Paper accepted for presentation at 2003 IEEE Bologna Power Tech Conference, June 23th-26th, Bologna, Italy

Stochastic Assessment of Unbalanced Voltage

Dips in Large Transmission Systems
G. Olguin, Student Member, IEEE and M.H.J. Bollen, Senior Member, IEEE

AbstractThis paper focuses on the simulation of unbalanced

voltage dips in a large transmission system. The theoretical
background about symmetrical components and the impedance
matrix is presented. Equations for during fault voltage are
derived. Simulations of balanced and unbalanced dips are
performed in a large transmission system taking into account the
transformer effect. Results are presented in several ways
including graphs and tables.
Index Terms power quality, stochastic methods, voltage dip,
voltage sag.

I. NOMENCLATURE
a
kf

v : Phase voltage in a at node k during a fault at node f.

Superscripts b and c identify phase b and c respectively.
Superscripts p, z and n identify positive, negative and zero
sequence components.

zijp : General entry of the positive sequence impedance

matrix, Z. Superscript n and z indicate the negative and zero
sequence entries respectively.
a : Fortescue transformation a-operator, given by

a=e

2
3

II. INTRODUCTION

OLTAGE dips -or voltage sags- are short duration

reductions of rms voltage. Typically two characteristics
describe them as a single event: magnitude and duration. At
the system level an additional characteristic is needed:
frequency of occurrence. In this paper we study magnitude and
frequency of dips in a large transmission system. The main
cause of voltage dips is the occurrence of remote faults in
transmission and distribution systems. For the purpose of this
paper, the remaining voltage during the fault characterizes a
voltage dip.
Dips are the most important voltage disturbance, being
responsible for most of the financial losses in the industry due
The authors acknowledge the financial support of Energimyndigheten,
Elforsk, and ABB Corporate Research under the Elektra program.
The authors are with the Department of Electric Power Engineering,
Chalmers University of Technology, Gothenburg Sweden. SE-412 96
(fax:+46.31.7721633,
e-mail:gabriel.olguin@elteknik.chalmers.se
mathias.bollen@elteknik.chalmers.se)

0-7803-7967-5/03/$17.00 2003 IEEE

to power quality deficiencies. Efforts need to be made to

describe the performance of the network in terms of dips.
Voltage dips are of stochastic nature because of the presence
of several random factors in the dip generation process. Power
quality monitoring can be used to obtain the average power
quality of the network, but many monitors need to be installed
ideally one at each bus- and for a long period of time. The
alternative is stochastic assessment [1, 5].
Stochastic assessment of dips combines the response of the
network to faults with a known fault frequency. Stochastic data
is combined with deterministic results to obtain the expected
number of dips at load points. The stochastic data is given by
reliability data in terms of fault rate, while the during-fault
voltages are the deterministic results. During-fault voltages are
the result of a short circuit simulation for a given set of
parameters: pre-fault voltage, network configuration, fault
positions, generation scheduling, etc.
Two stochastic methods have been presented as suitable
tools for voltage dip frequency estimation. The method of
critical distance is a simple way to determine the severity and
frequency of dips in radial systems [2]. The second method is
called Fault Positions and it is a straightforward way to deal
with dips [3, 4, 5, 6].
This paper will concentrate on the calculation of duringfault voltages for non-symmetrical faults. The statistical
processing of the results is similar as for symmetrical faults
and described in other papers, e.g. [3] and [4]. The paper will
also discuss ways of presenting the results for unbalanced dips.
III. IMPEDANCE MATRIX AND SHORT CIRCUITS
Matrix calculations are an efficient tool for computer-based
analysis in meshed systems. The calculation of the during-fault
voltages balanced and unbalanced- is based on three
principles: Superposition theorem, the node impedance matrix
and symmetrical components.
A. Symmetrical faults
Consider a network with N+1 nodes and its impedance
matrix, Z. The reference node is named zero and is chosen to
be the common generator node. Among the remaining N nodes
are the created nodes on lines needed to simulate short-circuits
at the chosen fault positions. According to the superposition
theorem, the voltage at node k during the fault at node f is
given by (1), where vpref(k) is the pre-fault voltage at node k and
vkf is the voltage-variation at node k due to the fault. It should

be noted that we are dealing with complex variables, however

a simplified notation has been adopted.
(1)
+ v
v =v
kf

pref ( k )

kf

The voltage variation vkf can be calculated by means of

the impedance matrix Z. The impedance matrix contains -in its
diagonal- the driving-point impedance zii of every node with
respect to the reference node. The driving point impedance of
a node is the Thevenins equivalent impedance between it and
the reference node. Hence, the driving point impedance allows
finding the short circuit current at each bus. Z also contains the
transfer impedances zij between each bus of the system and
every other bus with respect to the reference bus. They can be
interpreted as the voltages that exist on each of the other buses,
with respect to the reference, when a bus of the system is
driven by a unity injection current. Hence, the transfer
impedance allows finding the voltage-variation at node k due
to a short-circuit current driven at node f.
During a three phase short circuit at node f, the current
injected to the node f is given by (2), where vpref(f) is the prefault voltage at the faulted node f, zff is the impedance seen
looking into the network at the faulted node f and the minus
sign is due to the direction of the current.
(2)
v

if =

pref ( f )

z ff

Once the short-circuit current is known the change in

voltage at any node k due to this current- can be easily
calculated by using the transfer impedance zkf.
(3)
v

vkf = z kf

pref ( f )

z ff

It should be noted that vkf vfk because zff zkk.

The during-fault voltages are found applying (1), this is
adding to (3) the pre-fault voltage as shown in (4).
(4)
v pref ( f )

vkf = v pref ( k ) zkf

z ff

In contrast to the admittance matrix Y that is sparse, Z is a

diagonal dominant full matrix for a connected network. This
means that every bus in the system is subjected to dips due to
faults everywhere in the network, however the voltage drop
depends on the transfer impedance between the load point and
the faulted bus. In general the transfer impedance decreases
with the distance between the faulted point and load bus.
Hence load points will not seriously be affected by faults
located far away in the system. The voltage variation also
depends on the driving impedance at the faulted bus. The
driving impedance determines the weakness of the bus. The
stronger the faulted bus the larger the voltage drop and the
larger the affected area.
B. Nonsymmetrical Faults
When dealing with nonsymmetrical faults, symmetrical
components are used. Impedance matrix of positive Zp,
negative Zn and zero sequence Zz need to be built [7]. In

transmission systems the positive and negative sequence

matrices are usually considered equal. This assumption holds
for any static circuit, such as transmission lines and
transformers. It is not fully valid for rotating machines.
However, since the influence of the rotating machines in the
impedance matrix is small, the positive and negative sequence
matrices are nearly equal. In our simulations we assume these
two matrices to be equal, however the equations for
nonsymmetrical faults have been written for the general case.
The method described in the previous section can again be
used to calculate the during fault voltages. It is advantageous
to determine the sequence voltage variation due to the
corresponding sequence current and transform back these to
phase components.
Single phase fault: Phase a to ground
For a phase to ground fault, the short circuit current is
found by connecting the sequence networks in series. The
current flowing through the three sequence networks is the
same. The injected current in the positive, negative and zero
sequence networks is given by (5). The superscripts p, n and z
over if and zff indicate the positive, negative and zero sequence
components. The superscript a identifies the phase.
(5)
va

i fp =

pref ( f )
z
n
ff
ff

z +z +z
p
ff

; i fp = i nf = i zf

To find the sequence-voltage changes vkf we use the

transfer impedance zkf of each sequence network. The positive
sequence-voltage change at node k due to the phase to ground
fault is given by (6). Similar equations can be derived for
voltage changes of negative and zero sequence.
(6)
va

vkfp = zkfp

pref ( f )
z
ff

z + z + z nff
p
ff

The pre-fault voltage at the load point k, vpref(k), contains

only positive sequence component and is equal to the pre-fault
voltage in phase a. The during-fault sequence-voltages at node
k are given by (7), (8) and (9).
(7)
va

v kfp = v apref ( k ) z kfp

v = z
n
kf

n
kf

v = z
z
kf

z
kf

pref ( f )

z + z zff + z nff
p
ff

v apref ( f )

(8)

z ffp + z zff + z nff

v apref ( f )

(9)

z ffp + z zff + z nff

Phase voltages are obtained from the sequence voltages.

From (9) it is clear that phase voltages at node k will not
z
contain zero sequence component if the transfer impedance z kf
is equal to zero. This is particularly true when the observation
bus and the fault point are at different sides of a transformer
with delta winding.
To get the phase voltages, we transform back to phase
components. Equations (10), (11) and (12) present the results,

where a is the Fortescue a-transformation factor. These

expressions can be simplified assuming positive sequence
impedance equal to the negative one. Also the zero sequence
component can be neglected in some cases.
(10)
( z z + z p + z n ) va

vkfa = v apref ( k )

v =v

b
pref ( k )

v =v

c
pref ( k )

b
kf

c
kf

kf

kf

kf

z +z +z
z
ff

p
ff

pref ( f )

n
ff

( zkfz + a 2 zkfp + a zkfn ) v apref ( f )

z zff + z ffp + z nff

(11)

( zkfz + a zkfp + a 2 zkfn ) v apref ( f )

(12)

z zff + z ffp + z nff

The zero-sequence voltage rarely propagates to the

equipment terminals. Even if it appears at the equipment
terminals, the majority of the equipment is not affected by it
because they are usually connected in delta or ungrounded
star.
In absence of zero sequence voltage and assuming Zp = Zn,
(13), (14) and (15) give the phase voltages at the load point
during the fault.
(13)
va

vkfa = v apref ( k ) zkfp

pref ( f )
z
ff
p
ff

z +

z
2

Two Phase Fault: Phases b-c

A similar approach can be used to derive equations for
unbalanced dips due to phase-to-phase faults. For a phase-tophase fault only the positive and negative sequence networks
take part in the analysis. The zero sequence current is null and
so is the zero sequence voltage. The positive sequence current
is equal in magnitude to the negative sequence current, but in
opposite direction. The positive sequence voltage-change is
given by (16) and is equal to the negative sequence, but of
contrary sign. The zero sequence voltage-change is null.
a
(16)
p
p v pref ( f )
p
n
vkf = zkf p
;

v
=

v
kf
kf
n

z ff + z ff

As the network is considered balanced before the fault, the

pre-fault voltages contain only positive sequence voltage. The
phase voltages during the fault are obtained by transforming
back to phase components.
(17)
(z n z p ) va

vkfa = v apref ( k ) +

(14)

a
1 p v pref ( f )
b
b
vkf = v pref ( k ) + zkf
zz
2
z ffp + ff
2
a
V
1
vkfc = v cpref ( k ) + zkfp pref ( f z)
z
2
z ffp + ff
2

v c pref

drop. These dips are known as dips type Da meaning that the
main drop is in phase a [8]. In Section IV we will see that the
same single-phase fault is seen differently when a delta-wye
transformer is in between the observation point and the faulted
node.

vkfb = v bpref ( k ) +
(15)

vkfc = v cpref ( k ) +

kf

kf

pref ( f )

z +z
p
ff

n
ff

( a 2 zkfp + a zkfn ) v apref ( f )

(18)

z ffp + z nff

( a zkfp + a 2 zkfn ) v apref ( f )

(19)

z ffp + z nff

If Zp = Zn holds, the previous equations can be simplified.

Figure 2 presents the phasor diagram for this kind of dips.
They are known as dips type Ca meaning that the main drop is
in phases b and c [8].

v c kf

v c pref
v a kf
v b pref

v a pref

v b kf

Fig. 1. Dip type D for Zp=Zn and zero-sequence transfer impedance null

Note that (13), (14) and (15) show that phase voltage at the
observation bus can be calculated using a balanced short
circuit algorithm by adding a fault impedance equal to half of
the value of the zero sequence driving point impedance at the
fault point. These expressions only hold when the zero
sequence voltage is not relevant. Figure 1 shows a phasor
diagram for this dip. The faulted phase shows the main voltage

v c kf
v

kf

v a kf

v a pref

v b pref
Fig. 1. Dips type C for Zp=Zn

Two Phases To Ground Fault: Phases b-c-ground

In a two phases to ground fault sequence networks are
connected in parallel. The sequence currents are all of different
values. At the load point will always exist voltage of positive

and negative sequence. The zero sequence voltage will be

present if there is a path for the zero sequence current to flow
between f and k. Equations for this kind of unbalanced dip can
be derived in the same way already described.
IV. EFFECTS OF POWER TRANSFORMERS
In the previous sections equations for balanced and
unbalanced dips have been derived. However the effect of
power transformers on the resulting dip has not been taken into
account. Power systems contain power transformers with many
different winding connections. As a voltage dip passes through
a transformer, the relations between the voltage phasors will
change compared with the primary side. Two effects can be
identified [8]:
Change of the symmetrical phase. A dip can change its
symmetrical phase because of the labeling of the phases at
the secondary side of the transformer. This change is not
considered in this paper because our interest is in the
expected number of dips without discrimination of phases.
Change of the magnitude and/or phase because of the
winding connections. Only delta-wye is considered in this
paper. Two effects are considered: zero sequence blocking
and phase rotation. Zero sequence blocking is already
taken into account by modeling the zero sequence
transfer-impendence.
To model the phase rotation and its effect on magnitude we
consider ANSI/IEEE Standard [9]. According to this standard,
the low voltage side, whether in wye or delta, has a phase shift
of 30o lagging with respect to the high voltage side phase-toneutral voltage vector. Hence in passing through the
transformer from the fault side to the observation side, the
positive-sequence-phase voltage is shifted 30o in one direction,
and the negative-sequence voltage is shifted in the other
direction. These phase shifts can be incorporated into the dip
equations to consider the effect of transformers. However,
from the point of view of dip characterization what it is
important is the complex during-fault voltage with respect to
the pre-event voltage at the same bus, see (1). At the
observation point the pre-fault phase voltage is of positive
sequence and is the reference for the dip at that point. Hence
the positive-sequence voltage does not need to be shifted, but
the negative one needs to be shifted 60o to take into account
the effect of the transformer.
The phase angle shift introduced by the transformer in the
negative sequence can be written in terms of the Fortescue atransformer operator.
(20)
+ 60 o = a 2
o
(21)
60 = a
Consider for instance a phase-to-phase fault involving phase b
and c in the high voltage (HV) side of a transmission system.
The dip caused by this fault is seen at the HV sector as a Ca
dip and it was described in (17), (18) and (19). In the HV
sector, there is no need to shift the negative sequence because
the fault point and the observation bus are located at the same
sector. The same fault causes a dip at low voltage (LV), but in
this case a phase shift of +60o needs to be introduced into the

negative sequence. Using basic properties of the Fortescue atransformer operator and expressing the pre-fault voltage of
phase a in terms of phase c we get (22), (23) and (24).
Comparing these equations with (10), (11) and (12) we
conclude that the dip Ca is seen at the other side of the
transformer as a Dc dip. The similarity can be made clearer
taking the positive and negative sequence impedances equal
and neglecting the zero sequence component. In this case, the
phasors diagram of figure 1 is applicable with phase c as the
symmetrical phase.
(22)
(a z kfn + a 2 z kfp ) v cpref ( f )
a
a
vkf ( LV HV ) = v pref ( k )
p
n
z ff + z ff

vkfb ( LV HV ) = v bpref ( k )

vkfc ( LV HV ) = v cpref ( k )

(a z kfp + a 2 z kfn ) v cpref ( f )

z +z
p
ff

( z + z ) v cpref ( f )
p
kf

(23)

n
ff

n
kf

(24)

z ffp + z nff

Note that the previous analysis shows that a Ca dip at the

HV sector is seen as a Dc dip at the LV, but also that a Dc dip
at the LV sector is seen as a Ca dip at the HV sector. A similar
analysis shows that a Da dip at the LV sector is seen as a dip
Cb at the HV sector. Although the previous analysis allows
identifying the phases involved in the dip type, in practice the
phases involved depends on the re-labeling of phases. Hence
what can be concluded is that delta-wye transformers change
the dip type from D to C and vise versa.
These equations have been implemented in a simulation
tool in order to perform a stochastic assessment of balanced
and unbalanced voltage dips.
V. STOCHASTIC ASSESSMENT OF VOLTAGE DIPS
The main purpose of the stochastic assessment of voltage
dips is to get a clear picture of the expected number of dips
and their characteristics as duration, magnitude, phases
involved, etc. In this work we are mainly interested in
magnitude. The magnitude of a dip is given by the remaining
voltage during the fault and has already been presented for
balanced and unbalanced faults. Once the magnitude of the
potential voltage dips has been calculated the stochastic
assessment can be performed combining the deterministic
results of the short-circuits simulations with the stochastic data
regarding faults.
How often a particular voltage dip occurs at a given
location depends on several factors, the number of faults
occurring in the electrical neighbourhood being the most
important one. This electrical neighbourhood has been
already identified as the exposed area [3,4]. How often a dip
occurs at bus k due to faults at a given bus f depends on the
reliability of the bus f. If the fault rate of bus f is , then the dip
caused by this fault will be seen times per year. Lines present
a special problem because the during-fault-voltage depends
upon the fault location. Some fault positions are needed along
the lines in order to get an accurate evaluation of the
magnitude of the dip caused by faults on lines.

12

73

76

79

30

23

51

21

28

77
80

86

24

62
63

52

16
85

84

14

22

19

29

47

11

35

78

26

65

10
43

54
2

81

32

59

49

53

20

46

60

25

8
44

71

75

48

15

68

31

56

74

67
61

27
38

40

50

69
45

17
41

13

82

70
33

87

34

39

18

58

37
72

55

57

83

42

9
64
66
36

Fig. 1.Simplified Model of the National Interconnected Systems of Colombia, buses 15, 16, 63 and 68 are 500 kV buses.

Suppose that after simulating short circuits on buses and

along lines the resulting dip magnitudes at load points are
stored in a during-fault voltage matrix Vdfv. A particular entry
of this matrix would be an element vkf indicating the remaining
phase voltage at bus k during a fault at node f. Combining this
matrix with the fault rate of fault positions, the stochastic
assessment of voltage dips can be performed.
The matrix Vdvf contains relevant information. The exposed
area can be derived from the Vdfv matrix. This information is
contained in rows of Vdfv. The exposed area encloses the
observation bus and the network area in which the occurrence
of faults would cause a during fault voltage lower than a given
value at the observation bus.
VI. DESCRIPTION OF THE SYSTEM AND SIMULATIONS
A. System
The studied system is a simplified model of the National
Interconnected system of Colombia and has been described in
[3]. It contains 87 buses and 164 lines with a total length of
11651 km. A double-circuit 500kV line connects two 230kV
grids. See Figure 1.
The objective of this study is to show the possibilities of the
stochastic assessment. However the results cannot be directly

applied to this system because some assumptions have been

necessary to complete the simulations.
B. Modeling and Simulations
The system was modeled by means of the positive and zero
sequence impedance matrices. Negative sequence impedance
matrix was taken equal to the positive one. Lines were
modeled as medium-length lines perfectly transposed. Lines
shunt parameters are considered. Independence of sequence
components is assumed, meaning that the different sequences
do not react upon each other.
The simulations consider 781 fault positions with 87 faults
at buses and 694 faults at lines. Lines were divided so that
each line segment was no longer than 15 km. Fault rates were
assigned to fault positions corresponding to the type of fault.
The fault rate for fault positions on lines was determined from
the quotient between the expected annual number of faults on
lines and the number of fault positions on lines.
In order to count the number of dips occurring at the load
positions, the following criterion was adopted: each faultoriginated event, balanced or unbalanced, was counted as one
single dip. The minimum phase-to-neutral voltage was used to
characterize the resulting dip. During fault voltages were
registered for all buses and for each simulated fault. Three
simulation cases were developed as explained.

 Case A. This simulation only considers three phase faults.

The fault rate for lines is 0.0134 faults per year-km and is
the same for the 164 lines. The fault rate for buses is 0.08
faults per bus-year and is the same for the 87 buses.
Case B. Only single phase to ground faults were
simulated. Fault rates as in case A.
Case C. Four kinds of shunt faults were simulated: singlephase to ground, phase to phase, two phases to ground and
three-phase fault. The probabilities of these faults were
assumed to be 80% single phase, 5% two phases, 10% two
phases ground and 5% three phases. Fault rates as in Case
A.

Figure 2 presents the 85% exposed areas for buses 5, 15

and 87. The lowest of the phase to ground voltages is used to
characterize the event. Faults on buses or lines inside the areas
would cause a remaining voltage in the worst phase lower than
85% at the indicated buses. It should be noted that part of the
lines connecting buses inside the exposed areas might be
actually outside the exposed areas. Also note that the system
has been modeled as solidly grounded. Dashed line is used for
the single-phase fault exposed area, while solid line is used for
the three-phase exposed area. The single-phase fault exposed
areas for buses 5 and 87 are almost coincident with the threephase fault exposed area, however a bit smaller. The threephase fault exposed area is bigger than the single-phase fault
exposed area but they do not differ too much when there is no
power transformer inside the area. When the exposed area
contains power transformers its shape changes drastically. This
can be explained by the blocking of the zero sequence
component and the changing in the dip type due to the winding
connection of the transformers.
From Figure 2 we can see that equipments will be less
exposed to spurious trip due to unbalanced dips. For an
impedance-grounded system the single-phase fault exposed
area would be significantly smaller.

C. During Fault Voltages, Exposed Areas

The simulations give the during fault voltages for all buses.
However, only part of the results is presented in this paper due
to space limitations. A useful way to present the effect of faults
around the system on a given bus is by using the exposed area.
To derive conclusions about the exposed area for symmetrical
and nonsymmetrical faults; we present these areas for both
cases A and B.

12

73

76

79

30

23

51

21

28

77
80

86

24

62
63

52

16

85

84

14

20

46

47

11

35

78

68

31

56

74

67

61

27
38

40

65

10

43

54
26

81

32

59

49

71

75

60

25

8
44
53

48

29

22

19

15
15

50

69
45

17
41

13

82

70

87
87
1

33
34

39

18

58

37
72

55

57

83

42

9
64
66
36

Fig. 2. Exposed areas (85%) for buses 5, 15 and 87. Dashed line indicates exposed area for single-phase fault

55

The exposed areas presented in Figure 2 have been drawn

as closed sets to simplify the drawing, however they might be
isolated sets enclosing the parts of the system in which a fault
would cause a dip more severe than a given value. To illustrate
this, consider the lines connecting bus 87 and 27. Figure 3
shows the during fault voltage at bus 87 when a moving fault is
applied on the line connecting these buses. Single-phase fault
and three-phase faults are shown. It can be seen that for both
fault types part of the line does not pertain to the 85% exposed
area of bus 87. The term critical distance is used to describe
the length line exposed to faults that might lead to critical dips
[1,2]. For this line and three-phase faults, the critical 85%distance would be approximately the first 40% of line plus the
ending 10% of it.
Dip at bus 87 due to faults on line 87-27
1phase fault

3 phase fault

1.0
0.9

Magnitude pu

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

20

40

60

80

100

to a three-phase fault at the same position.

D. Dips Frequencies
Voltage dips are power quality events that might stop an
entire industrial process. Duration and retained voltage are
used to characterize dips. In this paper only magnitude of the
dip is studied, however as dips have been simulated for 230kV
and 500kV we would expect durations of less than 500 ms. In
order to characterize an individual bus of the system the
frequency of the event is what matters because the main
objective is to evaluate the risk associated to spurious trips of
sensitive processes connected at that bus. We now present the
expected number of dips for some selected buses.
TABLE I
CUMULATIVE DIP FREQUENCY AT BUS 87

Remaining
voltage vkf pu
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90

% line lenght

TABLE II
CUMULATIVE DIP FREQUENCY AT BUS 5

Fig. 3. During fault voltage at bus 87 due to faults on line 87-27

Figure 3 corresponds to a line connecting the load bus. The

data contained in the Vdfv matrix also allows the description of
the during fault voltage due to faults on lines not connecting
the load bus.
Consider the line connecting nodes 62 and 76. According to
the exposed areas shown in Figure 2, only part of this line is
inside the 85% exposed area for single-phase fault. This is
confirmed by Figure 4, which shows the resulting remaining
voltage for faults along line 76-62. Figure 4 also shows the
effect of the zero sequence blocking by the transformers.
Dip at bus 15 due to faults on line 76-62
1 phase fault

Magnitude pu

3 phase fault

1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0

20

40

60

80

Cumulative Dip Frequency at Bus 87

3 ph
1 ph
2 ph
2 ph-g
Total
0.04
0.13
0.00
0.04
0.21
0.13
0.01
0.11
0.34
0.10
0.12
0.67
0.02
0.21
1.02
0.16
1.70
0.10
0.26
2.22
0.28
2.06
0.12
0.39
2.85
0.45
3.20
0.23
0.62
4.50
0.82
5.55
0.46
1.44
8.26
1.28
14.02
0.99
2.33
18.62

100

% line lenght

Fig. 4. During fault voltage at bus 15 due to faults on line 76-62

Both Figure 3 and 4 confirm the fact that single-phase faults

are less severe than three-phases faults, in the sense that the
worst phase shows a higher remaining voltage than the dip due

Remaining
voltage vkf pu
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90

Cumulative Dip Frequency at Bus 5

3 ph
1 ph
2 ph
2 ph-g
Total
0.39
6.03
0.25
0.78
7.4
0.45
6.99
0.29
0.90
8.6
0.54
8.02
0.33
1.07
10.0
0.60
9.16
0.41
1.19
11.4
0.68
10.79
0.54
1.36
13.4
0.83
13.08
0.64
1.64
16.2
0.98
14.16
0.82
1.91
17.9
1.07
15.60
1.00
2.09
19.8

Figures 5 to 7 show the cumulative number of dips at buses

87, 15 and 5 due to balanced and unbalanced faults. As it
would be expected the contribution of single-phase faults is
dominant. However the contribution of the different type of
faults does not exactly follow the distribution probability of
faults. The exposed area again can explain this. As the exposed
area for three phase faults is larger than for any other fault the
resulting contribution of this faults to the total frequency is
more than the 5% that could be expected. Table I shows that,
for instance for bus 87, the three-phase faults contribute with
more than 29% of the total dips more severe than 0.6 pu. It
should be noted that bus 87 has several lines converging to that
bus. For buses at the end of a line (radial source) the
distribution probability of the faults is a reasonable indicator
of the contribution of unbalanced dips to the total number of
dips. Table II shows the contribution of the different faults to
the total number of dips at bus 5. Bus 5 is fed from the system

through one single line. The contribution follows the

distribution probability of faults.

Dips at bus 87

Events/year

3 ph

VIII. REFERENCES

2 ph-g

2 ph

1 ph

area for balanced dips. The contribution of balanced and

unbalanced faults to the dip frequency was also studied.

[1]

40
35
30
25
20
15
10
5
0

[2]
[3]

[4]
0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

V pu

[5]

Fig. 5. Cumulative Dip Frequency at bus 87

[6]

Dips at bus 15

Events/year

3 ph

1 ph

2 ph-g

2 ph

[7]

35
30
25
20
15
10
5
0

[8]
[9]

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

Bollen, M. H. J. Understanding Power Quality Problems: voltage sags

and interruptions , IEEE Press Series on Power Engineering, New York
2000.
Olguin, G. and Bollen, M. H. J., Stochastic Prediction of Voltage Sags:
an Overview , Probabilistic Methods Applied to Power Systems
Conference 2002, September 22-26, Naples Italy.
Olguin, G. and Bollen, M. H. J. The Method of Fault Position for
Stochastic Prediction of Voltage Sags: A Case Study , Probabilistic
Methods Applied to Power Systems Conference 2002, September 22-26,
Naples Italy.
Qader, M. R.; Bollen, M. H. J.; Allan, R. N. Stochastic prediction of
voltage sags in a large transmission system , Industrial and Commercial
Power Systems Technical Conference, IEEE , 1998 Page(s): 8 18
Conrad, L. Grigg, C. and Little, K., Predicting and Preventing
Problems Associated with Remote Fault Clearing Voltage Dips, IEEE
Transactions on Industry Applications, vol. 27, no. 1, pp. 167-172,
Jan/Feb 1991.
Heine, P. and Lehtonen, M., Influence of Subtransmission System
Characteristics on Voltage Sags , 10th International Conference on
Harmonics and Quality of Power, 6-9 October 2002, Rio de Janeiro,
Brazil.
Anderson, P. M., Analysis of Faulted Power Systems , The IOWA
State University Press 1973. First edition, 1973.
Zhang, Lidong Three-phase Unbalanced of Voltage Dips , Licentiate
Dissertation, Department of Electric Power Engineering, Chalmers
University of Technology, Gothenburg, Sweden, 1999.
ANSI C57.21.10.1988 American National Standard For Transformers
230kV and Below 833/958 through 8333/10417 kVA, Single-Phase,
and 750/862 through 60000/80000/100000 kVA, Three-Phase without
Load Tap changing; and 3750/4687 through 60000/80000/100000 kVA
with Load Tap Changing Safety Requirements .

V pu

IX. BIOGRAPHIES

Fig. 6. Cumulative Dip Frequency at bus 15

Dips at bus 5

Events/year

3 ph

2 ph

1 ph

Gabriel Olguin received his BSc in Electrical Engineering from University

of Santiago, Chile in 1994, MBA from University of La Serena, Chile and
MSc in Power Engineering from Federal University of Santa Catarina in
Florianpolis, Brazil in 1999. He was a planning engineer at EMEC Co. in
Coquimbo, Chile, where he leaded several studies in the planning and
distribution tariff setting areas of distribution systems. He is currently a
research assistant in the Department of Electric Power Engineering at
Chalmers University of Technology, Gothenburg, Sweden, working towards
his Ph.D. degree. His research interests are stochastic and statistics methods
applied to voltage sags studies.

2 ph-g

40
35
30
25
20
15
10
5
0
0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

V pu
Fig. 7. Cumulative Dip Frequency at bus 15

VII. CONCLUSIONS
Simulations of balanced and unbalanced dips have been
performed in a big transmission system. The effect of power
transformers on the voltages has been modeled. It was found
that for a solidly grounded system and in absent of power
transformers the exposed areas of dips originated by singlephase faults is similar but slightly smaller than the exposed

Math H. J. Bollen is professor in electric power systems in the Department

of Electric Power Engineering at Chalmers University of Technology,
Gothenburg, Sweden. He received the MSc and Ph.D. degrees from
Eindhoven University of Technology, Eindhoven, The Netherlands, in 1985
and 1989, respectively. Before joining Chalmers in 1996 he was a research
associate at Eindhoven University of Technology from 1989 to 1993, and
lecturer at University of Manchester Institute of Science and Technology
between 1993 and 1996. His research interests covers various aspects of
power quality and reliability. He has published a number of fundamental
papers on voltage dip analysis and a textbook on power quality. Math Bollen
is active in several IEEE working groups on power quality.