25 kV case.
errors. The length of the line was also varied to assess its Therefore, it can be concluded that the lower the nominal
impact on phase voltages. The results of the largest phase distribution voltage level, the higher is the error introduced by
voltage error (compared with the equivalent positive-sequence neglecting the non-transposition and line unbalances.
voltage) as a function of the length and the load are presented After substantial experimentation, we are able to establish a
in Tables VII and VIII. rule of thumb which states that, in average, the expected
TABLE VII. magnitude voltage errors in 12 kV non-transposed distribution
12.47 kV - Largest Voltage Error as a function of Line Length networks when assuming ideal transposition is between 0.04-
Length Mag Error Angle Error 0.08 %/MVA/km. For 25 kV lines, this error varies between
(km) (pu) (%) (deg) (%) 0.008-0.015 %/MVA/km. As an example, the maximum error
found on a 12 kV feeder of 10 km long with a load of 5 MVA
5 0.968 -0.7% -2.01 0.0%
installed at the end would be between 2 and 4%. This rule
10 0.932 -1.5% -4.23 -0.1%
does not apply to feeders having voltage regulators and
15 0.891 -2.6% -6.76 -0.3% switched capacitor banks nor does it apply to the error
TABLE VIII. introduced in the voltage angle.
12.47 kV, 5 km - Largest Voltage Error as a function of Line Loading
C. Distribution Load Unbalance
Load Mag Error Angle Error
It has been observed that current unbalances in distribution
(MVA) (pu) (%) (deg) (%)
systems can rise to 15%-20% in long radial and rural feeders.
3 0.968 -0.7% -2.01 0.0% These load unbalances may create voltage unbalances greater
6 0.941 -1.4% -4.46 -0.1% than 3.0%, which is harmful to some customer loads and
8 0.890 -2.2% -4.89 0.0% inhibits the correct operation of protection devices. There
exist commercial packages that are able to solve load
From tables VII and VIII, it can be observed that voltage unbalanced load-flow for weakly meshed distribution systems
magnitude errors increase exponentially with length and using fast iterative (backward-forward sweep) solvers.
loading for a wide range of operational voltages in Although available to most distribution utilities, there are a
 5

few that still model distribution unbalanced systems using loads in phases a and c are inverted for both the 12.47 kV and
conventional positive-sequence solvers. This section reviews the 25.2 kV test cases. The largest error in voltage magnitude
the effects of load unbalance on phase voltages and compares is reduced from 5.2% to 2.7% and from 2.9% to 0.3% for
the errors for the same two distribution systems presented in 12.47 kV and 25.2 kV systems respectively. This error
the previous section. cancelation effect must be seen as a coincidental result and
The 5 km long 12.47 kV system is modeled with the loads not as a consistently repeatable effect.
taken from the IEEE 4-Node Test Feeder [12] case and shown TABLE XIV.
in table XI. This load represents a current unbalance of 25% 12.47 kV Load Unbalance - Counteract Effect
which can be found in some rural distribution systems. In
Voltage Mag. Error Angle Error
addition, the 25.2 kV case was modeled with a load
representing an unbalance of 5.0% which is considered a Phase (pu) (%) (deg) (%)
more representative value for urban systems. Tables XII and a 0.959 2.7% -0.87 2.1%
XIII show the results and errors for the 12.47 kV and 25.2 kV
b 0.913 -2.1% -122.60 -0.6%
cases respectively. For the 12.47 kV line section, the largest
voltage magnitude error is 5.2%. This value contrasts with the c 0.922 -1.2% 113.33 -2.8%
1.4% error obtained when only electromagnetic unbalances Pos. Seq. 0.933 - -3.36 -
are considered, as shown in Table VIII. For the 25.2 kV case, TABLE XV.
a load unbalance of 5% results in the largest error in voltage 25.2 kV Load Unbalance – Counteract Effect
magnitude of 2.9% as shown in Table XIII. This is almost
double of the 1.5% error obtained with similar length but load Voltage Mag. Error Angle Error
balanced network shown in Table IX. Phase (pu) (%) (deg) (%)
TABLE XI. a 0.917 -0.3% -3.73 0.4%
Load Model for 12.47 kV and 25.2 kV Systems b 0.915 -0.4% -123.81 -0.3%
12.47 kV 25.20 kV c 0.925 0.6% 115.14 -0.6%
Phase
P (kW) Q (kvar) P (kW) Q (kvar) Pos. Seq. 0.919 - -4.14 -
a 1,275 790 2,800 920
b 1,800 872 2,970 976 Although voltage magnitude errors effects can be minimized
when load phasing is properly handled, this is difficult to
c 2,375 781 3,070 1,009
achieve in actual distribution systems.
Total 5,450 2,443 8,840 2,906
TABLE XII. IV. ERRORS INTRODUCED IN SHORT-CIRCUIT STUDIES
12.47 kV Load Unbalance Effect
The ANSI/IEEE standards [2] and [3] and the IEC 909 [1]
Voltage Mag. Error Angle Error guide describe the procedures and recommended assumptions
Phase (pu) (%) (deg) (%) to compute short circuit currents. A comparative assessment
a 0.885 -5.2% -5.93 -2.2% of these two standards in terms of system modeling, database
requirements, AC and DC current decay, and duty types for
b 0.973 4.3% -123.48 0.1%
an industrial complex system is presented in [13].
c 0.946 1.4% 119.12 2.1% This paper is not intended to perform a comparison or to
Pos. Seq. 0.933 - -3.36 - provide a description of either standard, but rather to compare
TABLE XIII. the steady-state symmetric fault currents for error assessment
25.2 kV Load Unbalance Effect for the 5-node test case shown in Fig. 2. This test case is
comprised of three balanced loads, a 100 km 138 kV
Voltage Mag. Error Angle Error
transmission line, a delta-grounded wye transformer, and a 5
Phase (pu) (%) (deg) (%) km 12.5 kV distribution line.
a 0.893 -2.9% -4.55 -0.4% Source HV LV
b 0.928 1.0% -124.31 0.1% 138 kV 12.5 kV
G
c 0.937 2.0% 116.38 0.4% Load C
Pos. Seq. 0.919 - -4.14 - Load A Δ-Yg Load B
Fig. 2. 5-Node Test Case System
Neglecting load unbalances can introduce significant errors
A. Comparison of short-Circuit Methodologies
in voltage magnitudes and angles. However, an interesting
error compensation effect is observed when the load is ANSI/IEEE and IEC standard methodologies for fault
“properly unbalanced”. Load unbalance errors introduced in calculations present differences which are thoroughly
the node voltages may cancel errors due to the non- discussed in [13]. Some of the main differences are listed in
transposition. To illustrate this concept, consider the results of Table XVI [14]. The conventional methodology shown in the
Tables XIV and XV. These tables show the errors when the same table is described in most power system textbooks [9].
 6

For the purpose of this section, changes in generator and test case has been modified in order to include actual and
motor impedances, asymmetry, and fault impedances are more detailed generator data. Two steam generators (CF1-HP
neglected. All calculations are in steady-state. The fault and CF2-HP) and one hydro generator (H15), taken from
current obtained using the three calculation methods are [11], were modeled in the EMTP and standard power systems
compared in Table XVII for three-phase, single-phase-to- analysis software. GENROU and GENSAL models were
ground, phase-to-phase, and double-phase-to-ground faults used to represent steam and hydro generators respectively. A
applied at the receiving end bus (load C) of the system in Fig. SEXS type exciter was added to every generator in both
2. simulators. The generator data used in this test is presented in
TABLE XVI. Appendix A. Although saturation effects in machines could
Comparison Between Short-circuit Standard Methodologies be modeled, they were neglected for the purpose of this test.
ANSI/IEEE IEC Conv. The transmission system connecting the three generators is a
Adjust motor impedances Yes Yes No 230 kV system including 6 transmission lines, three step-up
Use load-flow voltages Yes No Yes transformers and three loads as shown in the model of Fig. 3.
Include pre-fault load current No No Yes The accurate model considers a three-phase load balanced
Include fault impedance No No Yes system, but includes electromagnetic mutual couplings of the
Shunt faults (LLL, LG, LL, LLG) Yes Yes Yes non-transposed transmission lines and generator stators. The
Series faults (open phases, etc.) No No Yes transmission line data, not included in this paper for the sake
Symmetrical rms current Yes Yes Yes of space, is available to the reader upon request.
Asymmetrical rms current Yes Yes No
B. Maximum Clearing Time (MCT) Comparison
B. Comparison against the Accurate Model
The accurate test case was compared with its equivalent
The 5-node test system with balanced loads shown in Fig. 2 positive-sequence model (PSM) available in standard
is used here for comparisons with accurate computations also software. The results for the MCT for a three-phase fault and
performed with phasors in steady-state. The results of the different integration time steps are presented in Table XVIII.
short-circuit calculations are shown in Table XVII which The accurate/unbalanced model considers a three-phase
reveals that significant errors can be found in some specific representation of machines and non-transposed transmission
cases. For example, in the three-phase fault case, the lines. In the accurate/Pos-seq model, the transmission lines
conventional, ANSI/IEEE and IEC methods introduce are represented by sequence parameters taken from the PSM
absolute value errors of up to 7.8%, 11.4% and 6.5% model in [11]. A permanent three-phase fault (see fault switch
respectively when compared to the accurate model. It must be SW in Fig. 3) is applied at BUS7 and cleared by opening the
noted that the accurate model computes different currents in line connecting BUS5 to BUS7 in Fig. 3.
each phase, thus reflecting the line unbalance effects. Single- The results show that the MCT computed by the accurate
phase-to-ground faults were applied to each phase. For the model using a time step of 0.1 ms is almost 14% greater than
case when the short-circuit is applied to phase B the current is the value provided by the PSM. Even though the time domain
6.0%, 7.8% and 2.7% smaller than the accurate value for the simulation provides a reasonable result for a time step of 1.0
conventional, ANSI/IEEE and IEC methods respectively. ms, the obtained MCT cannot be considered as a valid and
Although not negligible, the differences are smaller with the accurate value since time domain solutions rely on smaller
IEC method for most types of faults. Therefore, it can be time steps and the error in switching instants becomes
concluded that neglecting load currents may counteract the significant. On the other hand, the PSM provides the same
effect of considering 1.0 pu as pre-fault voltages in the IEC MCT of 268 ms for both time steps as shown in Table XVIII.
fault current calculation method. It can be also observed that 2.5% of the total error is due to
It is important to emphasize that for series compensated the balanced representation of the unbalanced and non-
lines, standard symmetrical components based computation of transposed transmission lines. The resultant 11% of difference
short-circuit currents can give completely wrong results. This is due to the machine model (full dq0 representation) and the
is due to the coupling between sequence networks caused by error accumulated by the phasor solution used in the PSM.
the nonlinear effects of the arrester protecting the It should be noted that in time domain simulations each
capacitor [15]. phase switch opening is delayed, consistent with actual switch
operation, until the phase current crosses zero. The full
V. ERRORS INTRODUCED IN TRANSIENT STABILITY STUDIES investigation of the source of the differences between
A. Transient Stability 9-Bus Test Case traditional transient stability simulations (using phasors) with
An improved model of the 9-bus test case presented in [11] the more accurate time-domain simulations will be part of a
is used to assess the errors introduced by positive-sequence sequel paper.
(or balanced) system representation, machine models, and
solution methods in transient stability studies. The presented
 7

TABLE XVII.
Comparison of Short-circuit Currents for Different Faults and Methodologies

Fault Phase Accurate Conventional ANSI/IEEE IEC

(A) (A) (%) (A) (%) (A) (%)
A 2,397 2,446 2.0% 2,350 -2.0% 2,481 3.5%
Three-phase B 2,653 2,446 -7.8% 2,350 -11.4% 2,481 -6.5%
C 2,334 2,446 4.8% 2,350 0.7% 2,481 6.3%
A 1,744 1,702 -2.4% 1,669 -4.3% 1,762 1.1%
Single-phase-to-ground B 1,810 1,702 -6.0% 1,669 -7.8% 1,762 -2.7%
C 1,761 1,702 -3.3% 1,669 -5.2% 1,762 0.1%
B 2,110 2,118 0.4% 2,035 -3.5% 2,148 1.8%
Phase-to-phase (B-C)
C 2,110 2,118 0.4% 2,035 -3.5% 2,148 1.8%
B 2,274 2,255 -0.8% 2,144 -5.7% 2,263 -0.5%
Double-phase-to-ground (B-C)
C 2,178 2,255 3.5% 2,144 -1.6% 2,263 3.9%

BUS2 BUS7 BUS8 BUS9 BUS3

1.03/_9.3 1.03/_3.7 1.02/_0.7 1.03/_2.0 1.03/_4.7

LF LF
1 2 2 1
out
out

AVR
LF
AVR

SM 18/230 13.8/230 SM
C_P100_Q35
in
in
+

SM3
SM2 13.8kV
18kV 128MVA
SW

192MVA PVbus:LF3
+

PVbus:LF2

BUS5
BUS6
1.00/_-4.0
+

1.01/_-3.7
LF
LF
A_P125_Q50 B_P90_Q30

BUS4
1.03/_-2.2
13.8/230
2
1

BUS1
1.04/_0.0
LF
SM1

SM

Slack:LF1
128MVA
13.8kV
out

in

AVR

Fig. 3. Modified 9-Bus Test Case System [11]

TABLE XVIII. VI. CONCLUSIONS

Comparison of MCT for a Three-phase Fault at Node 7 Assumptions in traditional load-flow studies are accurate for
Method Model Time step MCT Diff. balanced systems, but significant errors in node voltage
(ms) (ms) (%) magnitudes and phase angles are introduced when load
0.1 310 -13.5% unbalances or non-transposed lines are present. Errors in load-
Unbalanced
Accurate 1.0 304 -11.8% flow studies can be as large as 15%.
Pos-seq 1.0 & 0.1 301 -11.0% Short-circuit studies present diverse results depending on
PSM Pos-seq 1.0 & 0.1 268 - the standard used for fault current calculations. Errors in fault
currents may exceed 10% in some types of faults.
It has been found that important differences are also
introduced in transient stability studies for three-phase faults,
making conventional simulation tools more conservative than
 8

accurate simulations. Differences in the order of 14% were REFERENCES

observed for a 9-Bus test case studied when computing the [1] International Std. IEC 60909, “Short Circuit Current Calculation in
maximum clearing time. Although this error is substantial, it Three Phase AC Systems”, Geneva: International Electrotechnical
Commission, 1988.
is only valid for the studied test case and cannot be [2] IEEE Std. 551 “Recommended Practice for Calculating Short-Circuit
generalized to all transient stability studies and systems. Currents in Industrial and Commercial Power Systems (Violet Book)”,
Further research is needed to draw more general conclusions. 2006.
Power systems operating close to their maximum stability [3] ANSI/IEEE Std. 141, “IEEE Recommended practice for Electric Power
Distribution for Industrial Plants (Red Book)”, 1986.
limits and loading capabilities require more accurate modeling [4] J. Mahseredjian: “Simulation des transitoires électromagnétiques dans
in order to optimize the utilization of their transmission assets. les réseaux électriques”, Édition ‘Les Techniques de l'Ingénieur’,
As advanced simulation algorithms are very efficient, there Dossier n°D4130, 2008.
[5] J. Mahseredjian, S. Dennetière, L. Dubé, B. Khodabakhchian and L.
are no impediments for the power industry to incorporate Gérin-Lajoie, “On a new approach for the simulation of transients in
advanced models to perform various system studies and power systems”, Electric Power Systems Research, Volume 77, Issue
obtain more accurate results. 11, September 2007, pp. 1514-1520.
[6] [Link]
distribution/power-technologies-international/software-solutions/
APPENDIX [7] [Link]
[8] R. Horton, M. Halpin, K. Wallace, “Induced Voltage in Parallel
A. Generator Data for Transient Stability Test Transmission Lines Caused by electric Field Induction”, IEEE 11th
Generators Data GEN-1 GEN-2 GEN-3 Conference on Transmission & Distribution Construction, Operation
and Live-Line Maintenance, October 2006.
Gensal Genrou Genrou [9] P.M. Anderson, “Analysis of Faulted Power Systems”, IEEE Press
Type -
H15 CF2-HP CF1-HP power system engineering series, 1995.
[10] P. Kundur, “Power System Stability and Control”, EPRI Power System
BUS # 1 2 3
Engineering Series, McGraw-Hill, Inc., USA, 1994.
Rated (MVA) 158.0 192.0 128.0 [11] P.M. Anderson, A.A. Fouad, “Power System Control and Stability”,
Nominal (kV) 13.8 18.0 13.8 IEEE Press power system engineering series, Wiley-Interscience, 2003.
[12] IEEE Distribution System Analysis Subcommittee, “Radial Distribution
T'do (sec) 5.200 5.000 5.890
Test Feeders”, 2001,
T"do (sec) 0.029 0.043 0.034 [Link]
T'qo (sec) 0.000 1.500 0.600 [13] A. J. Rodolakis, “A Comparison of North American (ANSI) and
European (IEC) Fault Calculation Guidelines”, IEEE Transaction on
T"qo (sec) - 0.000 0.000 Industry Applications, Vol. 29, No. 3, May/June 1993, pp. 515-521.
(MW- [14] PSAF Version 3, “Short-circuit (ANSI & IEC), Arcflash, Cymbreak,
H 3.180 3.108 2.383
sec/MVA) User’s Guide and Reference Manual”, CYME International T&D Inc.,
Damp. (pu) 0.000 0.000 0.000 July 2006.
[15] J. Mahseredjian, P. J. Lagacé, S. Lefebvre, and A. Chartrand,
Xd (p.u.) 0.920 1.670 1.680 “Superposition technique for MOV-protected series capacitors in short-
Xq (pu) 0.510 1.640 1.610 circuit calculations”, IEEE Trans. on Power Delivery, July 1995, Vol.
X'd (pu) 0.300 0.315 0.232 10, no. 3, pp. 1394-1400.

X'q (pu) 0.000 0.958 0.320

BIOGRAPHIES
X"d (pu) 0.220 0.225 0.171
Jaime A. Peralta (S’07) was born in Santiago, Chile in 1969. He received his
Xl (pu) 0.130 0.186 0.095
[Link]. degree in electrical engineering from the University of Chile in 1994
X0 (pu) 0.120 0.101 0.101 and his [Link]. from École Polytechnique de Montréal in 2007. He is
X2 (pu) 0.255 0.224 0.171 currently pursuing his Ph.D. in electrical engineering at the École
R2 (pu) 0.045 0.028 0.026 Polytechnique de Montréal. He is presently working as Asset Management
Planning Leader at BC Hydro, British Columbia, Canada. Jaime is a
Ra (pu) 0.002 0.004 0.002 registered Professional Engineer in the province of British Columbia.

Exciter Data Francisco de León (S’86, M’92, SM’02) received the [Link]. and the [Link].
degrees in Electrical Engineering from the National Polytechnic Institute
GEN # 1, 2 & 3
(Mexico), in 1983 and 1986 respectively, and obtained his Ph.D. degree from
Type - SEXS the University of Toronto, Canada, in 1992. Currently, he is an associate
Ta/Tb - 0.1 professor in the Department of Electrical and Computer Engineering of
Tb (sec) 10 Polytechnic Institute of NYU in Brooklyn, NY. His main interest include the
K - 100 steady state and transient analysis of power systems, the electromagnetic and
Te (sec) 0.1 thermal design of machines, and the definitions of power in nonlinear and
Emin (pu) -7 unbalanced circuits.
Emax (pu) 7
Jean Mahseredjian (M’87, SM’08) graduated from École Polytechnique de
Montréal with [Link]. (1985) and Ph.D. (1991). From 1987 to 2004 he
worked at IREQ (Hydro-Québec) on research and development activities
related to the simulation and analysis of electromagnetic transients. In
December 2004 he joined the faculty of electrical engineering at École
Polytechnique de Montréal.