Cac Bai Bao Khoa Hoc 2007-2012 (Phan II CUNG CAP DIEN)

eBook for You
I HC BCH KHOA H NI
B MN H THNG IN
Tng hp cc bi bo khoa hc giai on 2007-2012
PHN II
eBook for You
CUNG CP IN
B mn H thng in - i hc Bch Khoa H Ni
82

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 1, JANUARY 2008
347
Fault Distribution Modeling Using Stochastic

Bivariate Models for Prediction of Voltage
Sag in Distribution Systems
AbstractThis paper presents a new method regarding fault distribution modeling for the stochastic prediction study of voltage
sags in the distribution system. 2-D stochastic models for fault modeling make it possible to obtain the fault performance for the whole
system of interest, which helps to obtain not only sag performance
at individual locations but also system sag performance through
system indices of voltage sag. By using the bivariate normal distribution for fault distribution modeling, this paper estimates the
inuence of model parameters on system voltage sag performance.
The paper also develops the modied SARFIX regarding phase
loads that create better estimation for voltage sag performance for
the distribution system.
Index TermsBivariate normal distribution, distribution
system, fault distribution modeling, phase loads, power quality
(PQ), stochastic prediction, voltage sag frequency.
I. INTRODUCTION
MONG power-quality (PQ) phenomena, the voltage sag
(dip) is dened in IEEE1159, 1995 as a decrease in rms
voltage to between 0.1 and 0.9 of the nominal voltage at the
power frequency for the duration of 0.5 cycle to 1 min. There has
been a greater interest in voltage sags recently due to problems
caused by the performance of sensitive electronic equipment
that is widely used.
Research about the voltage sag is usually related to a basic
process known as a compatibility assessment [1], [2] which
includes three steps.
Step 1) Obtain the voltage sag performance of the system of
interest.
Step 2) Obtain equipment voltage tolerance.
Step 3) Compare equipment voltage tolerance with the
voltage sag performance and estimate the expected
impacts of the voltage sag on the equipment.
Current research has shown evidence that obtaining the
voltage sag performance still needs more improvement. The
Manuscript received August 2, 2005; revised December 5, 2006. This work

was supported by the Korea Foundation for Advanced Studies International
Scholar Exchange Fellowship for the academic year of 20042005. Paper no.
TPWRD-00456-2005.
B. Q. Khanh is with the Electric Power System Department, Faculty of Electrical Engineering, Hanoi University of Technology, Hanoi, Vietnam (e-mail:
bq_khanh-htd@mail.hut.edu.vn).
D.-J. Won is with the School of Electrical Engineering, INHA University,
Incheon 402751, Korea (e-mail: djwon@inha.ac.kr).
S.-I. Moon is with the School of Electrical Engineering and Computer Science, Seoul National University, Seoul 151-742, Korea (e-mail: moonsi@plaza.
snu.ac.kr).
Digital Object Identier 10.1109/TPWRD.2007.905817
information about the voltage sag is mainly obtained by

monitoring and stochastic prediction. With recently advanced
computer-aided simulation tools, the stochastic prediction of
voltage sag becomes the preferable approach that can obtain
the results at required accuracy for various network topologies
and operational conditions. The method of fault positions
and the method of critical distances are known as the most
widely used methods for stochastic prediction studies.
It is notable that regardless of which method is used, a stochastic prediction study always has to solve two critical problems: 1) the modeling of causes leading to voltage sags and
2) the simulation of the power system for computing voltage sag
characteristics. Among important cause of voltage sags, shortcircuit faults in the power system account for the largest part and
the assessment of the voltage sag performance based on fault
distribution modeling is a well-known approach. However, it is
very difcult to build up accurate fault modeling because the
data of faults can only obtained by monitoring and, thus, it has
the same uncertainties as to what the monitoring of voltage sags
can generate.
This paper presents a new approach on fault distribution modeling for the stochastic prediction of voltage sags in the distribution system using the method of fault positions. The simulation of the distribution system and fault distribution modeling
are made on MATLAB for computing not only site indices, but
also system indices of voltage sags.
II. FAULT DISTRIBUTION MODELING
Modeling the fault distribution is to determine the short-circuit fault frequency (i.e., fault rate or the number of short-circuit
faults per year) for all fault types at all possible fault positions
throughout the system of interest. It consists of the selection of
fault position and fault type and the distribution of fault rate for
the selected fault positions and fault types.
Fault positions are generally chosen in a way that a fault position should represent short-circuit faults leading to sags with
similar characteristics [2]. For the distribution system with typical radial network topology, small line segments, and distribution transformers along the trunk feeders, it is possible to apply
only one fault position for each distribution transformer and one
fault position for each line segment.
Different fault types should be applied to each fault position
mainly depending on the number of phases available at the selected fault positions. The fault rate of each fault type is normally referred from the observed historical data.
0885-8977/$25.00 2007 IEEE
83
eBook for You
Bach Quoc Khanh, Dong-Jun Won, Member, IEEE, and Seung-Il Moon, Member, IEEE

The fault rate mainly depends on fault position, fault type,

and fault cause. While two earlier factors have been discussed
at length in past research, the distribution of the fault rate for
the selected fault positions has received less interest. The most
common assumption that has been argued so far is that because
the fault can occur anywhere in the system, stochastically, it is
possible to model the fault rate as the uniform distribution [3],
[4]. In this sense, the fault rate at each position is identical to
the component failure rate that is based on component reliability. However, in reality, many factors can lead to faults, not
just the component failure, and fault rates at different positions
in the system are rarely the same. Recently, a report [5] proposed some interesting 1-D models of fault distribution along
individual line segments (between two nodes). However, this research could not consider the distribution of transformer faults.
Furthermore, by using 1-D fault distribution, it is hard to obtain a system index about voltage sag performance since there
are plenty of line segments in the distribution system. The new
method of fault distribution modeling proposed by this paper
carefully analyzes concerned fault causes and builds up a suitable modeling of the fault distribution for the whole system of
interest from which system indices can be obtained.
III. NEW FAULT DISTRIBUTION MODELING BASED
ON FAULT CAUSES FOR DISTRIBUTION SYSTEMS
Although there are a variety of causes that result in faults in
distribution systems, it is possible to group them into two parts:
namely 1) equipment failures and 2) external causes.
Equipment failure is basically due to defects that are probably created during manufacture, transportation, and installation. Equipment failure depends on the time of being placed
into operation, the aging period, and maintenance conditions.
According to the reliability theory, it is often characterized by
the component failure rate. There are several distribution functions to model this parameter but the most common one is the
exponential distribution which assumes the component failure
rate to be constant. This value is equal to the average failure
rate during the useful life of the bathtub curve [6]. Therefore,
if the same type of equipment is used throughout the system
(e.g., the same type of distribution transformers used in the distribution system), it is possible to assume that the failure rate
of equipment follows the uniform distribution depending on the
equipment type although it still may cause some errors (e.g., not
all equipment is put into operation at the same time or has the
same maintenance conditions).
Besides equipment failure, there are many other causes from
the ambient environment that also may lead to faults in power
systems. This paper calls them the external causes. Some can inuence the fault performance of the power system in a large area
such as severe weather (wind storms, lightning, etc.). Meanwhile, others mainly have local impacts, such as trees and animals (birds, mice, etc.). Human factors (scheduled interruption,
human errors, mischief, and vandalism) can cause faults that
only inuence the power system in small parts as well as severe faults for a large power system. All of these causes occur
randomly and they can be simulated by stochastic models. 1-D
stochastic
models seem to not be suitable as explained before.
Fig. 1. Example of bivariate normal distribution.
This paper proposes the idea of using 2-D stochastic models instead (e.g., the bivariate normal distribution model as illustrated
in Fig. 1).
For large power systems, it is hard to obtain a converged 2-D
fault distribution model for various causes in a large area. However, for small-to-medium-size networks, such as the section of
distribution network fed from a bulk-point distribution substation, of which the monitored historical data of fault performance
shows that faults due to external causes occur concentratively on
one location (e.g., some lines pass through a small area which
is at high risk for faults due to industrial pollution or trunk fall),
it is the favorite condition to obtain a converged 2-D fault distribution model.
IV. PROBLEM DEFINITION AND SOLUTION
A. Case Study Denition
To illustrate the new method of fault distribution modeling
in the stochastic prediction of voltage sag in the distribution
system, this paper uses the IEEE 123-bus radial distribution
feeder [7] as the test system. It can be seen as the distribution
system is fed from a bulk point. It does not narrow the scope of
application of the study with the following assumptions.
Since line segments in the test system come in one, two,
and three phases, distribution transformers at load nodes
are the single phase type for separate single-phase loads.
For three-phase loads, the connection of the distribution transformer is 4.16-kV grounded wyelow-voltage
grounded wye.
Voltage sags are only caused by faults in the test system.
If the test system is supposedly a section of a large distribution system, only faults occurring in it are considered.
The faults in sections fed from other distribution substations can be skipped as the transformer impedance in distribution substations, in reality, is rather high. Similarly, the
faults in low-voltage networks are also ignored because of
the large impedance of distribution transformers. This assumption only neglects voltage sags caused by faults in the
transmission system. It will be considered if the stochastic
prediction of voltage sag in large transmission systems [4]
is included.
In terms of reliability, the test system is modeled on two
main components: lines and distribution transformers. The
reliability of any other distribution equipment is supposedly included in the reliability of these two components.
The fault positions are selected as mentioned in Part II. For
transformers, one fault position at each load node (i.e., the
nodes connected with distribution transformers) is applied.
84
eBook for You
348

KHANH et al.: FAULT DISTRIBUTION MODELING USING STOCHASTIC BIVARIATE MODELS
B. Fault-Rate Modeling
Faults are random events and as previously indicated, they
can be simulated by stochastic distribution models. According
to the analysis in Section III, the fault rate of each fault type at
each fault position is equal to the sum of equipment failure rate
and fault rate due to external causes. The equipment failure rate
is supposed to follow the uniform distribution model. Therefore,
for the fault position of the transformer , the failure rate is calculated as follows:
tion because it is the most common stochastic model which has

such critical advantages as it accepts continuous variables and is
easy to build up the distribution based on monitored historical
data. Besides that, it is also simple to convert to other models
using continuous variables.
So the fault rate at each fault position is as follows.
For the transformer
(3)
For the line segment
(4)
where
contributory percentage of faults due
to external causes
;
,
weighted factors of the fault rates of

the transformer and the line segment
that follow the bivariate normal
distribution model depending on fault
positions.
The joint probability density function of bivariate normal distribution is expressed as follows:
where
(5)
(1)
where
,
number of transformer faults of the test system;
total distribution transformers;
contributory percentage of equipment failure.
means and standard deviations of two

variables , ;
correlation coefcient. If the
coordinates of fault positions are
independent variables
.
The line failure rate is normally expressed in the number of

faults per year per foot (or meter) length. However, because of
the short length of line segments, the line failure rate is calculated for the whole line segment as follows:
(2)
where
number of line faults of the test system;
total line segments;
length of the line segment (in feet).
The distribution of the fault rate due to external causes depending on fault positions is supposedly in compliance with the
2-D stochastic
model. This paper uses bivariate normal distribuB mn H thng in - i hc Bch Khoa H Ni
The probability for a fault to occur at the fault position

within an area
can be calculated as follows:
(6)
and is large enough,

If
then the distribution is normalized as follows:
(7)
For the distribution system, geographically, if network nodes

are disposed relatively uniform, it will be possible to apply
the
85
eBook for You
For lines, one fault position is also applied for each line
segment. Due to the short line segments, this paper selects
the fault position at the end of each line segment (For the
test system, there are 122 line segments and 87 load nodes.
Therefore, 209 fault positions in total are selected).
Fault types (single phase to ground, phase to phase, two
phases to ground, and three phases to ground) are applied
to fault positions depending on the number of available
phases. The fault impedance is assumed to be negligible.
The fault rate of a distribution transformer is a random
variable depending on the position of the load node it is
connected to. The fault rate of a line segment is also a
random variable depending on the fault position and the
length of this line segment.
Based on the previous denitions and assumptions, the computation of voltage sags at all load nodes on the primary side
of distribution transformers throughout the test system is performed on MATLAB [8]. The voltage sag frequency at each load
node is obtained when applying the fault rate to each fault position. The fault rates at the fault positions are calculated based on
the new fault distribution modeling presented in Part B. Finally,
related voltage sag indices are calculated.
349

350
following approximation where and

the fault position .
Faults rate for the transformer
are the coordinates of
TABLE I
SYSTEM FAULT-RATE BREAKDOWN
(8)
Fault rate for the line segment

(9)
PQ indices are used to estimate the quality of supplied electric energy for the power system. To date, many PQ indices
have been proposed for various PQ events. A well-known index
of voltage sag is the system average rms voltage variation frequency index for voltage sag down to under X% of the nominal
voltage value
. It is often used for evaluating the PQ
of a three-phase power system based on monitored limited segmentation [3]. The assessed system is segmented so that every
point in the system is contained within a section monitored by
an actual PQ measuring instrument.
In distribution systems, because various phase loads (phase
to neutral, phase to phase, three-phase loads) are available,
asymmetrical faults, which account for most faults, never result
in voltage sags to all single-phase loads (e.g, phase A-to-ground
faults may not cause voltage sags to the loads connected between
phase B and neutral or phase C and neutral or loads connected
between phase B and phase C). Therefore, using
regardless of the number of phases involved, may not exactly
reect the voltage sag performance of the distribution system.
From the demand sides, the indices are more interesting because
they can estimate the voltage sag performance for phase loads.
In order to take the availability of various phase loads in the
distribution system into account, this paper newly develops
in regard to phase loads as follows:
(10)
(11)
(12)
where
,
number of sags down to under

X% that phase-to-neutral
(A,B,C), phase-to-phase (A-B,
B-C, C-A), or three-phase load
experiences;
number of phase-to-neutral
(A,B,C), phase-to-phase (A-B,
B-C, C-A), or three-phase
customers served from the
system of interest.
Fig. 2. Mapping of the IEEE123-bus radial distribution test feeder.
V. RESULT DEMONSTRATION AND ANALYSIS

A. Procedures of Stochastic Prediction
The process of stochastic prediction study is performed
through the following steps.
First, the system fault rate (the total of faults occurring in the
test system over a certain period of time) is assumed to be an
arbitrary number, say 500 faults. This value is just for calculation and easier graphic demonstration of the results. Besides
that, contributory percentages of different fault types are also
assumed as follows:
single phase to ground (N1): 80%;
two phase to ground (N11): 10%;
two phase together (N2): 8%;
three phase to ground (N3): 2%
and the component fault rates are supposed to be
transformer: 50%;
line: 50%.
The listed percentages shown are, in fact, based on actual survey
data [9]. Based on the aforementioned assumptions, the system
fault rates of transformers and lines for different fault types due
to different fault causes (equipment failure or external causes)
are calculated and shown in columns 2 and 3 of Table I. Parameters ( ,
) that are included make it possible to consider
the inuence of fault causes due to external factors.
Second, the fault rate of each fault type is calculated for each
fault position using the fault distribution models as stated
in
86
eBook for You
C. Development of Voltage Sag Indices

Fig. 3. Sag frequency spectrum and SARFI of different phase loads for the case the mean value is at node 13 and deviation = = 0:51
B. Evaluation of Inuences of the Fault Distribution Modeling

on the Voltage Sag Performance
The fault distribution modeling uses several parameters. In
practice, it is possible to adjust these parameters so that the resulting model is suitable for the fault performance of the distribution system of interest. However, the variation of these parameters also makes the voltage sag performance change accordingly. In modeling fault distribution, this paper also considers
the following options of fault distribution for estimating the inuences of fault distribution on voltage sag performance.
Change contributory percentages of the fault due to external causes (change
or
). In this paper, three options
, 50%, and 100% are considered.
Switch the position of the mean value ( , ) of the bivariate normal distribution. This paper considers four options of the mean value at nodes 13, 51, 67, and 85 as indicated in Fig. 2.
Vary the deviations ,
of the bivariate normal distribution. This paper also considers the options of the deviation that are equal to 0.2, 0.5, and 0.8 of the maximum
value among deviations
.
C. Results Analysis
Based on aforementioned procedures of stochastic prediction,
the following are remarkable results.
In Fig. 3, the indices of voltage sag for different phase loads,
including voltage sag frequency spectrums, corresponding
,
, and
for X ranging
from 10% to 90% of the nominal voltage are shown. In this
case study,
,
.
Besides that,
for the whole test system for different mean values (at nodes 13, 51, 67, and 85) of the fault
distribution models regardless of the number of involved
phases are also depicted in Fig. 4. Obviously, there are big
differences between
of different phase loads or
between
of phase loads and
of the whole
system.
of phases A, B, and C are different
because the number of single-phase loads on each phase are
different.
are rather low as single-phase loads
just experience sags due to single-phase-to-ground faults on
the same phase. Generally,
are greater because
phase-to-phase loads are impacted by more faults (faults on
two phases) than phase-to-neutral loads (faults on one phase).
For phase-to-phase loads, there is a little deep sag frequency;
meanwhile, the shallow sag frequency rises greatly because almost phase-to-ground faults (80% system fault rate) just cause
shallow sags to phase-to-phase loads.
for three
phases is the greatest and
for
is equal
to 500 sags per load because three-phase loads will experience
voltage sag for any fault type. The aforementioned remarks
also explain why
, dened for phase loads, is for more
useful indices for estimating the voltage sag performance in the
distribution system where many single-phase loads exist.
Fig. 4 also shows that different positions of the mean value of
fault distribution models result in different spectrums of voltage
sag frequency. It is notable that if the position of mean value gets
closer to the bulk point of supply, the deep sag frequency will
increase, that is, mainly because of the radial network topology
of the distribution system.
87
eBook for You
Table I. The test system with actual dimensions in feet is mapped

out in Fig. 2. The fault positions are assigned with coordinates.
Third, the voltage sag magnitude and phase shift at all load
nodes are computed for all selected fault positions. With the application of fault rates to the selected fault positions, the voltage
sag frequencies corresponding to different characteristics are
obtained. The voltage sag frequency is calculated for the following:
individual load nodes;
all possible phase loads, including phase-to-neutral,
phase-to-phase, and three-phase loads;
the whole test system.
351

Fig. 4. Sag frequency spectrum and SARFI of the whole system for different mean positions for the case that the deviation is = = 0:51
Fig. 5. Voltage sag frequency spectrum of the load-bus 63 on phase A for different deviations. The mean value is at node 67 (upper) and node 13 (lower).
Fig. 7. Voltage sag frequency spectrum and SARFI for the whole system for
different deviations for the mean value at node 67.
Fig. 8. Voltage sag frequency spectrum and SARFI for the whole system for
different deviations for a mean value at node 13.
Fig. 6. Voltage sag frequency spectrum for loads on phase A for different deviations. The mean value is at node 67 (upper) and node 13 (lower).
Figs. 5 and 6 plot the voltage sag frequency for load node
63 (see
Fig. 2) on phase A and for all loads on phase A for
different deviation values of fault distribution

in the case the mean values are identical to
the coordinates of node 13 and node 67. Similarly, Figs. 7 and 8
demonstrate the voltage sag frequency spectrum and 88
eBook for You
352

Fig. 9. Voltage sag frequency distribution for sags lower than 10%, 40% to 50%, 60% to 70%, and 70% to 80%,
node 67.
for the whole test system also for different deviation values
and for the mean values at
node 13 and node 67. Increasing the deviation values
and
will turn the normal distribution into the uniform distribution. It causes shape variations to the voltage sag frequency
spectrum. The clear increase of the frequency of deep sags is
shown in all cases of the sag performance demonstration. If
the mean position of the distribution model is located at node
13, which is very near the bulk point, the frequency of sags
below 10% is even raised by about 50% for the small deviation
. That is also explained as the result of
the radial network topology of the distribution system.
The spectrum of the voltage sag frequency for different case
studies (from Figs. 38) is quite similar in which deep sags account for a large number mainly due to short feeders in the distribution system. The frequency of 40% to 60% sags is also high
as the network topology consists of one trunk line with many lateral taps in the middle. That means the point of common coupling of many load nodes is on the middle of the trunk line. Few
load nodes connected to the trunk line near the bulk point of
supply (the distribution substation) explain why the shallow sag
frequency is very low. Fig. 9 gives us a closer look at the voltage
sag frequency distribution for different sag magnitudes. It is,
without doubt, that deep sag frequencies appear at the nodes
on branches connected close to the far end of the trunk line.
Voltage sags 40% to 50% are distributed rather uniformly excepting nodes near the bulk point. The shallow sag frequencies
mainlyBoccur
at several nodes near the bulk point of supply.
mn H thng in - i hc Bch Khoa H Ni
= = 0:51
= 50%, mean at
VI. CONCLUSION
This paper presented a new method of fault distribution modeling in the stochastic prediction of voltage sag for the distribution system using 2-D distribution models. When using 2-D
distribution models for modeling fault distribution, parameters
of the distribution model should be selected properly to match
the monitored historical data of fault performance of the system
of interest. By using the bivariate normal distribution for modeling fault distribution, this paper also analyzed the inuences
of its parameters on voltage sag performance. It is notable that
the alteration of the deviation value of the distribution has a
much stronger impact on sag performance, especially for the
deep sag frequencies pattern than switching the position of the
mean value. The more concentrated occurrence of faults on one
location in the distribution system of interest will increase the
number of deep sags. The results are also evidence that the typical radial network topology of the distribution system is also
another important reason for the high frequency of deep sags.
2-D stochastic models, such as the bivariate normal distribution used for modeling fault distribution, can provide a good
overview of fault performance of the whole system of interest.
Thus, it is possible not only to analyze the relation between
faults and voltage sags at individual locations of the system,
such as a specic load node or a segment of line, but also to
compute system indices of voltage sags, such as
.
The application of 2-D stochastic models has some limits to
the size of the system of interest. For the sections of the
dis89
eBook for You
353

354
tribution system, of which the size is so large as the one supplied from a bulk distribution substation, it is practical to use
this fault distribution modeling. The accuracy will be further
improved for the distribution systems, of which the topology
features the uniform arrangement of components. In addition,
the stochastic prediction of the transmission system should be
included if the inuence of fault occurring in the transmission
system on voltage sag performance in the distribution system of
interest is considered.
The presence of different phase loads in the distribution
for the whole system without
system indicated that
considering the number of phase of the loads cannot reect
voltage sag performance properly. To have a better assessment
of the voltage sag, this paper develops modied
regarding phase loads. The results proved that there are
,
, and
big differences between
for different phase loads and
for the
whole system. This modication of
is more practical
from the customers point of view when power-supply contracts
are set up.
[11] M. R. Qader, M. H. J. Bollen, and R. N. Allan, Stochastic prediction

of voltage sags in reliability test system, presented at the PQA-97 Europe, Elforsk, Stockholm, Sweden, Jun. 1997.
[12] J. A. Martinez-Velasco and J. Martin-Arnedo, Stochastic prediction
of voltage dips using an electromagnetic transient program, presented
at the 14th PSCC, Sevilla, Spain, Jun. 2002, Paper 4, Session 24.
Dong-Jun Won (M05) was born in Korea on January 1, 1975. He received the B.S., M.S., and Ph.D.
degrees in electrical engineering from Seoul National
University, Seoul, Korea, in 1998, 2000, and 2004, respectively.
Currently, he is a Full-Time Lecturer with the
School of Electrical Engineering with INHA University, Incheon, Korea. He was a Postdoctoral Fellow
with the Advanced Power Technologies Center,
Department of Electrical Engineering, University of
Washington, Seattle. His research interests include
power quality, dispersed generation, renewable energy, and hydrogen economy.
Seung-Il Moon (M93) received the B.S. degree

in electrical engineering from Seoul National University, Seoul, Korea, in 1985 and the M.S. and
Ph.D. degrees in electrical engineering from The
Ohio State University, Columbus, in 1989 and 1993,
respectively.
Currently, he is an Associate Professor of the
School of Electrical Engineering and Computer
Science at Seoul National University. His special
elds of interest include power quality, exible ac
transmission systems (FACTS), renewable energy,
and dispersed generation.
90
eBook for You
REFERENCES
[1] R. C. Dugan, M. F. McGranaghan, and H. W. Beaty, Electric Power
System Quality. New York: McGraw-Hill, 1996.
[2] M. H. J. Bollen, Understanding Power Quality ProblemsVoltage
Sags and Interruptions. New York: IEEE Press, 2000.
[3] D. L. Brooks, R. C. Dugan, M. Waclawiak, and A. Sundaram, Indices
for assessing utility distribution system RMS variation performance,
IEEE Trans. Power Del., vol. 13, no. 1, pp. 254259, Jan. 1998.
[4] M. R. Qader, M. H. J. Bollen, and R. N. Allan, Stochastic prediction
of voltage sags in a large transmission system, IEEE Trans. Ind. Appl.,
vol. 35, no. 1, pp. 152162, Jan./Feb. 1999.
[5] J. V. Milanovic, M. T. Aung, and C. P. Gupta, The inuence of fault
distribution on stochastic prediction of voltage sags, IEEE Trans.
Power Del., vol. 20, no. 1, pp. 278285, Jan. 2005.
[6] R. E. Brown, Electric Power Distribution Reliability. New York:
Marcel-Dekker, 2002.
[7] IEEE Distribution Planning Working Group Report, Radial distribution test feeder, IEEE Trans. Power Syst., vol. 6, no. 3, pp. 975985,
Aug. 1991.
[8] W. H. Kersting, Distribution System Modeling and Analysis. Boca
Raton, FL: CRC, 2002.
[9] T. A. Short, Electric Power Distribution Handbook. Boca Raton, FL:
CRC, 2004.
[10] G. Olguin, Voltage dip (sag) estimation in power system based on stochastic assessment and optimal monitoring, Ph.D. dissertation, Dept.
Energy Environ., Div. Elect. Power Eng., Chalmers Univ. Technol.,
Gotteborg, Sweden, 2005.
Bach Quoc Khanh received the B.S. and Ph.D. degrees in power network and systems from Hanoi University of Technology, Hanoi, Vietnam, in 1994 and
2001, respectively. He received the M.S. degree in
system engineering from the Royal Melbourne Institute of Technology (RMIT), Melbourne, Australia, in
1997.
He is currently a Lecturer with the Faculty of
Electrical Engineering, Electric Power System
Department, Hanoi University of Technology. He
was a Postdoctoral Fellow with the Power System
Laboratory, School of Electrical Engineering and Computer Science, Seoul
National University, Seoul, Korea. His special elds of interest include power
distribution system analysis, DSM, and power quality.
The 2009 ASEAN Symposium on Power and Energy Systems - EEE.RC.ASPES 2009
Prediction of Voltage Sags in Distribution

Systems With Regard to Tripping Time of
Protective Devices
Bach Quoc Khanh (Hanoi University of Technology)
Index Terms-- power quality, voltage sag, distribution

system, equipment compatibility curve, fault distribution
modeling, tripping time.
I. INTRODUCTION
MONG power quality phenomena, the voltage sag

(dip) is defined at IEEE1159, 1995 as a decrease in
RMS voltage to between 0.1 and 0.9 of the nominal voltage
at the power frequency for the duration of 0.5 cycle to 1
minute. Interests in the voltage sag have been getting much
greater recently due to its problems causing on the
performance of sensitive electronic equipments that are
widely used.
Researches about the voltage sag are usually related with
a basic process known as a compatibility assessment [1]
which includes three steps: i. Obtain the voltage sag
performance of the system of interest, ii. Obtain equipment
voltage tolerance, iii. Compare equipment voltage tolerance
with the voltage sag performance and estimate expected
impacts of the voltage sag on the equipment. Researches to
date have already evidenced that obtaining the voltage sag
performance is still needing much further improvement. The
information about the voltage sag is mainly obtained by
monitoring and stochastic prediction [1]. This paper
presents a method of predicting voltage sags in distribution
system using SARFIX-CURVE that is derived from SARFIX
with regard to tripping time of protective devices currently
used in power distribution networks in Vietnam.
II. INDICES FOR VOLTAGE SAG ASSESSEMENT

Voltage sag assessment often bases on its characteristics:
magnitude and duration. There are many indices proposed
for voltage sag quantification [1], [2] and one of frequently
used indices is SARFIX that is defined as follows
SARFI X
N
i
X (i )
(1)
where
X rms voltage threshold; possible values 10-90%
nominal voltage
NX(i) Number of customers experiencing voltage sag with
magnitudes below X% due to measurement event i.
N number of customers served from the section of the
system to be assessed
Despite being widely used, SARFIX only considers the
magnitude of voltage sag and, of course, its value is maybe
much greater than the actual number of tripping electrical
appliances, especially when the duration of sags is small
enough (less than a half second). To take the sag duration
into account, SARFIX is developed into SARFIX-CURVE [2],
[4], [6] which is defined below
Figure 1. ITI curve for susceptibility of computer equipment.

Bach Quoc Khanh is with Electric Power System Department,
Electricity Faculty, Hanoi University of Technology, 1 Dai Co Viet Rd.,
Hanoi, Vietnam (e-mail: bq_khanh-htd@mail.hut.edu.vn).
B mnNo.
H thng
Paper
PS-1in - i hc Bch Khoa H Ni
91
eBook for You
Abstract-- This paper presents a method of assessing a

power quality phenomena in distribution systems - voltage sag.
The voltage sag performance is obtained by the problem of
stochastic prediction of voltage sag in power systems [2] basing
on System Average RMS variation Frequency Index (SARFIX).
However,
SARFIX is modified into SARFIX-CURVE that
considers not only the magnitude of voltage sag, but also its
duration. The resulting SARFIX-CURVE provides a better
understanding of the influence of voltage sag on the electric
loads. The duration of voltage sag is modeled regarding the
tripping time of protective devices in distribution systems. The
paper also applies this method to assess voltage sag
performance of the 22kV feeder 482-E14 of 110/35/22kV Giam
substation in Hanoi city, Vietnam.
'
X (i )
i 1
SARFI X CURVE
(2)
where
N X' (i ) : Number of customers tripped when experiencing
voltage sag with magnitudes below X% due to measurement
event i.
SARFIX-CURVE corresponds to voltage sags below an
equipment compatibility curve. So far, frequently used
curves are CBEMA, ITIC and SEMI [1]. Obviously,
SARFIX-CURVE can provide a better understanding of the
influence of voltage sag on the operation of electric
equipment in electric networks. This paper presents the
method of calculating SARFIX-CURVE using ITI curve (Figure
1) as a case study.
III.
REDICTION OF VOLTAGE SAG IN DISTRIBUTION SYSTEM
A. Problem definition
The problem of stochastic prediction of voltage sag can
obtain the voltage sag performance of a specific electric
system by using data of events leading to sags. In fact, more
than 90% sag events are resulted from short-circuits and it is
possible to use fault modeling and short-circuit calculation
tools to predict voltage sags in the power system (Figure 2).
tPD
E
~
Vt
ZF
ZS
Vt
Short circuit
tPD
Load at PCC
E
VSag
Figure 2. Model of voltage sag

prediction in power systems
This work uses the method of fault position [1] for

voltage sag prediction in distribution systems with following
significant steps
- Modeling the fault distribution on of a given
segment of distribution system (see part B)
- Calculating the short-circuit current and voltage
sags at all influenced load nodes.
- Cumulating system voltage sags with different
characteristics and obtaining SARFIX.
- Cumulating system voltage sags that cause
equipment to trip and obtaining SARFIX-CURVE.
To obtain SARFIX-CURVE, this work uses the typical
tripping curve (tPD = f(IF)) of protective devices like fuses,
feeder circuit breakers currently used in distribution
systems. Each sag is plotted as a point characterized by a
pair of co-ordinates (magnitude of voltage sag and tripping
time). If the point falls out of voltage tolerant envelop
(Figure 1), the sag is cumulated to calculate SARFIX-CURVE.
B. Fault Distribution Modeling
Modeling the fault distribution is to determine the shortcircuit fault frequency (i.e. fault rate or the number of shortB mnNo.
H thng
Paper
circuit faults per year) for all fault types at all possible fault
positions throughout the system of interest [3]. It consists of
the selection of fault position and fault type and the
distribution of fault rate for selected fault positions and fault
types.
Fault positions are generally chosen in the way that a
fault position should represent short-circuit faults leading to
sags with the similar characteristics [1]. For the distribution
system with typically radial network topology, small line
segments and distribution transformers along the trunk
feeders, it is possible to apply only one fault position for
each distribution transformer and also one fault position for
each line segment.
Different fault types should be applied to each fault
position mainly depending on number of phases available at
the selected fault positions. The fault rate of each fault type
is normally referred from the observed historical data.
Fault rate mainly depends on fault position, fault type
and fault cause. For a segment of distribution system that is
geographically seen as small area, it possible to assume that
fault rate of each fault type follows uniform distribution for
all fault positions. [3]. In this sense, the fault rate at each
position is identical to component failure rate that is based
on component reliability. In reality, uniform fault
distribution is a practical assumption for distribution
systems because the service area of a certain distribution
line outgoing from a distribution substation is normally
small.
C. Assumptions
Besides fault distribution modeling, for the distribution
system, following assumptions are possibly considered [3].
- Voltage sags are only caused by faults in the
distribution system.
- If the distribution system is supposedly a section of a
large distribution system, only faults occurred within it are
considered. The faults in sections fed from other distribution
substations can be skipped as the transformer impedance in
distribution substations, in reality, is rather high. Similarly,
the faults in low voltage networks are also ignored because
of the large impedance of distribution transformers. This
assumption only neglects voltage sags caused by faults in
the transmission system. It will be considered if the
stochastic prediction of voltage sag in large transmission
systems [7] is included.
- In terms of reliability, the distribution system is
modeled on two main components: lines and distribution
transformers. The reliability of any other distribution
equipment is supposedly included in the reliability of these
two components.
- The fault positions are selected as mentioned in the Part
III.B. For transformers, one fault position each load node
(i.e. the nodes connected with distribution transformers) is
applied. For lines, one fault position is also applied for each
line segment. Because of short line segments, the paper
selects the fault position at the end of each line segment.
- Fault types (single phase to ground, phase to phase, two
phases to ground and three phases to ground) are applied to
fault positions depending on the number of available phases.
92
eBook for You
D. System voltage sag calculations

Short-circuit calculations and resulting voltage sag
magnitude at load nodes in distribution systems is
performed by MatLab programming that used in [3]. The
program consists of two modules
- Short circuit calculation
- Fault distribution modeling
Its block diagram is briefly depicted as Figure 3
START
DETERMINE FAULT LINE

Find nodes and branches on
fault current carrying line
ON FAULT-LINE CALCULATION
Calculate fault current IN and
sags VS at nodes on fault line
OFF FAULT-LINE CALCULATION

Calculate voltage sags VS at
nodes not on fault line
are also assumed as follows

24kV bus of
110kV Giam
substation
Circuit
breaker
Fuse
Distribution
transformer
SARFIX CALCULATION
Distribution
transformer
Figure 4. Brief description of

24kV feeder protection system
- Single phase to ground (N1) :
65%
- Two phase to ground (N11) :
10%
- Two phase together (N2) :
20%
- Three phase to ground (N3) :
5%
and the component fault rates are supposed to be
- Transformer :
50%
- Line :
50%
The tripping curve used for this work is the typical
inverse curve of in-service protective devices in distribution
systems like fuse-cutout for distribution transformer
protection, overcurrent relay for 24kV line feeder. The
common formula of tripping curve is
t PD
TRIPPING TIME
tPD = f(IN)
Fuse
a
(I * )b 1
(3)
eBook for You
The fault impedance is assumed to be negligible.
where
I*: Ratio of fault current IN and pickup current IP.
a, b: Constants that are selectable.
Sag quantification by magnitude
calculation
SARFIX-CURVE CALCULATION
Sag quantification by duration
STOP
Figure 3. Block-diagram of voltage sag prediction
and SARFIX-CURVE in distribution systems
IV. A CASE STUDY
A. Case study definition
This work illustrates the method by predicting voltage
sag performance and resulting SARFIX-CURVE for a 24kV
feeder network in Hanoi, Vietnam. Preliminary data is as
follows
The network segment under consideration: Feeder 482E14, 24kV, underground cable, outgoing from 110/35/22kV
Giam substation. Its a radial network with 99 nodes and 98
branches. Fault positions can be selected at load nodes for
distribution transformer fault and at all nodes for line fault.
Besides, contributory percentages of different fault type
B mnNo.
H thng
Paper
V. RESULT DEMONSTRATION AND ANALYSIS

Firstly, the system fault rate (the total of faults occurring
in the test system over a certain period of time) is assumed
to be an arbitrary number, say, 100 faults. This value is just
for calculation and easier graphic demonstration of the
results. The system fault rate is then distributed uniformly to
all fault positions as assuming in Part III.B. Short-circuit
calculation is made at every fault positions and resulting
voltage sags at all load nodes are identified by their
magnitudes. Besides, the fault current is used to determine
voltage sag duration as per (3) and each voltage sag
identified above are again checked to see whether it is to fall
inside the voltage tolerant envelope of ITI curve or not. If it
is inside, it is taken into account for calculating SARFIXCURVE. Finally two indices SARFIX and SARFIX-CURVE are
obtained and plotted in the same graphics for analysis. The
results are depicted on two graphics. Figure 5 depicts the
system voltage sag frequency spectrum. Figure 6 depicts
SARFIX and SARRFIX-CURVE.
The results also indicate some following remarks
- Deep sag frequency rises highly due to the radial
network topology with short distances of cable
lines in distribution systems.
- 40-50% sag is also a little greater than other sags
because the feeder consists of one trunk line with
many lateral taps in the middle. That means the
93
45
Sag
Sag leading load failure
40
35
developed in further researches. For a larger network, a

more suitable fault distribution should be considered [3],
[5]. In addition, a combination of the problems of predicting
voltage sags in distribution systems and transmission system
[7] will provide a more comprehensive understanding of
voltage sag performance of a power system.
30
25
VII. REFERENCES
20
[1]
15
[2]
10
5
[3]
0
0-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
VSag (percentage of Un)
[4]
Figure 5. System voltage sag frequency spectrum

SARFI
[5]
120
SARFIX
SARFIX-CURVE
100
[6]
80
60
[7]
40
[8]
20
0
<10
<20
<30
<40
<50
<60
<70
<80
<90
VSag (percentage of Un)

-
Figure 6. SARFIX and SARFIX-CURVE

Voltage sags with X greater than 70% are very few
because the system and distribution transformer
impedances normally are much higher than
distribution lines.
SARFIX and SARFIX-CURVE are slightly different
because the tripping time of protective devices in
distribution systems is typically 0.5 seconds and
frequencies of voltage sag of 70-80% and 80-90%
are very small. In ITI curve, sags with the
magnitude X lower than 70% nominal voltage
feature very short duration (less than one cycle)
and thus they are certainly taken in to account for
calculating SARFIX-CURVE.
M.H.J. Bollen, Understanding power quality problems - voltage sags

and interruptions, IEEE Press, 2000.
Recommended practice for the establishment of voltage sag indices,
Draft 6, IEEE P1564, Jan 2004.
Bach Quoc Khanh, Dong Jun Won, Seung Il Moon, Fault
Distribution Modeling Using Stochastic Bivariate Models For
Prediction of Voltage Sag in Distribution Systems, IEEE Trans.
Power Delivery, pp. 347-354, Vol.23, No.1, January 2008.
Juan A. Martinez, Jacinto Martin-Arnedo, Voltage Sag Studies in
Distribution Networks - Part II: Voltage Sag Assessment, Part III Voltage Sag Index Calculation, IEEE Trans. Power Delivery, pp.
1679-1697, Vol. 21, No. 3, July 2006.
Jovica V. Milanovic, Myo Thu Aung, C. P. Gupta, The Influence of
Fault Distribution on Stochastic Prediction of Voltage Sags, IEEE
Trans. Power Delivery, pp. 278-285, Vol. 20, No. 1, Jan 2005.
D. L. Brooks, R. C. Dugan, Marek Waclawiak, Ashok Sundaram,
Indices for Assessing Utility Distribution System RMS Variation
Performance, IEEE Trans. Power Delivery, vol.13, no.1, pp.254-259,
Jan. 1998.
M.R.Qader, M.H.J.Bollen, and R.N.Allan, Stochastic Prediction of
Voltage Sags in a Large Transmission System, IEEE Trans. Industry
Applications, vol.35, no.1, pp.152-162, Jan./Feb. 1999.
M.R.Qader, M.H.J. Bollen and R.N.Allan, Stochastic Prediction of
Voltage Sags in Reliability Test System, PQA-97 Europe, Elforsk,
Stockholm, Sweden, Jun. 1997.
VIII. BIOGRAPHIES
Bach Quoc Khanh received B.S. and Ph.D. degrees in power network
and systems from Hanoi University of Technology, Hanoi, Vietnam in 1994
and 2001 respectively. He received M.S. in system
engineering from RMIT, Melbourne, Australia in
1997. He is a teaching staff of Electric Power
System dept., Electrical Engineering Faculty,
Hanoi Univeristy of Technology. His special
fields of interest include power distribution
system analysis, DSM and power quality.
VI. CONCLUSIONS
This paper presented a method of assessing voltage sags
in distribution systems with regard to tripping time of
protective devices. The assessment bases on SARFIX-CURVE
that combines SARFIX and equipment compatibility curves.
Therefore, the results of assessment provide a better
understanding of the influence of voltage sag on loads.
This method is also found useful for power quality
assessment and power supply contracting principles for
power distribution utilities in Vietnam in the process of
electricity market establishment because the management of
distribution system is becoming financially separated from
the power system.
The application of the method has some limits that can be
B mnNo.
H thng
Paper
94
eBook for You
System voltage sag frequency
point of common coupling of many load nodes is

on the middle of the trunk line.
TP CH KHOA HC & CNG NGH CC TRNG I HC K THUT h S 77 - 2010
NH GI ST GIM IN P NGN HN
TRN LI TRUYN TI IN 220KV VIT NAM
PREDICTION OF VOLTAGE SAGS IN THE 220KV TRANSMISSION SYSTEM OF VIETNAM
Bch Quc Khnh
Phng Th Anh
Trng i hc Bch Khoa H Ni
Cng ty CP T vn Xy dng in I
TM TT
Bi bo trnh by phng php nh gi mt hin tng cht lng in nng (CLN) trn li
truyn ti in (LTT) l st p ngn hn (SANH - voltage sag) [1]. M hnh nh gi SANH da trn
phng php d bo ngu nhin SANH [2] trong h thng in (HT). Vic nh gi ny da trn ch
tiu tn sut SANH trung bnh ca HT vi c tnh X (SARFIX) v SARFIX-CURVE [3] cho php xt n
khng ch c trng bin ca SANH m cn c c trng thi gian tn ti SANH. i tng tnh
ton l h thng truyn ti in 220kV ca Vit Nam theo tng s 6 vi t l sut s c ngn mch
thc t ca nm 2008. Vic nh gi ny l mt c gng u tin nh lng ha tnh hnh mt hin
tng cht lng in nng ph bin trn mt li in din rng thc t gip cho vic nh gi cht
lng in nng ni chung ca h thng in Vit Nam hin nay.
This paper presents a method of predicting a power quality phenomena in distribution systems,
voltage sag [1]. The calculation of voltage sag performance follows the model of stochastic prediction
of voltage sag in power systems [2]. The voltage sag performance is predicted basing on the System
Average RMS variation Frequency Index (SARFIX) and SARFIX-CURVE [3] that considers not only the
characteristics - magnitude, but also the characteristics duration of voltage sag. The objective of
research is the whole 220kV transmisson systems in Vietnam as per the 6th master-plan with actual
data of fault rate of the year 2008. This prediction is the first effort of quantifying the voltage sag
performance for such a large transmission system that helps assess the power quality of the electric
power system in Vietnam now.
CLN c cung cp v nh gi tc ng ca
CLN i vi ph ti. Vic xc nh yu cu
CLN ca cc ph ti thuc v cc nh sn
xut thit b dng in m in hnh l c tnh
chu in p ca ph ti CBEMA, ITIC hoc
SEMI [1] (Hnh 1).
I. T VN
Theo IEEE-1159, 1995, SANH (voltage
sag) l hin tng CLN trong gi tr in
p hiu dng ca li in st gim cn t 0,1
n 0,9 in p nh mc trong thi gian t 0,5
chu k n 1 pht [1]. SANH c th lm cho
cc thit b in nhy cm nh in t cng
sut, cc b iu tc hay my tnh c nhn
ngng hoc lm vic khng mong mun. Hin
tng ny li rt hay xy ra, trong khi nng
cao hiu sut qu trnh hay vic ng dng cc
cng ngh mi, cc thit b in ng dng in
t cng sut ngy cng c s dng nhiu, do
SANH ngy cng c quan tm nghin
cu. Trc khi xem xt nhng gii php khc
phc tc ng ca SANH, yu cu nh gi
SANH trong HT lun c t ra. Qu trnh
nh gi CLN ni chung v SANH ni ring
thng tri qua ba khu ch yu [1] l i. Nhn
dng tnh hnh CLN c cung cp, ii. Xc
nh yu cu CLN ca cc ph ti, iii. So
snh yu cu CLN ca ph ti vi tnh hnh
Trong khi , vic nhn dng tnh hnh

CLN l nhim v ca pha cung cp in.
Vit Nam, bt u c nhng nghin
cu chuyn su v nh gi tnh hnh SANH
trong HT [2, 3], tuy nhin vic nh lng ha
tnh hnh SANH trn HT thc t Vit Nam
vn cha c thc hin. Nguyn nhn chnh
hin nay l khng c mt c s d liu v
CLN ni chung v SANH ni ring ca HT
Vit Nam do h thng gim st v lu tr thng
tin v CLN vn cn rt thiu. Bn cnh vic
gim st CLN, mt cch gin tip xc nh
tnh hnh SANH trn HT c th dng m hnh
d bo CLN da trn cc nguyn nhn sinh ra
n. Trong cc nguyn nhn ny, trn 90%
72
95
eBook for You
ABSTRACT
SANH l do s c ngn mch trong HT. Do

, c th nh gi SANH thng qua m phng
v tnh ton ngn mch trn HT theo phng
php im s c [1, 2].
Vng
an ton
Vng mt an ton
Hnh 2. Cc c trng ca SANH

2.2. Xy dng m hnh im s c i vi
li in truyn ti 220kV ca Vit Nam
Vng mt
an ton
Hnh 1. ng cong chu in p ca nhm

thit b SEMI (Semiconduactor Equipment and
Materials International group)
Bi bo trnh by phng php d bo
SANH trn ton b li in truyn ti 220kV
ca HT Vit Nam theo Tng s VI dng
phng php im s c vi s liu s c ngn
mch trn li truyn ti 220kV thc t ca
nm 2008 v s dng ch tiu SARFIX v
SARFIX-CURVE [3,4,8].
II. XY DNG M HNH BI TON
2.1 Phng php im s c dng cho d
bo SANH trong li in truyn ti 220kV
Theo phng php ny, gi thit SANH
gy ra l do ngn mch trong HT. Khi , c
trng bin ca SANH (Hnh 2) c xc
nh bi v tr v loi s c ngn mch [1, 4].
c trng thi gian tn ti SANH th ph thuc
vo thi gian loi tr ngn mch ca cc thit b
bo v.
eBook for You
- Chn v tr s c : Ch xt s c ngn mch

trn li 220kV. Cc ngn mch xy ra li
c in p thp hn c th gi thit l t nh
hng n li 220kV do tng tr cc my bin
p khu vc v a phng l kh ln. V bn
thn li 220kV rt ln nn trong nghin
cu ny cha xt cc s c ngn mch trn li
500kV v ti cc ngun in. Bin SANH
ti cc nt ph ti ph thuc vo v tr im
ngn mch. V nguyn tc s c c th xy ra
ti bt c u trn li 220kV, tuy nhin nu
trong mt phm vi ca li in m ngn mch
u dn n SANH ti cc nt ph ti c
cng c trng bin th ch cn chn mt v
tr in hnh. S h thng truyn ti in
220kV theo tng s VI [9] tnh n nm
2008 gm 66 trm 220kV, 98 nhnh ng dy
220kV vi tng chiu di l 7988km. S c
ngn mch trn li in truyn ti c xt
cho c ng dy v trm bin p nn i vi
s c trm bin p xt tt c 66 nt c trm
220kV, cn s c trn ng dy, ty theo tng
chiu di mi nhnh ng dy m xt mt
hoc vi im s c trn nhnh . Nhn chung,
cc im s c cch nhau t 10km n 40km.
Tng s im ngn mch trn ng dy l 169
im.
- Chn loi s c ngn mch : Li in cao p
l 3 pha, vai tr cc pha nh nhau nn khi xt
cc dng ngn mch th xt 4 dng vi t l
phn b sut s c [10] :
Ngn mch 1 pha - t :
65%
Ngn mch 2 pha :
20%
Ngn mch 2 pha - t :
10%
Ngn mch 3 pha :
5%
Cc c trng trn y ca SANH c

xc nh ti cc nt ph ti ca li truyn ti
220kV l cc trm bin p 220kV t xc
nh cc ch tiu SARFIX v SARFIX-CURVE cho
c h thng truyn ti in 220kV ca Vit
Nam.
73
96
- Phn b s c ngn mch : S c ngn mch

mang tnh ngu nhin ph thuc vo nhiu yu
t [2] nn sut s c nhn chung khc nhau i
vi tng loi s c v v tr s c. Trong nghin
cu ny, do s liu thng k v phn b s c
trn li truyn ti 220kV cha chi tit nn
s phn b s c c xut theo m hnh
phn b u. Theo thng k ca tng cng ty
truyn ti in quc gia li truyn ti 220kV
trong nm 2008 c tng s 45 s c xy ra ti
cc nt trm 220kV v 143 s c trn cc
ng dy 220kV. Sut s c ng dy l
0,0179 s c/km/nm v ca trm bin p l
0,682 s c/trm/nm. Phn b s c cho tng
loi ngn mch i vi ng dy v my bin
p c cho Bng 1.
tng loi s c ngn mch gy ra. Gn sut s

c cho tng v tr v tng loi s c s rt ra
c tn sut SANH ti tng nt ph ti do s
c ang xt gy ra. Lp li vic tnh ngn mch
v SANH vi cc im s c khc ri tng hp
li ta c tn sut SANH vi cc c tnh bin
khc nhau do s c ti tt c cc im ngn
mch trn li truyn ti 220kV gy ra, v cui
cng ta rt ra c ch tiu SARFIX ca ton h
thng.
Start
M phng phn b s c trn
li truyn ti 220kV
Bng 1. Phn b s c theo loi s c
N(1)
N(1,1)
N(2)
N(3)
Sut s c
ng dy
Trm bin p
92,95
29,25
28,60
9,00
14,30
4,50
7,15
2,25
M phng li in, tnh

ngn mch bng PSS/E
Tnh SANH cho tng nt ti
bng PSS/E
Tnh SARFIX cho li truyn
ti in 220kV
- Chn v tr nt ph ti cn xc nh SANH :
Trn li truyn ti, nt ph ti l cc nt c
trm 220kV cp in xung cc li c in p
thp hn. Li 220kV li c dng mch vng
nn nhn chung trn mi nhnh ng dy
220kV, bo v c t ti c hai u v khi
xy ra s c ngn mch trn nhnh ng dy
no th nhnh s b c lp ring. Do , tt
c cc nt (66 trm 220kV) trn li in u
b SANH khi s c, khng c nt no b mt
in duy tr v ta phi tnh SANH cho 66 nt
ny.
eBook for You
Loi
Tnh SARFIX-CURVE cho

truyn ti in 220kV
Stop
Hnh 3. S khi nh gi SANH trn li

truyn ti in 220kV Vit Nam
- Xc nh SARFIX-CURVE : xc nh SARFIXCURVE, phi xt n thi gian loi tr s c ca
h thng bo v li 220kV v dng c tnh
chu in p la chn. i vi li 220kV ca
Vit Nam hin nay, bo v chnh l bo v ct
nhanh (so lch dng in hoc tng tr cc
tiu) vi tng thi gian ct ngn mch t 120ms
n 150ms. Trong nghin cu s dng c tnh
chu in p ca cc ph ti nhy cm l SEMI,
v vi thi gian loi tr s c nh trn, cc
SANH c bin di 70% u ri vo vng
mt an ton v lm cc ph ti nhy cm ngng
lm vic. Do , khi xc nh SARFIX-CURVE,
vi X t 70% n 100% in p nh mc th
SARFIX-CURVE khng i.
2.3 Tnh ton ngn mch v xc nh c

trung bin SANH trong li in truyn
ti 220kV ca Vit Nam
Vic tnh ngn mch v SANH ti cc
nt ph ti trong li truyn ti 220kV c
thc hin bng chng trnh PSS/E. S khi
cc bc tnh ton nh hnh 3.
- Xc nh SARFIX : Vic chn v tr v xc
nh sut s c cho tng v tr v tng loi s
c c thc hin nh 2.2. Dng chng
trnh PSS/E tnh ngn mch ti tng im s c
vi tng loi s c v t xc nh bin
SANH ti tt c 66 nt ph ti do tng im v
74
97
III. PHN TCH KT QU
SARFIX
Thc hin trnh t tnh ton nh s

khi hnh 3, sau y l tm tt mt s kt qu
ng lu :
- Tn sut SANH trung bnh ca mt nt ph ti
bt k :
Hnh 4 v 5 biu din kt qu tnh ton
tn sut SANH ti nt trm 220kV Mai ng,
Thnh ph H Ni.
Tn sut SANH
N(3)
N(1,1)
N(2)
N(1)
SARFIX-CURVE
X
X
Hnh 7. SARFIX v SARFIX-CURVE ca li

truyn ti in 220kV tnh cho nm 2008
Hnh 4. Tn sut SANH nt 220kV Mai ng

theo tng khong c trng bin
Tn sut SANH
eBook for You
Hnh 4 biu din tn sut SANH theo

tng khong c trng bin ca SANH.
Hnh 5 biu din tn sut SANH khi SANH c
bin nh hn tng mc c trng bin .
- Tn sut SANH trung bnh h thng :
i vi mi ph ti, tn sut SANH
trung bnh theo tng khong c trng bin
SANH c cho Hnh 6. V cui cng l ch
tiu SARFIX ca ton b li truyn ti in
220kV Vit Nam c cho Hnh 7.
T kt qu cho ta mt s nhn xt ng
ch sau :
- Tn sut SANH ng vi tng loi s c tng
ng vi tn sut ca tng loi s c.
- SANH nng (70%-90%) c tn sut kh ln

v tn sut SANH d ca nt c th l 220kV
Mai ng hay trung bnh cho tng nt ch
khong 25 ln/nm rt nh so vi tng s s c
trn li 220kV l 188. l v li 220kV tri
ton quc nn ngn mch xy ra tng min t
nh hng n cc ph ti ti cc min khc.
Hnh 5. Tn sut SANH nt 220kV Mai ng

theo c trng bin ly tin
Tn sut
- Tn sut SANH nt 220kV Mai ng ln

hn SARFIX ca ton h thng v li 220kV
min Bc c nhiu ph ti hn do a bn rng
hn.
Hnh 6. Tn sut trung bnh SANH theo tng

khong c trng bin SANH
75
98
li truyn ti in cn c xt thm cc
SANH do ngn mch phn ngun v li
truyn ti in 500kV. Hn na, nghin cu
cng cn c th pht trin vic xem xt cc yu
t nh hng n phn b s c khi li truyn
ti in ca Vit Nam tri trn mt phm vi
rng ln vi tnh hnh s c khc nhau. Cc m
hnh ngu nhin vi cc lut phn b xc sut
ph hp vi tnh hnh xy ra s c thc t c
th c xem xt [2, 6, 8].
V. KT LUN
Bi bo trnh by phng php nh
gi SANH trn li truyn ti in 220kV ca
Vit Nam thng qua ch tiu SARFIX v
SARFIX-CURVE. y l c gng u tin nh
lng ha vic nh gi tnh hnh SANH ni
ring v CLN ni chung trong HT Vit Nam
trn mt din rng.
Nghin cu trong bi bo cng cn c
pht trin thm. Kt qu nh gi SANH trong
1.
M. H. J. Bollen; Understanding power quality problems - voltage sags and interruptions; IEEE
Press, 2000.
2.
Bach Quoc Khanh, Dong Jun Won, Seung Il Moon; Fault Distribution Modeling Using Stochastic
Bivariate Models For Prediction of Voltage Sag in Distribution Systems; IEEE Trans. Power
Delivery, Vol.23, No.1, pp.347-354, Jan. 2008.
3.
Bach Quoc Khanh; Prediction of Voltage Sags in Distribution Systems With Regard to Tripping
Time of Protective Devices; Proceeding, EEE.CR.ASPES2009, Tech. Section 2.1., Hua Hin,
Thailand, Sep. 28-29, 2009.
4.
D. L. Brooks, R. C. Dugan, Marek Waclawiak, Ashok Sundaram; Indices for Assessing Utility
Distribution System RMS Variation Performance; IEEE Trans. Power Delivery, Vol.13, No.1,
pp.254-259, Jan. 1998.
5.
M.R.Qader, M.H.J.Bollen, and R.N.Allan; Stochastic Prediction of Voltage Sags in a Large

Transmission System; IEEE Trans. Industry Applications, Vol.35, No.1, pp.152-162, Jan./Feb.
1999.
6.
Juan a. marTNez-Velasco; Computer-Based Voltage Dip Assessment in Transmission and

Distribution Networks, Electrical Power Quality and Utilisation, Journal Vol.XIV, No.1, 2008.
7.
J.V.Milanovic, M.T.Aung and C.P.Gupta; The Influence of Fault Distribution on Stochastic

Prediction of Voltage Sags; IEEE Trans. Power Delivery, vol.20, no.1, pp.278-285, Jan. 2005.
8.
Recommended practice for the establishment of voltage sag indices, Draft 6, IEEE P1564, Jan
2004.
9.
Tng s pht trin H thng in Vit Nam, Bn IV, Vin Nng lng, 2006.
10. T. A. Short; Electric Power Distribution Handbook, CRC Press, 2004.

a ch lin h :
Bch Quc Khnh - Tel: 0904.698.900, email: bq_khanh-htd@mail.hut.edu.vn

B mn H thng in, Khoa in, Trng i hc Bch khoa H Ni
S 1, i C Vit, H Ni
76
99
eBook for You
TI LIU THAM KHO
Prediction of Voltage Sag in The Transmission

System of Vietnam, A Case Study
Bach Quoc Khanh, Nguyen Hong Phuc
Index Terms--transmission system, power quality, voltage sag

frequency, stochastic prediction, fault distribution, fault clearing
time, ITIC, SEMI curve.
I. INTRODUCTION
and estimate expected impact of voltage sag on the equipment.

The permissible voltage tolerance for electric equipment,
normally defined by the manufacturers and the well-known
PQ curves for susceptibility of computer equipment displays
are CBEMA, ITIC or SEMI [1] whereas power quality
assessment of power supply system is utilities duty. This paper
is the first effort to assess the voltage sag performance in the
transmission system of Vietnam by using the method of
stochastic prediction of voltage sags [1], [2], [3] using
SARFICURVE-X that is derived from SARFIX with regard to
fault clearing time of protective devices currently used in the
transmission system in Vietnam.
II. INDICES FOR VOLTAGE SAG ASSESSMENT
Voltage sag assessment often relies on voltage sag
characteristics: magnitude and duration. There are many
indices proposed for voltage sag quantification. [1], [4] In this
paper the authors use one of the frequently used indices,
SARFIX. It is defined as follows
mong power quality phenomena, voltage sag (dip) is

defined by IEEE 1159 (1995) as a decrease in RMS
voltage to between 0.1 to 0.9 of nominal voltage at power
frequency for duration of 0.5 cycle to 1 minute. Interests in
voltage sag has been getting much greater recently in Vietnam
due to its impact on the performance of sensitive electronic
equipment like variable speed drives, computer-controlled
production lines that are widely used, especially in industry.
Although voltage sags are more common in distribution
system, many causes leading to voltage sag are derived from
transmission systems. An assessment of voltage sag in
transmission systems is important for utilities and customers
in Vietnam now.
Voltage sag assessment normally comes prior looking for
the solution of voltage sag mitigation. Voltage sag assessment
is usually related with the basic process known as a
compatibility assessment [1] which includes three steps: (i).
Obtain the voltage sag performance of the system of interest,
(ii). Obtain equipment voltage tolerance and (iii). Compare
equipment voltage tolerance with the voltage sag performance
Bach Quoc Khanh is a faculty member with Electric Power Systems
Department, Electrical Engineering Faculty, Hanoi University of Science and
Technology, 1 Dai Co Viet Rd., Hanoi, Vietnam (e-mail: bq_khanhhtd@mail.hut.edu.vn).
Nguyen Hong Phuc is a master student with Electric Power System
Department, Electricity Faculty, Hanoi University of Science and Technology,
1 Dai Co Viet Rd., Hanoi, Vietnam (e-mail: cllx2003@gmail.com).
978-1-61284-788-7/11/$26.00
2011
B mn H thng in - i hc Bch
KhoaIEEE
H Ni
SARFI X
X (i )
(1)
where
X rms voltage threshold; possible values 10-90% nominal
voltage
NX(i) Number of customers experiencing voltage sag with
magnitudes below X% due to measurement event i.
N number of customers served from the section of the system
to be assessed
Despite being widely used, SARFIX only considers the
magnitude of voltage sag. Unfortunately, the magnitude value
maybe much greater than the actual number of tripped
electrical appliances, especially when the duration of sags is
small enough (less than a half second), such as for
transmission system in Vietnam where the total fault clearing
time of protection system is typically less than 5 to 7 cycles of
the mains frequency. To take the voltage sag duration into
account, SARFIX is developed into SARFICURVE-X [5], [6]
which is defined below
m
SARFI CURVE X
'
X (i )
i 1
(2)
where
N X' (i ) : Number of customers tripped when experiencing

voltage sag with magnitudes below X% due to measurement
100
eBook for You
Abstract In this paper, a novel effort for prediction of

voltage sag in the entire transmission system of Vietnam is
presented. As the Vietnamese electricity industry moves toward
the electricity market, prediction will help utilities have early
assessment of power quality in transmission system. The
proposed prediction approach uses a fault position method in
which the fault distribution in the transmission system is created
based on an actual fault occurrence in the entire 220kV and
500kV transmission system throughout Vietnam that took place
in 2008. The research also makes use of the SARFICURVE with
ITIC and SEMI curve, which takes into account of the actual
fault clearing time of protective devices used in transmission
system in Vietnam. By using SARFICURVE, a better assessment of
voltage sag performance is obtained in the transmission system
with regard to loads voltage tolerance.
event i.
If we plot voltage sag as a point with co-ordinates being its
magnitude and duration on the graph of the equipment
compatibility curve, SARFICURVE-X corresponding to voltage
sags falling out of the equipment voltage tolerant area (Fig. 1)
will be obtained. So far, well known curves are CBEMA, ITIC
and SEMI [1]. Obviously, SARFICURVE-X can provide a better
understanding of the influence of voltage sag on the operation
of electric equipment in electric networks. This paper presents
the method of calculating SARFIX-CURVE using ITIC and SEMI
curve (SARFIITIC-X and SARFISEMI-X) as case studies.
SARFIX-CURVE
disqualified
SARFIX-CURVE
qualified
Fig 1. ITI curve for susceptibility of computer equipment

III. PREDICTION OF VOLTAGE SAG IN
THE TRANSMISSION SYSTEM OF VIETNAM
A. Problem definition
The problem with stochastic prediction of voltage sag is
that it can only obtain the voltage sag performance of a
specific electric system by using data of causal events leading
to sags. In fact, more than 90% sag events are resulted from
short-circuits, hereby called faults, and it is possible to use
fault modelling and short-circuit calculation tools to simulate
and predict voltage sags in the power system. This work uses
the method of fault position [1] for voltage sag prediction in
the transmission systems with following significant steps
1. Modeling the fault distribution of the transmission system
of Vietnam event modeling (Sub section B)
2. Calculating the short-circuit current and voltage sags at all
influenced load nodes event indices (Sub section C)
3. Quantifying voltage sag frequency at load nodes (site
indices) and cumulating system sags with different
characteristics and obtaining SARFIX (system indices)
4. Cumulating system voltage sags that cause equipment to
trip and obtaining SARFICURVE-X.
To obtain SARFIX-CURVE, the voltage sag duration that
depends on the fault clearing time of protective system should
be considered. This work takes the typical tripping time of
protective devices (instantaneous protective relay) and high
voltage circuit breakers currently used in the transmission
system in Vietnam into its calculation.
B. Fault Distribution Modeling and Assumptions

- Fault distribution modeling: Fault distribution modeling
considers the occurrence of all faults in the whole transmission
system of Vietnam that cover 500kV and 220kV networks.
The scope of the transmission system of Vietnam starts from
the points of energy receiving from generating centers or
interconnection points with the transmission system of South
China to load nodes that are step-down 220kV substations. An
individual fault (short-circuit) is characterized by a pair of
parameters: fault position, fault type and its occurrence is
assigned a fault rate. All faults with their assigned rate of
occurrence build up a fault distribution model. Following are
analyses of each fault characteristics for the transmission
system of Vietnam.
- Fault position: The fault can occur anywhere in the
transmission system including 500kV and 220kV networks.
Since load nodes of the transmission system are 220kV stepdown transformers, faults in 110kV networks and distribution
networks should not considerably impact on voltage sags in
transmission system because of large impedance of 220kV
step-down transformers. Faults at the power generating
sources should be included in the faults at the 220kV step-up
transformers. Therefore, this work only considers faults that
occur in the transmission system. According to [1], [3], [7],
basing on the concept of area of vulnerability, fault
positions should be generally chosen in the manner that a fault
position should be the representative for other nearby shortcircuit faults in a portion of network that cause voltage sags to
load nodes with the similar characteristics (similar
magnitudes). Voltage sag magnitude normally divides in 9
ranges : 0-0.1, 0.1-0.2,, 0.8-0.9 p.u. Similar manitudes mean
the magnitudes that fall inside a same range of magnitude
above said. Faults in the transmission system are divided into
two groups. That are overhead line OHL faults (or faults on
branches) and transformer faults (faults on substations). In the
transmission system of Vietnam given in VI Master Plan [10]
for the year 2008, 63 substations 220kV will be seen as load
nodes for voltage sag assessment. The transmission system
(Fig. 2) includes the 500kV network (11 nodes as 500kV
substation and 17 branches of OHL with total length of
3246km) and the 220kV network (63 nodes as 220kV
substations and 103 branches of 220kV OHL with total length
of 6414km). In Figure 2, the number of 220kV substation is 51
that are under the management of National Power
Transmission Corporation (NPT). Other twelve 220kV
substations are under the management of power generation.
Therefore, transformer fault positions will be 11 for 500kV
substations and 63 for 220kV substations respectively. For
OHL faults, fault positions are selected depending on the
length of each branch. According to the above said principle
of fault position selection, we divide the line branches into
some segments and each segment is represented by one fault
position, normally at one of two ends of the line segment. For
220kV OHL, the line segment length shoud be from 10km to
40km depending on the line branch length. For 500kV OHL,
each line segment should be 50km. In this case study, fault
positions are selected at 76 locations for 500kV OHL and 169
locations for 220kV OHL. Therefore, there are 319 fault
positions in total.
101
eBook for You
220/110/35kV
Mai Dong substation,
2x250MVA, Hanoi
Vietnam National Power Transmission

Corporation
eBook for You
Total 500kV OHL length: 3441km

Total 220kV OHL length: 76541km
Number of 500kV substation: 11
Number of 220kV substation: 51
Total 500kV transformer capacity: 8756MVA
Total 500kV transformer capacity: 14761MVA
Fig 2. The Transmission System of Vietnam in 2008

102
- Fault type: This calculation considers all types of short

circuit with well known contributory percentages of different
fault type are assumed as follows
Single phase to ground (SP-G):
65%
Two phase to ground (PP-G):
10%
Two phase together (P-P):
20%
Three phase to ground (3P-G):
5%
For the transmission system that requires high reliability
and stability, short-circuits are prone to permanent fault.
Therefore, in this work, transitory faults are not considered.
Regional Power
Transmission
Company
Line fault rate

(per km.year)
500kV
220kV
Substation
fault rate
(per year)
PTC1 (North)
0.00093
0.02504
0.0397
PTC2 (North Center)
0.00562
0.00536
0.0408
PTC3 (South Center)
0.00173
0.01279
0.0161
PTC4 (South)
0.0077
0.00808
0.0229
NPT
0.00407
0.01478
0.0306
It is noticeable that the fault rates stated in Table 1 are for

all four fault types as mentioned above. Therefore, for each
fault type, the fault rate should multiply by contributory
percentage of different fault types. For the fault that represents
OHL faults within a line segment, fault rate should be
calculated based on the length of the line segment.
- Selection of load nodes for voltage sag calculation: In the
transmission system, load nodes are 220kV substations
feeding to downstream 110kV and medium voltage networks.
The topology of transmission network is complicated and
many branches also have switching devices at both ends.
When a fault occurs on a certain branch (a line or a
transformer), the two switching devices at both ends of that
branch will trip and isolate it from the network. Therefore,
many load nodes normally experience voltage sags. Only the
loads on or nearby the fault position (for transformer fault)
suffers an interruption. So, voltage sags at all 63 load nodes
had to be considered in this work.
- System loading condition when faults occur: It is also
notable that for short-circuit calculation in the transmission
system where limited power sources are connects to, the shortcircuit current and voltage sags depend heavily on the prefault loading condition when the fault occurs. The heavier the
Select the load node (among

63 nodes) for sag calculation
Select the fault position

(among 319 positions)
Select the fault type

(SP-G, PP-G, P-P, 3P-G)
Short-circuit
calculation and
determine voltage sag
magnitude at selected
load node by PSS/E
Fault distribution
modeling, determine
fault rate of the fault
under calculation
Calculate the frequency

of voltage sag at
the selected load node
Sag frequency
spectrum by
the fault under
calculation
(event index)
Sag frequency
spectrum at
selected load
node by all
faults
Sag frequency
spectrum at all
load nodes by
all faults
(site index)
Fig 3.
Block diagram
of the problem
of prediction of
voltage sag in
the transmission
system of
Vietnam.
Are
all fault type
selected ?
eBook for You
- Fault rate: The occurrence of short circuits depends on

many factors [3] and the rates of occurrence of different faults
(fault position, fault type) are normally not the same.
However, because, in reality, recorded fault data does not
consider detailed fault distribution, this work assumes that
fault distribution for each fault type follows uniform model
within each regions in Vietnam. For example, phase-to-ground
faults remain unchanged anywhere in the section of
transmission system within a region. The transmission system
is Vietnam is divided in four regions. The data of fault
performance recorded by NPT and its subsidiary agencies
(Power Transmission Companies, PTC) for 2008 is shown in
the Table 1 below.
TABLE 1. REGIONAL FAULT RATE PERFORMANCE
START
False
True
Are
all fault position
selected ?
False
True
Are
all load nodes
selected ?
False
True
SARFIX
calculation
Check ITIC
curve ?
True
SARFIX-CURVE
(system index)
STOP
103
load on the system is, the higher short-circuit current will be

generated and the deeper voltage sags will be at load nodes.
Therefore, the most interested prefault loading condition is
obviously that of full loaded and this work performs the shortcircuit calculation in the maximum loading condition.
(per unit) intervals for different fault types. Fig. 5 is voltage

sag frequency spectrum for all fault types. Fig. 6 is the
cumulative voltage sag frequency.
Fig 6. Cumulative Voltage Sag Frequency (per year)

at 220kV Mai Dong Substation, Hanoi, Vietnam
For other load nodes, the calculation is similarly performed
and then we obtain voltage sag frequency spectrum of all other
load nodes. Finally, the average frequency spectrum per load
node is calculated and plotted on the Fig. 7 and SARFIX of the
whole transmission system of Vietnam is calculated as the
formula (1). The voltage sag performance of transmission
system SARFIX is shown in Fig. 8.
eBook for You
- SARFIX calculation: With fault distribution modeling for

the transmission system proposed in Part B, this work
performs short-circuit calculation using the program PSS/E for
a certain individual fault (fault position, fault type) and then
voltage sag magnitude at a selected load node is calculated.
After assigning fault rate to this fault, the frequency of sag at
the selected load node resulted by this fault will be obtained.
By repeating this calculation for all other faults (fault position
and fault type), and gather them together, we obtains the
frequency spectrum of voltage sag with different magnitude
characteristics at the selected load nodes caused by all faults in
the transmission system. Fig. 4, Fig. 5 and Fig. 6 show an
example of voltage sag performance for an individual load
node (220kV Mai Dong substation in Hanoi, Fig. 3). Fig. 4
shows voltage sag frequency spectrum by sag magnitude
NEW
Sag Magnitude
(p.u)
C. Short circuit calculation and voltage sag determination for

the transmission system of Vietnam
Short circuit calculation and voltage sag determination for
the whole transmission system of Vietnam is carried out by
program PSS/E (Power System Simulation for Engineering).
The block diagram of the calculation is depicted in Fig. 3.
SP-G
PP-G
P-P
Sag magnitude
(p.u)
Fig 4. Voltage sag frequency spectrum (per year)
by fault types at load node 220kV Mai Dong substation
Sag Magnitude
(p.u)
3P-G
Fig 7. Transmission system average voltage sag frequency

by magnitude characteristics
SARFIX
SARFIITIC-0.7
SARFISEMI-0.5
Fig 5. Voltage sag frequency spectrum (per year) for all fault
events at 220kV Mai Dong Substation, Hanoi, Vietnam
SARFISEMI-X
Sag Magnitude
(p.u)
Sag Magnitude
(p.u)
SARFIITIC-X
Fig 8. SARFIX and SARFICURVE-X of

the transmission system of Vietnam
104
D. Result Analysis
From the results, therere some following remarks:
- The SARFIX and SARFICURVE-X values obtained from this
calculation are useful for utilities as a benchmark for reducing
the frequency of fault for solving the problem of voltage sag.
This result also helps customers know the voltage sag
performance and choose suitable location for less voltage sag
frequency.
- The frequency of voltage sag as the result of an individual
fault type is proportional to fault rate of that fault type for
shallow sags (Fig. 4).
- Shallow sags (0.7-0.9 p.u) feature a rather high frequency
while the frequency of deep sags is very small. Furthermore,
the frequency of voltage sag with X lower than 0.9 for either
the 220kV Mai Dong substation (about 33 times, Fig. 5) and
the system average load node (about 22 times, Fig. 7) is also
khanh
Sag magnitude
(p.u)
Sag magnitude
(p.u)
Figure 9. The difference between ITIC curve and SEMI curve.

Within this range, ITIC ridethrough voltage is 0.7 p.u whereas
this voltage level for SEMI F47 is just 0.5 p.u. Therefore, with
the total fault clearing time (160ms to 180ms) for the
transmission system in Vietnam, only voltages sag with X
lower than 0.5 p.u are qualified for SARFICURVE-X using the
SEMI curve (SARFISEMI-X). With X greater than 0.5 p.u,
voltage sags fall inside SEMIs ridethorugh area and not
qualified for SARFISEMI-X. So, for X from 0.5 p.u to 0.9 p.u,
the value of SARFISEMI-X remains unchanged and equal to
SARFIITIC-0.5. SARFISEMI-X is also shown on Fig 8.
Sag magnitude
(p.u)
Fig 9. Voltage sag frequency of selective load nodes

(220kV substations) throughout of Vietnam
105
eBook for You
- SARFIITIC-X calculation: SARFIX-CURVE can be achieved by

taking fault clearing time of protective system into account.
For the transmission system of Vietnam, the primary functions
currently used for transformer protection is biased differential
protection using differential relays of SIEMENS (SIPROTEC
7UT613) or ALSTOM (MiCOM P340). For OHL line
protection, the primary functions currently in use are also the
differential protection as above said using the telecommunication links of power line carrier or fibre-optical
ground wire integrated in power carrying lines or the distance
protection using differential relays of SIEMENS (SIPROTEC
7SA6) or ALSTOM (EPAC 3000, MiCOM P440). All those
protective relay system is of instantaneous tripping type that is
typically less than 100ms. The switching devices are almost
SIEMENS, SCHNEIDER or ABB products manufactured in
Europe with typical breaking time of 40ms for 500kV to 60ms
for 220kV circuit breakers. Besides the above mentioned
operating times of protective relays and circuit breakers,
additional time delays are also included for auxiliary relay
trips and operating time of tele-protection with total additional
operating time not exceeding two more cycles (20-24ms).
Therefore, the total fault clearing time is 160ms to 180ms that
defines the voltage sag duration. If posing this duration on the
ITIC curve, its obviously that only sags lower than 0.7 p.u.
will be out of load voltage tolerance and qualified for
SARFIITIC-X. The upper 0.7 p.u. sags with duration defined by
the above said fault clearing time definitely fall inside the
voltage tolerance envelope and thus, they are not qualified as
SARFIITIC-X. Therefore, SARFIITIC-X is a part of SARFIX with
X lower than 0.7 p.u. as also shown on the SARFIX chart (Fig.
8). For X from 0.7 p.u to 0.9 p.u, the value of SARFIITIC-X
remains unchanged and equal to SARFIITIC-0.7.
If we use SEMI curve for assessment of sag duration, it is
noticeable that there is a small difference between ITIC curve
and SEMI curve for X from 0.5 cycle to 10 cycles (Fig. 9).
IV. CONCLUSIONS
This paper presented the first effort of predicting voltage
sag performance for a large transmission system as a case
study of Vietnam. In this work, fault distribution modeling is
proposed basing on actual fault performance for different
regions in Vietnam. Using SARFIITIC-X gives a better
assessment of voltage sag influence on loads operation. The
results of this work will be a useful reference for utilities in
power system quality assessment toward electricity market
operation. This research still needs to develop as faults in the
generation part has yet to take into consideration. Besides, if a
better fault data is achieved (by monitoring), a more detailed
fault distribution can be made and finally a better voltage sag
performance can be obtained.
[3]
Bach Quoc Khanh, Dong Jun Won, Seung Il Moon, Fault Distribution
Modeling Using Stochastic Bivariate Models For Prediction of Voltage
Sag in Distribution Systems, IEEE Trans. Power Delivery, pp. 347354, Vol.23, No.1, January 2008.
[4] D. L. Brooks, R. C. Dugan, Marek Waclawiak, Ashok Sundaram,
Indices for Assessing Utility Distribution System RMS Variation
Performance, IEEE Trans. Power Delivery, vol.13, no.1, pp.254-259,
Jan. 1998.
[5] Juan A. Martinez, Jacinto Martin-Arnedo, Voltage Sag Studies in
Distribution Networks - Part II: Voltage Sag Assessment, Part III Voltage Sag Index Calculation, IEEE Trans. Power Delivery, pp. 16791697, Vol. 21, No. 3, July 2006.
[6] Bach Quoc Khanh, Prediction of Voltage Sags in Distribution Systems
With Regard to Tripping Time of Protective Devices, Proceeding,
EEE.CR.ASPES2009, Tech. Section 2.1., Hua Hin, Thailand, September
28-29, 2009.
[7] J.V.Milanovic, M.T.Aung and C.P.Gupta, The Influence of Fault
Distribution on Stochastic Prediction of Voltage Sags, IEEE Trans.
Power Delivery, vol.20, no.1, pp.278-285, Jan. 2005.
[8] P. Saninta, S. Premrudeepreechacharn Assessment and prediction of
voltage sag in transmission system in northern area of Thailand,
Proceeding, 13th International Conference Harmonics and Quality of
Power, ICHQP, Sept.28-Oct.1 2008, Wollongong, NSW, Australia.
[9] E Inan, B. Alboyaci, C. Leth Bak, A Case Study Of Turkish
Transmission System For Voltage Dips, The Journal on Power and
Energy Engineering, Vol 1, No. 2, April 2010
[10] National Institute of Vietnam, Master Plan VI, 2006.
VII. BIOGRAPHIES
Bach Quoc Khanh received B.S., M.S.
and Ph.D. degrees in power network and
systems from Hanoi University of Science
and Technology, Hanoi, Vietnam in 1994,
1997 and 2001 respectively. He has been a
faulty member of Electric Power System
dept., Electricity Faculty, Hanoi University
of Science and Technology since 2002. He
is currently a visiting scholar in the Center
for Advanced Power System, IESES,
Florida State University. His research
interests include power system analysis,
DSM, power system quality, distributed
generation.
Nguyen Hong Phuc received BS in Electrical Engineering Faculty,
University of Thai Nguyen in 2006, Vietnam. He is currently a master student
in the Electric Power Systems Department, Electricity Faculty, Hanoi
University of Science and Technology, Hanoi, Vietnam.
V. ACKNOWLEDGEMENT
It is acknowlegded that this work received helps from
Phung The Anh (Msc), Nguyen Anh Tu (Msc) with Power
Engineering Consulting Joint-Stock Company No. 1, Hanoi,
Vietnam for data collection and power system simulation as
well as technical consultancy from Professor David A. Cartes,
Senior Member, IEEE, and Dr. Bhuvaneswari Ramachandran
with the Center for Advanced Power Systems, Institute for
Energy Systems, Economics and Sustainability, Florida State
University, USA.
VI. REFERENCES
[1]
[2]
M.H.J. Bollen, Understanding power quality problems - voltage sags and

interruptions, IEEE Press, 2000.
M.R.Qader, M.H.J.Bollen, and R.N.Allan, Stochastic Prediction of
Voltage Sags in a Large Transmission System, IEEE Trans. Industry
Applications, vol.35, no.1, pp.152-162, Jan./Feb. 1999,
106
eBook for You
much lower than total faults in transmission system (about 110

times per year). Thats because the goegraphical shape of
Vietnam is rather long (about 1700 km) and thin in the middle
(the narrowest is just 60 km) and the short-circuit faults occur
on one region has almost no impacts on voltage sag for loads
in other regions.
- Cumulative frequency of voltage sag for the node 220kV
Mai Dong substation in Hanoi (33 times for X < 0.9 p.u) is
higher than the SARFI0.9 (22 times) because therere more
load nodes (220kV substations) located surrounding Hanoi
and vicinity. In the center and in the south of Vietnam, the
density of 220kV substation is lower than in the north and
faults has less impact on voltage sag of load nodes. Fig. 10
shows sag frequency charts for selective load nodes in the
north (upper), in the center (in the middle) and in the south
(lower) of Vietnam that indicates the above said difference.
- Also because of high frequency of 0.7-0.9 p.u. voltage sag,
SARFIX-CURVE is very much lower than SARFIX despite
voltage sags with the magnitude up to 0.7 p.u are qualified
enough for SARFIX-CURVE. Therefore, voltage sags due to
faults in the transmission system of Vietnam have less
influence on loads than faults in distribution system when the
frequency of deep sag is normally very high [6]. It is a
remarkable finding in power quality assessment in the power
system of Vietnam.
&&&&"#'- $-$!'- - # &'% )' $#- *&'"- % #- - #%( #'*$- #- &- &'% )'- #%' $#-
/1.1#$08 ++-8 +-0)208 8 )58 0000()28 * 8 4*'2"8 0"8

38 2*8 3'208 $)8 $02-$32$*)8 0602(8 *))28 5$2#8 5$)8 08
$02-$328 ")-2$*)8 8 $)8 5#$#8 0*(8 3)-2$)8 2*-08 * 8
2#$08 8 2#)*'*"68 -8 *)0$-8 *'2"8 0"8 #-2-$02$08
+)8 *)8 2#8 0602(08 0#*-2$-3$28 +*5-8 2#28 $)'308 2#8
*)2-$32$*)8 * 8 8 +*5-8 $)%2$*)8 5#)8
3'208 *3-8 #8
*32+328 +*5-8 +-*38 68 5$)8 08 08 $08 8 3)2$*)8 * 8 2#8
5$)8 0+8 2#28 $08 02*#02$8 $)8 )23-8 68 30$)"8 2#8 +-*$'$02$8
(*'$)"8 * 8 5$)8 0+8 28 2#8 8 $)02''2$*)8 0$28 2#8
3)-2$)268 * 8 5$)8 -"$(8 $08 3$'28 $)2*8 2#8 +-*'(8 * 8 02*#02$8
+-$2$*)8 * 8 4*'2"8 0"8 $)8$02-$32$*)80602(8 5$2#8 5$)8 )-"68
-0*3-08*))28 #8 ++-8 000008 4*'2"8 0"8 +- *-()8
$)8 8 EH *08 028 -8 8 *))28 )25*-&8 (*'$)"8
)8 3'28 )'60$08 -8 *)8 30$)"8
8 M&3& *'2"8 0"8
#-2-$02$08 )8 -,3)68 -8 '0*8 2-($)8 $)8 -'2$*)8 2*8
5$)8 0+8 *-8 $!-)28 0)-$*08 * 8 8 0$7$)"8 )8 +'()28
26+08 * 8 5$)8 +*5-8 ")-2$*)8 )8 2#8 $(+28 * 8 2#8 0602(8
0*3-8 +$268* 82#8 $02-$32$*)80602(8* 8$)2-028
g s $02-$328 )-2$*)8 $02-$32$*)8 602(8

*5-8 3'$268 *'2"8 "8 $)8 088
. .4=;6&?#=.64
*1* RlRN{dN xvR{ Rm F{R Rvldq^ Zvm

{FPddvqFll NRq{FldRP Llj Rm d` ^RqR{Fdvq
N
xlFqNvqqRNRPv`R{FqmddvqqRv{j vF\{Rmv{R
PRNRq{FldRP Rm d` mFllR{ ^RqR{Fdq^ qd qFmRl
Pd{dLRP ^RqR{Fdvq &* NvqqRNRP v Pd{dLdvq
qRv{jv{RRqPd{RNlv`RxvR{Nvqmxdvq "RdPRF
F{dR vT RN`qdNFl FqP RNvqvmdN LRqRX v dlddR RqP
R{ FqP `R Rqd{vqmRq `R dqFllFdvq vT &* dq xvR{
RmFlv{FdRN`FllRq^RvdlddR dqF{dvFxRNvq
qRv{jxlFqqdq^PRd^qFqPvxR{Fdvq=`R&*N`FllRq^Rv
qRv{j vxR{Fdvq F{R PR v `R TFN `F `R Pd{dLdvq
Rmd{FPddvqFllPRd^qRPd`dq^lRRqR{^v{NRFqP
`Rx{RRqNRvT&* N`Fq^RRmT{RzRqNxvR{Yvdq
Rd`R{ RFP FR v{ {FqdRq FR !mvq^ qmLR{ vT
vxR{FdvqFl dR NFRP L &* dqvlRmRq dmxFN vq
xvR{zFld`FRLRRqvT^{RFNvqNR}vdlddR
:GA$?/>BN @'$'/J'%N '"?G ?MN
@9 @4ef% ,/AN L<?2N /AN AG>><?C'%N "MN
0'B: 7N %H$!B0<:N <H:& B0<:N (<?N B,'N 0A/B0:*N $,<4!?N ?<*@ 7N @4:4@4dd$
$,N G<$N , :,N /AN L/B-N 5'$D?/$N <L'?N MAE'8AN '> ?B9':BN 5'$D?/$ 5N
:*/:''@1:*N $G6BMN :</N :1J'@A1FMN <)N$0':$'N :%N '$-:<5<*MN
0'BN % N !:</N /'B:!7N '7!04N
XSU;IOSS5SSQQ-;;-AQ(55A5;;
B
mn H thng in - i hc Bch Khoah___
H Ni
e /N =N
;R^F{Pdq^ `R x{vLlRm vT vlF^R F^ N`F{FNR{ddN vT

vlF^R F^ dq Pd{dLdvq qRv{j NFRP L TFl F{R
`d^`l dqYRqNRP L Rm `v{ Nd{Nd xvR{ Yvdq^
`{v^` `R TFl F Rll F `R NvqX^{Fdvq vT x{vRNdvq
Rm !ll `RR FxRN F{R Flv FTTRNRP L `R x{RRqNR vT
&* DE FqP DE F{R Fmvq^ mFq {RRF{N`R Rxlv{dq^
dmxFNvT&*vq`v{ Nd{NdFqPvlF^RF^xR{Tv{mFqNRdq
Pd{dLdvq Rm ,vRR{ `RR v{j mFdql NvqdPR{RP
&* F NvqRqdvqFl qN`{vqv ^RqR{Fv{ {R^F{PlR vT
RqR{^ NvqR{dvq RN`qvlv^dR RxRNdFll {RqRFLlR RqR{^
{Rv{NRldjRdqPxvR{ <vmRv`R{NvqdPR{RPdmxFNvT
&* RN`qvlv^dR vq x{vLlRm vT xvR{ Rm L `R
TvNRP mFdql vq RFP FR x{vLlRm N` F FqFld vT
vlF^R FLdld RqR{^ lvR DE v{ {RldFLdld DE DE
=`R mvdFdvq vT `d v{j NvmR Zvm `R TFN `F vlF^R
F^ NFRP L TFl PRxRqP vq `R Rm `v{ Nd{Nd
xvR{ <<#: `F dqn PRxRqP dqxF{ vq xvR{dqhRNdvq
Zvm &* `RqTFlvNN{ Bd`PdTTR{Rq&*RN`qvlv^dR
`R Fmvq vT xvR{ dqhRNRP Zvm &* mF F{
Nv{{RxvqPdq^ v vmR qNR{Fdq TFNv{ fq`R{Rq dq `R &*
=`d v{j dll x{RRq F qR FqFld `F NvqdPR{ `R
qNR{FdqvTdqPRqR{^{Rv{NReq`Rx{vLlRmvTvlF^R
F^ FRmRq Tv{F Pd{dLdvq RmNvqqRNRP d`dqP
LFRP&*B7*
..MOG,C4D348=F=T=GF
J ,?<E175J(17JDD5DD=5>EJ
@vlF^R F^FRmRq d F xvR{ zFld x{vLlRm `F d

Fll{RlFRPvFLFdNx{vNRjqvqF`R NvmxFdLdld
FRmRq DE dq `dN` vLFdqdq^ `R vlF^R F^
xR{Tv{mFqNR vT `R Rm vT dqR{Rd vqR vT N{ddNFlRx
Bd` {RNRqlFPFqNRP NvmxR{ FdPRP dmlFdvqvvl `R
vN`FdN x{RPdNdvq vT vlF^R F^ LRNvmR `R x{RTR{FLlR
Fxx{vFN` (v{ vlF^R F^ NFRP L TFl `R mR`vP vT
TFlxvddvqFrP`RmR`vPvTN{ddNFlPdFqNRF{Rjqvq
F `R mv dPRl RP mR`vP Tv{ vN`FdN x{RPdNdvq
PdR4vmFR{`FmR`vPdRPFvN`FdNx{RPdNdvq
P FlF FvNdFR d` v dmxv{Fq Fj F `R
TFl mvPRldq^ FqP `R `v{ Nd{Nd NFlNlFdvq vT `R xvR{
Rm vT dqR{R 6q `d x{Rdv {RRF{N` DE F`v{ `F
PRRlvxRPx{vNRP{RTv{vN`FdNx{RPdNdvqvTvlF^RF^
dq Pd{dLdvq Rm d` PRFdlRP TFl Pd{dLdvq
mvPRldq^ FqFld FqP NvmxFdvq vT vlF^R F^ dqPdNR F
107
eBook for You

PdTTR{Rq lRRl vT xR{TvuFqNR RRq dR FqP Rm

dqPdNR (v{`dv{jdqNRTvNdmFdqlvq`RdmxFNvT
B89 qNR{Fdq vq vlF^R F^ vql RRq dqPdNR vT
vlF^R F^ F{R NvqdPR{RP =`F mRFq `R TFl xR{TvuFqNR
d mvPRlRP Tv{ F RlRNRP xvddvq FqP `R Nv{{RxvqPdq^
vlF^R F^ F F RlRNRP dR F{R vLFdqRP Tv{ FqFld =`R
vlF^R F^ FRmRq d PdNRP dq `R {RlFdvq d`
B89qNR{FdqTFNv{
J *5DEJ(IDE5=J -9E8J -9>4J1D54J
=`d v{j N`vvR `R .''' 4vPR =R (RRPR{ DE F

`R R Rm . d mvPRlRP L <dRmRq :vR{ <Rm
<dmlGv{ !PFqNRP &d{dLdvq 'q^dqRR{dq^ :{vPNdd
=vvl:<<!&':=FRxv{RPdq(d^ :<<!&':=
dFqRTTRNdRvvlTv{Pd{dLdvqRmmvPRldq^FqPdF
x{vRP xvdLlRTv{mvPRldq^vT .''' 4vPR=R (RRPR{

dq DE B89RNaqvlv^NFqFlvLRmvPRlRPL:<<!&':=
d` dmxldXRP mvPRl `F F{R dll FNNRxFLlR Tv{ `v{
Nd{Nd FqFld ;Rxv{ PRldR{RPL :<<!&':= Tv{L{FqN`
N{{Rq FqP qvPFl vlF^R dq `v{ Nd{Nd FqFld `Rlx
TFRq `R x{vNR vT qRv{j mvPRldq^ FqP FqFld vT &9
dmxFN<vmR{RmF{jTv{`dmvPRldq^F{RFTvllv
:<<!&':=vqlFNNRxRzRqNRFlRTv{`{RR x`FR
ldqRmvPRldq^ v`R fs`R{RqqRzFl mFl Nvxldq^dq`R
x`FR dmxRPFqNR mF{dNR Tv{ Pd{dLdvq ldqR DE d qv
NvqdPR{RP.qTFNddFNNRxFLlRTv{`v{ Nd{NdFqFld
&d{dLRP lvFP ldqR F{R mvPRlRP dq^ lmxdq^ lvFP
mvPRldq^ DE =`R{RTv{R F qmLR{ vT XNdvqFl lvFP qvPR
lFLRldq^!F{RFPPddvqFll^RqR{FRPdq`RRRm
eBook for You
r
"
8nj<Tld Ftcdz=444Tdz|8ddcdxn|lZ1:zrtcdqdctsKRR*14KTN'4
B89d`{RR x`FRRlRmRq ddvqlxlFNRPvq`RmFdq
TRRPR{ FqP `R `{RR x`FR lFR{Fl &9 lvNFdvq d qvuFll
PRRudqRP LFdq^ vq RR{Fl vxdmFl xvR{ Yv xlFqqdq^
TvulFdvq DE ,vRR{ Tv{ `R x{xvR vT `d v{j
B89 d RlRNdRl lvNFRP F qvPR v{ F
NFR PdR 4vPR TRF{R `R &9 qRF{ Rm v{NR
qvPR FqP F{Rdq`RmdPPlR vT mFdqTRRPR{FqPqvPR
ddqFa{RR x`FRlFR{FlRqP &9dNvqqRNRPvmRPdm
vlF^R TRRPR{ `{v^` Pd{dLdvq {FqTvuR{ (v{ qvPR
`d xFxR{ mvPdXR `R dR vT C3( {FqTvuR{ F
j@!Tv{NvqdPR{ds^`R&9d`xRqR{Fdvq
4vFPF`{RRxRvTB89F{RRlljqvq DE DEF
XRP xRRP dqPNdvq mFN`dqR F{dFLlR xRRP PvLl TRP
dqPNdvq mFN`dqR FqP F{dFLlR xRRP qNb{vqv ^RqR{Fv{
&RFdlRP mvPRl vT `RR `{RR xR vT dqP xvR{ {Rv{NR
`FR LRRq x{RRqRP dq F{dv xLldNFdvq ,vRR{ d`
{R^F{P v &9 {RxvqR v TFl dq Pd{dLdvq Rm `R
`v{ Nd{Nd xvR{ dqhRNRP L &9 dq `R qRv{j `Rq F
TFlvNN{mFdqlPRxRqPvq`RxRvTmFN`dqRPRd^qRP
Tv{ B89 .mxFN vT xvR{ RlRN{vqdN TFNdlddR dq `R
dqR{NvqqRNdvqLRRRqB79FqP`RqRv{jF{RqR^ld^dLlR
=`R{RTv{R `dxFxR{R`RmvPRlvTqN`{vqvmFN`dqR
FqPdqPNdvqmFN`dqR FFdlFLlRdq:<<!&':=FdmxldXRP
mvPRlvTB89xR
... EG14C?F:9*UCT6W4FTB/>-&9VF-4OT*=FT\
J 56>9>7J (35>1C9?DJ?CJ(IDE5=J(8?CE9C3G9EJ%?H5CJ!>J
9DEC92GE9?>J(IDE5=J -9E8J -9>4J1D54J
=`R N`H{FNR{ddN vT vlF^R F^ mF^qdPR F F lvFP qvPR

d{RlRPLFqRRqvT`v{ NgNd dqxvR{Rm =`R{R
F{R F{dv TFNv{ FTTRNdq^ `R`v{ Nd{Nd N~Rq FqP `R
NFq LR NlFdXRP dqv v ^{vx (d{ ^{vx vT TFNv{ F{R
`F dqYRqNdq^ `R qRv{j NvqX^{Fdvq ldjR qRv{j
vxvlv^ RlRmRqxR qR{FlRF{`dq^mR`vP `R{RF
`Rv`R{ ^{vx vT TFNv{ F{R `FNFdq^F{dFdvq v <<#:
N` F &9xvR{ dqhRNdvq PqFmdN lvFP Rm v{NR
NFxFNd Bd``RdqFllFdvqvT&9dq`RqRv{j <<#:d
NvqRzRql N`Fq^RP =`R <<#: PRxRqP vq `R x{R TFl
RFP FRvT `R Pd{dLdvqRm dq `dN`RlRN{dNlvFP
NFq LR TRP L Rd`R{ `R Rm v{NR T{vm {Fqmddvq
Rm FqP &9 .q `R RFP FR `R xvR{ ^RqR{FRP L
&9 mF LR Nvq{vllRP L `R vxR{Fv{ ,vRR{ Tv{ `R
&9 P{dRq L{RqRFLlR RqR{^ {Rv{NR ldjR dqP RqR{^
`R^RqR{Fv{Nvq{vlF{RPRd^qRPvvxR{FRdq`RmvPR`F
NFq x{vPNR `R lF{^R xvdLlR RqR{^ dq v{PR{ v mFdmdR
RNvqvmdNFPFqF^R =`R mv N{ddNFlTFNv{dqYRqNdq^`R
xvR{ PRRlvxRP L F dqP RqR{^ NvqR{dvq Rm d `R
108
dqP RlvNd FqP &* vx xvR{ NFq LR mvPRlRP F F

TqNdvq vT eqP RlvNd =`R{RTv{R `d xFxR{ dll NvqdPR{
`R vlF^R F^ PR v TFl eq Pd{dLdvq Rm dq `R
{RlFdvqd`B&*dqPxRRPF{dFdvq
;R^F{Pdq^ `R xvR{ N{R vT `R {LdqR `R N{R `FxR
NFqF{dRPLFfq^vqxRRPNvq{vlxRxdN`Nvq{vlv{Fll
Nvq{vl <vTF{x{vxvFlTv{mvPRldq^`d{RlFdvq`FRLRRq
FNNRxdq^`RxvlqvmdFlTv{mFdvq DE DE DE DEFqP`d
v{j R F xdNFl xvR{ N{R vT F xdN` Nvq{vllRP dqP
{LdqR(d^F`RTvllvdq^RzFdvq DE
!
47
!7
J ?>D945C9>7J -9>4J +>35CE19>FJ!>J1G<EJ"?45<9>7J
8njOdq_|ntsad|ddsnsczvddcY _scZHIzt|v|vtdxL

$% ,7 , 57, 5
37
*%7 &7
"7
&7

`R{R
#7 ;FRPvxxvR{vT B&*
# dqdqPxRRP4v{mFll d o
BdqP xRRP F `dN` `R B&* vx xvR{ {RFN`R
{FRPFlR=`dFlRv[RqTFlldq`R{Fq^RvT o
x~Y # vdqPxRRP=`dFlRNFqLR o
1@J 1J FqPIF{RNvqFq`FF{RNFlNlFRPFTvllv DE

,17
,7
,7

7
7
7

7

7

=`d v{j NvqdPR{ `R Tvllvdq^ PdTTR{Rq NRqF{dv vT

B&* FxxldNFdvq `F dmxFN `R B&* xvR{
Nvq{dLdvq v <<#: .q RFN` NRqF{dv vlF^R F^
N`F{FNR{ddN F{R PRR{mdqRP dq {RlFdvq d` dqP xRRP
F{dFdvqeq`R{Fq^RFmRqdvqRPdq
;R^F{Pdq^B&*lvNFdvqTv{dR qvPR
FqP F{R Fjdq^dqvNvqdPR{FdvqTv{NvqqRNdq^d`&*
{Fdq^jB &*x{R TFlx{vPNRPxvR{djB
;R^F{Pdq^ B&* xRqR{Fdvq lRRl B&* d` {Feq^
jBjBFqPjBFqvPRdFxxldRP
;R^F{Pdq^ `R dmxFN vT Rm v{NR NFxFLdld `{RR
xvdLdlddR d` v{NR dmxRPFqNR x {vq^ x
{F`R{ {vq^ FqP x RFj F{R PdRP =`R jB
{FRPB&*d`x{R TFlTlllvFPdq^dlvNFRPFqvPR
Bd`v &*v{d` &* `F dqhRN F NvqFq xvR{ vR{

`R dmR dq Pd{dLdvq Rm F TFl RRq NFq vql LR
PRXqRP L TFl xvddvq FqP TFl xR ({vm `F PRXqddvq
`v{ Nd{NdNFlNlFdvqdxR{Tv{mRPFqP`R{Rldq^vlF^R
F^N`F{FNR{ddNFlvFPqvPRF{RvLFdqRP =`dvlF^RF^
xR{Tv{mFqNR d qdzR Tv{ `d TFl RRq FqP vlF^R F^
ZRzRqNd ` RzFlv`RTFl{FRFd^qRPv `R^dRq
TFl RRq ,vRR{ d` `R RdRqNR vT B&* dq
Pd{dLdvq Rm `R FLvR mR`vP vT vlF^R F^
zFqdXNFdvqdqvlvq^R{`RNFR &Rv`RN`Fq^ReqdqP
xRRP vR{ `R dmR `R Fmvq vT xvR{ xxldRP L &* v
`RqRv{jN`Fq^RFNNv{Pfs^l !F{RlRRq`v^``R
TFlRRqF{RPRXqRPL`RFmRTFlxRFqPF`RFmR
TFl xvddvq `Rd{vNN{{RqNRF `R PdTTR{Rq dqFqNFR
PdTTR{Rq vlF^R F^ N`F{FNR{ddN <v Fq FPPddvqFl TFNv{
`vlP LR dqNlPRP `Rq PRXqdq^ F TFl RRq =`F d `R
x{vLFLdld vT dqP xRRP F B&* lvNFdvq `F Nv{{RxvqP
`RdqFq`RTFlRRqvNN{ .qv`R{v{P FTFlRRq
d` TFl xvddvq TFl xR FqP Fd^qRP TFl {FR
x{Rdvl PRXqRP dq `R NFR d`v B&* `vlP LR
L{RFjPvq \{`R{ dq `R v NFllRP L RRq vT TFl d`
{R^F{P v `R dqP {R^dmR F `R dR vT B&* dqFllFdvq
BfsP d vN`FdN dq qF{R FqP dqP xRRP {R^dmR NFq LR
mFqdTRRPdqvN`FdNmvPRl @F{dvx{vLFLdldTqNdvq
R{R XRP d` `R XRlP PFF v dPRqd] dFLlR FddNFl
Pd{dLdvqTv{{Rx{RRqdq^dqP{R^dmR .dTvqP`F`R
BRdLll Pd{dLdvq NFq LR RP v PRN{dLR `R dqP
F{dFdvq dq F {R^dmR d` Fq FNNRxFLlR FNN{FN lRRl DE
DE .q BRdLll Pd{dLdvq `R F{dFdvq dq dqP xRRP F{R
N`F{FNR{dRPL `Rx{vLFLdldPRqdTqNdvqxP FqP
`RNmlFdRPd{dLdvq\qNdvq

)

0-7

`R{R
/77 <`FxRdqPR J7 <NFlRfqPR
=`R {RldFLdld vT BRdLll xPT dq dqP {R^dmR FqFld
109
eBook for You

-
;R^F{Pdq^xR vTB&* v xR
dqPNdvqmFN`dqR
FqPqN`{vqvmFN`dqRd``RFmR{Fdq^ jB Tll
lvFPRPF{RRRPTv{NvqqRNdvqv`RqvPR
"Fdq^ vq `R FLvR x{vxvRP NRqF{dv vlF^R F^
N`F{FNR{ddN F qvPR d NvmxRP d` `R `v{ Nd{Nd
dFxxldRPFqvPR !l`v^`qvPRd qvFlvFPqvPR
d d RRq F F N{ddNFl qvPR `Rq NvtsRNRP d` B&* FqP
vlF^RF^NFqLRNvqdPR{RPF`dlvNFdvq !llxvdLdlddR
vT TFl xR <dq^lR ldqR v ^{vqP lfqR v ldqR v ldqR v
^{vqPFqP `{RR x`FR v ^{vqP F{RNvqdPR{RP . d Flv
qvdNRFLlR `F Tv{ `R x{xvR vT `d v{j vlF^R F^
xR{Tv{mFqNR d vql PRR{mdqRP LFdq^ vq `R RRq dqPR
=`R Rm dqPR NFq LR vLFdqRP F[R{ fmdlF{ RRq dqPR
NFlNlFdvq F{R mFPR Tv{ Fll TFl RRq dq `R Rm vT
dqR{RFqPdRdqPdNRvTFlllvFPqvPRF{RdqR^{FRP
J
PSxSqP wq `S JOO{JO dq SdmJdq_ JqP ] )d_

dqPSOJqPSUpS`SPd{dMdwq`JxS ;J dOJllSP;JlSd_`
xPT `dO` x{SSq `S mw {SJlddO Jxx{wdmJdwq wT dqP
xSSP x{wXlS =`S JlS wT ] OJq __S `S JS{J_S dqP
xSSP =`STwllwdq_)d_ PSxdOBSdMllxPTd`JlSwT
] JqP `JJ{SSPJOJSPdSdq`dw{jTw{JqJldq_
`SdmxJOwTdqPqOS{JdqwqwlJ_SJ_xS{Tw{mJqOS
`
a
cw
` aw
y
b
^_
m
o{
??T2
82
7W22
8L>2
8 2
RP2
DT2
2
cd+
z
hi
v
ef+x
8nj Tt|_qf_q|x_|d_sc[1:zt|v|vtdx_j_nsz|nsczvddc
)7
HP
8nk C! [enaqqvcf
`
4 `xnec@ O`qenklcn{|xna|nus
a
4 #4`xnec

B`S{S
5A TJl ZSzSqO wT `S TJl SSq 9J `J wOO{ `Sq
dqPxSSPJB&*lwOJdwqSzFl z
5 =wFlTFl{FSFFTFlxwddvqF#J J
'7 :S{NSqF^S wT `S TFl xS Nw{{SxwqPdq^ w `S TFl
SSq 9 J ;STS{{dq_w /0J `dw{jFmS`SxS{NSqF^SwT
`STFlxSdq`SSSmdJTwllw
<dq_lSx`FSw_{wqP:4
J
=wx`JSw_{wqP::4
J
=wx`FSw^S`S{::
J
=`{SSx`FSv^{wqP:4
J
i@ BSdMllxPTF`SdqPxSSP y
)d^ J dll{FSFNFSPdq`dN``SPd{dMdwqwTTFl
{JSwT`STJlSSqJqwPS J J_JdqdqPxSSPJqPd
{Sldq_ wx xwS{ wT B&* lwOJSP J qwPS J dq_
BSdMll Pd{dMdwq d` ;J JqP $ =`S xwS{ O{S wT
`SdqP{MdqS J dOwqdPS{SPd` %& :.J ,=DJ
,&=DJ JqP ,A=DJ "Jdq^wq`S{SlJdwqMSSSqTFl
Pd|dMdwq JqP dqP xSSP JqP `S xwS{ O|S wT dqP
{MdqS SNJqd`P{J`S {SlJdwqMSSSqB&* wx
xwS{JqP`SNw{{SxwqPdq_TJl{FS)d_
6T Ow{S `S JMwS wJl TJl {JS Pd|dMdwq J_Jdq
B&* wx xwS{ dll MS M{wjSq Pwq \|`S{ Tw{
dqPddPFl TJl xS FNNw{Pdq_ w (7 Tw{ wMJdqdq^ `S
wOO{|SqOS{FSwTdqPddPFlTFlSSq
[

XYZ
%
%
n.+7
8nj N' Tlexeq`|ntsae|des|t|`qf`q|x`|d`|stce% "

_sc[J:zt|v|vtex
.dFlwqwdOSFMlS`F`STFlP{FdwqdPSXqSPMTFl
NlSF{dq_ dmS FqP dq Pd{dMdwq Sm d PwS qw lF S{
lwq_ xdOFll lS `F J SOwqP Bd`dq O` `w{ dmS
P{Fdwq dd xwdMlSwFmS`F`SFlS wTdqPxSSP
dqN`Fq^SP=`S{STv{SB&*wxxwS{dFlwFMlSP
Tw{ dqPddPFl TFl SSq FqP {Sldq^ wlF^S F^
mF_qdPSF{SXSPTw{SFN`TFlSSq
.@ 0*R7STU2=7R0*C-UC*T=GF*F2QRUCT+F*C\R=R
"Fdq_ wq `S FMwS x{wxwSP TFl SSq mwPSldq_ d`
{S_F{P w dqP qNS|Fdq wlF_S F_ N`F{FNS{ddN FqP
wlF_SJ_ZSzSqOTw{`SqwPS J PSw`S`w{ Od{Od
SSqF`SqwPS J F{SPSS{mdqSPMdq^`STqNdwqwT
`w{ Nd{Nd NFlNlFdwq dq :<<!&':= &dTTS{Sq NSqF{dw
x{wxwSP dq <SNdwq ... F{S NwqdPS{SP )wllwdq_ F{S
{SmJ{jJMlS{SlKPJqJld

J !=B13EDJ 7 -9>4J)B554J ,1C91E9?>J
Bd` x{dmF{ TwO wq `S dmxFO wT dqPddPFl FlS wT

dqP xSSP wq wlF_S F_ xS{Tw{mFqOS `S dqP {S^dmS d
FmSP w Twllw `S OFS P F dll{FSP dq )d_ J =`S
{Sldq_ wx xwS{ FlS wT B&* dq PdTTS{SqNSqF{dw
wT B&* JxxldOFdwq F{S PSS{mdqSP M J B&* xS d
mwPSlSP J `S qN`{wqw _SqS{Fw{ )w{ `S OSqJ{dw wT
B&*FxxldNFdvqx{wxwSPdq:F{!wT<SNdwq... `SFMwS
wMFdqSP wx xwS{ FlS wT B&* F{S SP Tw{ `w{
Od{Od OJlOlJdwq JqP `S Ow{{SxwqPdq_ wlJ_S J_
N`J{JNS{ddOJ{SUpJllJN`dSSP
)d_ J PSxdN `S wlJ_S J_ mJ_qdPS wT qwPS J Tw{
wmS PdTTS{Sq TJl xS J qwPS J x`JS ! w _{wqP
x`JS!JqP"w^{wqP x`JS!wx`FS"JqP`{SSx`JS
110
eBook for You
"Jdq^wqdqPxSSPxPT SNJqmwPSl`SPd{dMdwqwT
wNN{{SqNS {FS wT M SSq wT TJl dq Pd|dMdwq Sm
d` B&* .qSJP wT `S JmS Sm `w{Nd{Od xwS{
JxxldSPwJTJlSSqd`J_dSqTJl{JSJJ{dJMlSdqP
xSSP PSxSqPSP Sm `w|Nd{Nd xwS{ d JxxldSP JqP `S
TJl{FSwT`STFlSSq`wlPMSPd{dMSPT{`S{dqldqS
d` dqP xSSP x{wMFMdld Pd|dMdwq mwPSl BSdMll
mwPSl =`S{STw{S {S_F{Pdq_`S TFl SSq FqwPS J Tw{
dqFqNS dT`SwFlTFl{JSFd_qSPw `FTFlxwddwqd
FmSPwMS J +Tw{FqFld `STFl{JSPd{dMSP
wFM SSqwTTFl`FwOO{`SqB&*wxxwS{
d x{wPNSP F F xSNdXN dqP xSSP `wlP MS NFlNlFSP M
FTwllw
C3J
,r
j 6V ;
<
>
-EM
8nj"Wtq|_jdz_j_|stcd%!ftx
cngdxds|f_q||vdz_|stcd% "
GO
8nj4 Wtq|_jdz_j_|stcd%!ftx
cngdxds|ZH:zndz_|stcd%!
(d^$`v`RvlF^RF^FqvPRPRv`{RRx`FR
v ^{vqP TFl F qvPR Tv{ NFR vT PdTTR{Rq B89
qvmdqFl dR PdWR{Rq /0 xRqR{Fdvq NvtqRNRP Flv F
qvPR (d^ $^^R{vq^R{dmxFNvTB89dR`Fq
`R F{dFdvq vT B89 vx xvR{ =`F d LRNFR B89
dRdmmRPdFRldmxFNvqB89{FqdRqFqPL {FqdRq
dmxRPFqNR`F`d^`ldqYRqNRvq`v{ Nd{NdN{{RqFqP
R{mdqFlvlF^RQvx`Rq`v{ Nf{NdvNN{
B3K
J !=B13EDJ $ *85J +>35CE19>FJ $ -9>4J'579=5DJ
! FqFlRPdq:F{ " <RNdvq ... dqPxRRPd F {FqPvm

F{dFLlRFqPdq`Rx{vLlRmvTvN`FdNx{RPdNdvqvTvlF^R
F^ RNFqPRXqR`RTFlvNN~RqNRFF{FqPvmRRqFqP
TFl {FR d Flv mvPRlRP dq^ dqP xRRP x{vLFLdld
Pd{dLdvq &dTTR{Rq dqP {R^dmR NFq LR mvPRlRP F `R
BRdLll xPT d` PdTTR{Rq dqPdNR /7 FqP J =`R F{R dq }
NvqR{RPv`R{RlFdvqLRRRqTFl{FRFqPB89vx
xvR{(d^ =`R{RTv{RFTFlRRqdPRXqRPL
(FlxRFqvPR
B89 vx xvR{ .q `d FqFld B89 d xxvRP
vNvtsRNvqvPR
(Fl {FR Nv{{RxvqPdq^ v v FLvR TFNv{ FqP
NFlNlFRPLTv{mlF$
(v{ dqPddPFl TFl RRq `v{ Nf{Nd NFlNlFdvq d
xR{Tv{mRP FqP vlF^R F^ mF^qdPR d zFqdXRP @vlF^R
F^ ZRzRqN RzFl `R {FR vT `R TFl RRq " {RxRFdq^
`d NFlNlFdvq Tv{ FllTFl xR FqP Fll NFR vT fsP xRRP
PRxRqPRP B89 vx xvR{ FqP dqR^{Fdq^ vlF^R F^
ZRzRqNRvLFdq`R`vlRZRzRqNxRN{mvTvlF^R
F^d`PdTTR{RqN`F{FNR{ddNvTF^mF^qdPR
(d^ PRxdN `R ZRzRqN xRN{m vT vlF^R F^ Tv{
`RNFRPvTB89Fdll{FRPdq(d^ BdqP{R^dmRd
mvPRlRP L ;FlRd^` xPT d` 0)$ =`d NFRP d RRP
Tv{vxRvTB89FdqPNdvqmFN`dqRFqPqNcvqv
^RqR{Fv{ d` `R FmR B89 qvmdqFl dR j@! =`R
{Rlxvdqv`FdqPNdvqmFN`dqRxRB79{Rldq
LRR{vlF^RF^yR{Tv{mFqNRmv{R`FllvF^
0 2/21.*
0/ 1.*
67
[
DL
7
Z

?
8
@
i 6K
+-
J
qQ

: -FN
8nj%Wtq|_jdz_j_|stcd%!
ftxcnffdxds|ZH:qtb_|ntsz
8nj&Wtq|_jdz_j_|stcd%!ftx
cnffdxds|zz|drztxbdb_v_bn|
(d^ x{RRq `R dmxFN vTB89 lvNFdvq B89dRP

j@! d FxxldRP Tv{ PdTTR{Rq lvNFdvq 4vPR d
qRF{R `R Rm v{NR FqP dqP xRRP `F `R {vq^R{
dmxFN vq vlF^R F^ xR{Tv{mFqNR 6`R{ B89 lvNFdvq
vql NFR `FxR N`Fq^R dq vlF^R F^ mF^qdPR PR v
dN`dq^vqFqPvTT d`v{d`vB89 BfsPxRRP`F
Flmv qv dmxFN vq vlF^R F^ LRNFR B89 lvNFdvq d
qRF{ TFl lvNFdvq =`F d ` qvPR d RlRNRP Tv{
PRmvq{FR `R dmxFN vT dqP xRRP vq vlF^R F^
xR{Tv{mFqNR dq qR FqFld dq xF{ " vT `d <RNdvq (d^
PRN{dLR dmxFN vT `R `v{ Nd{Nd NFxFNd vT `R Rm
v{NR vqvlF^RF^ ! j@! {FRPB89dNvqqRNRP
v qvPR =`R {Rl dqPdNFR `F `R RFjR{ `R Rm
v{NR d `R mv{R B89 fmxFN vq vlF^R F^ `R `FxR{
N`Fq^RfsvlF^RF^`RqB89dN`RvqFqPvT
u} q}k}a .tl
8nj8xdwdsbzvdb|xrtftq|_jdz_j_|stcd%!
n|l|ldnscxdjnrdrtcdqdcaO_qdnjlvcf
4
(d^ FqP (d^ `v `R dqR^{Fdvq vT vlF^R F^
xR{Tv{mFqNR PR v Fll TFl xR FqP B89 vx xvR{
FlRNv{{RxvqPdq^ v `R`vlR{Fq^R vTB89vxR{Fdq^
dqPxRRPTv{`RBRdLllxPTd`PdTTR{RqdqPdNR /7FqP J
.q Lv` (d^{R `R {Fq^R vT mF^qdPR lvR{ `Fq x
FNNvq Tv{ vlF^RF^PRv`RTFl vNN{`RqdqP
{LdqRdNvLRNFRvTRd`R{R{lvv{R{`d^`dqP
xRRPvdPR`RN dqvN vdqPxRRP{Fq^R@vlF^R
F^ZvmxvxF{RPRvFllTFlxRRNRxdq^
dq^lRx`FRv^{vqPTFl @vlF^RF^`d^`R{`Fqx
F{R `R {Rl vT dq^lR x`FR v ^{vqP TFl =`R ZRzRqN
vT vlF^R F^ dq `d {Fq^R d zdR `d^` LRNFR `R {FR vT
dq^lR x`FRv^{vqP TFl FNNvq Tv{vTvFlTFl{FR
.q (d^ `R BRdLll xPT d NvqdPR{RP Tv{ k ;FlRd^`
111
eBook for You
v ^{vqP .q `d NFR P F B89 d` qvmdqFl dR vT

j@! d NvqqRNRP v qvPR =`R Rm v{NR d
xxvRPvLR{vq^dmxRPFqNRRzFl x .dRFv
{RFldR `F Rd`R{ F R lv dqP xRRP lvR{ `Fq N fs
xRRP FqP F R{ `d^` dqP xRRP `d^`R{ `Fq N v
xRRPB89dqvNvqqRNRPv`RqRv{jFqPvlF^RF^
F{R `FxRl PRRx ,vRR{ vqNR B89 d NvqqRNRP v `R
qRv{j RRq d x{vPNR lv xvR{ vlF^R F^ F{R
d^qdXNHql dmx{vRP "RdPR dq^lR x`FR v ^{vqP TFl
h NFR `Fllv F^ L v`R{ TFl xRNFR PRRx F^
=`R N`Fq^R dq B89 vx xvR{ ^RqR{Fll `F lR dmxFN
vq `R vlF^R F^ mF^qdPR `Fq dN`fs^ vq FqP vTT `R
B89NvqqRNdvqv`RqRv{j
J
xPV FqP % J J J .q (d^ F R NFq XqP `F

dqN{RFdq^ % `Rv{RdNFll mRFq eqN{RFdq^ `R vlF^R F^
FR{F^RdqPxRRPv{dqv`R{v{P`RmFdmmTFl{FR
mvRvF{P`R{Fq^RvTB&9`d^`R{vxxvR{ !F
{Rl `R ZRzRqN vT `Fllv F^ LRNvmR `d^`R{ .q
(d^ BRdLllxPTd{R^F{PRPd`% HqP /7N`Fq^Rdq`R
{Fq^R vT J =`R {dR vT /7 mRFq `R
mFdmm vT1A ^R `d^`R{ FqP `R Pd{dLdvq LRNvmR
mv{R NvmxFN <v d` `R dqN{RFR dq /7 `d^` ZRzRqN vT
vlF^RF^dx{vqRv`R{Fq^RvTPRRxF^
% + +
#!&)+
x{RRqRP mR`vP NFq LR F x{RldmdqF{ zFqdXNFdvq vT

dmxFN vT fsP qNR{Fdq vq vlF^R F^ xR{Tv{mFqNR 0q
Pd{dLdvqRmd`B89NvqqRNRP
+-BFGZC41:D4FT
. d FNjqvlRP^RP `F `d v{j {RNRdRP RN`qdNFl
xxv{ Zvm :{vT 8!#F{R <Rqdv{ 3RmLR{ !J &{
<2<{dFFFFqP FqP {RRF{N`R{ d` `R #RqR{ Tv{
!PFqNRP :vR{ <Rm .qdR Tv{ 'qR{^ <Rm
'NvqvmdNFqP<FdqFLdld(lv{dPF<FR?qdR{d?<!
'
'
(

#

(
8nj<<Wtq|_jdz_jfxdwdsb_|stcd%!n|lZ1:ns|ldnscxdjnrdz
rtcdqdcaO_qdnjlvcfn|lcngdxds|_qdztf|ldzb_qdnscd
%+ +
$"'*+

"
"
&

o
%+ +
(+ +
+
8njWtq|_jdz_jidwdsb_|stcd%!n|lZ1:ns|ldnscxdjnrdz
rtcdqdcaZdnaqqvcf.(# n|lcngdxds|_qdztf |ldzl_vdnscd
X/GF-CUR=GF
=`dxFxR{x{RRqFqRFRmRqvTvlF^RF^PRv
TFl dq Pd{dLdvq Rm d` B89NvqqRNRP ;FqPvm
F{dFdvq vT dqP xRRP mFjR N`Fq^R dq B89 vx
xvR{ dqhRNRP v `R qRv{j `Rq TFl vNN{ FqP `R
vlF^R F^ xR{Tv{mFqNR d N`Fq^RP FNNv{Pdq^l =`R xFxR{
LdlP `R vN`FdN mvPRldq^ vT `R qNR{Fdq vT dqP
{R^dmR dq PRXqdq^ TFl RRq FqP TFl Pd{dLdvq
mvPRldq^ vT `R x{vLlRm vT vN`FdN x{RPdNdvq vT vlF^R
F^ =`R{RTv{R `R vlF^R F^ xR{Tv{mFqNR d FqFlRP dq
{RlFdvqd`dqP{R^dmRF&9lvNFdvq&dTTR{RqNRqF{dv
vT B89 FxxldNFdvq F{R NvqdPR{RP =`R {Rl ^^R `F
N`Fq^RdqvlF^RF^N`F{FNR{ddNPRvdN`dq^vqFqP
vTT B89 F{R `FxR{ `Fq PR v N`Fq^R dq dqP xRRP
PRxRqPRP vx xvR{ vT B89 BdqP {R^dmR d` `d^`R{
FR{F^R dqP xRRP dll ^dR mv{R dmx{vRmRq Tv{ vlF^R
F^ =`R{RRF{N`{RlF{RLhRNRPv`R{RlFdRlldmdRP
NFxFLdld Tv{ B&9 mvPRldq^ dq :<<!8':= FqP `R
]<^ D;@ ,tqqds(+-.("( 4 )2+4 0$"/4 +)$'-4 )$. 4 -4

(4(.++0*.")(-4 =444Kxdzz
]^ Z;Bdxz|nsj"-.+"0.")(4 3-.'4 )$"( 4 (4 ($3-"-4 -O-Kxdzz
CC-
]^ R_|l_on|l D_|ld "(4 (+!4 0('(.$-4 -)0+4 ($3-"-4
(4)()'"-4 RvxnsjdxWdxq_j,dxqns;dncdqadxj"
] ^ CZ_sj-Rnsjl* Bzn_p"(4 )2+4 3-.'-4 **$".")(-4 )4
)'*0..")($4(.$&" (4 RvxnsjdxWdxq_j,dxqns;dncdqadxj

]!^ Z;Bdxz|nsjO_cn_q1nz|xna|ntsTdz|8ddcdxz"-.+"0.")(4 3-.'4
($3-"-40)''"..4*)+.4 K4RRrrdxDdd|nsj
]"^ A_s * D_x|nsdWdq_zbtA_bns|t D_x~ns*ydct4DTK Dtcdq ftx
*s_qznz tf 1nz|xna|dc :dsdx_|nts =rv_b| ts Wtq|_jd R_jz
+)"( -4 =s|dy_|nts_q -tsfdxdsbd ts Ktdx Rz|drz Tx_sznds|z
=KRT# nsCts8x_sbdtsAsd ##
]#^ A* Rnq_;,8srnq_tBC ,|qdxKxx=rv_b| tf 1nz|xna|dc
:dsdx_|nts ts |ld =444 Ftcdz Tdz| 8ddcdx n|l Gdxbxxds|
Kxt|db|nts +)"( 4 &m Ftx~l *rdxnb_s Ktdx Rrvtznr
F*KR# Rdv|#b|#
]%^ ,_blNtbBl_sl1tsjAsZtsRdsj=>Dtts8_q|1nz|xna|nts
DtcdqnsjUznsjR|tbl_z|nb,n_xn_|dDtcdqz8txKxdcnb|ntstfWtq|_jd
R_j ns 1nz|xna|nts Rz|drz 4 +(-.")(-4 )(4 )2+4 $"1,4
vv #! WtFtA_s_x%
]&^ T Tlnxnsjdx*Kd|dxzzts_scTKd|x:xnccnz|xa_sbdxdzvtszdtf
nsc |xansdz dwnvvdc n|l nscb|nts jdsdx_|tx _sc ctaqfdc
nscb|nts jdsdx_|tx+)"( -4 =444 K4R :dsdx_q Ddd|nsj
tqvv! ! #
]^ \D*|_484>R__c_sORdd|l_v_|lD D*R_r_q_4gdb|tf
Znsca_zdc 1: Rd_zts_qn| _sc Usbdx~_ns| ts 1nz|xna|nts Rz|dr
Ctzzdz+)"( 4 }m Ftx~l *rdxnb_s Ktdx RrvtznrF*KR
% %Rdv|%
]<<^ O B_xpnK ;O ,nqqns|ts* Rnrvqnhdc Znsc Ktdx:dsdx_|nts
Dtcdq ftx Odqn_anqn| 4_q_|nts 4 +(-.")(-4 )(4 (+!4
)(1+-")(4 WtqFt v v!! Asd"

]<^ K :ntxzd|~tB 8 U|zxtjn1ddqtvrds| tf * Fd Kxtbdcxd 8tx
Odqn_anqn|DtcdqnsjtfZnscTxansd:dsdx_|txz4 +(-.")(-4
)(4)2+4**+.0-4(43-.'-4 WtqK*R<Ft<A_s&%
]^ -CD_z|dxzAD|_qd:R|xa_bR-xbnb _sc F@dspnszR|_|nz|nb_q
d_q_|nts tf tq|_jdz ns cnz|xna|nts zz|drz n|l dradccdc nsc
jdsdx_|nts%4+)"( -4 :dsdx_|ntsTx_szrnzznts_sc1nz|xna|nts
vv'#Wtq #Ft Aq
]< ^ * F -dqnp4sdxj t|v| dz|nr_|nts ftx zr_qqzb_qd nsc vtdx
jdsdx_|txz znsj Zdnaqqxdvxdzds|_|nd nsc c_|_)0+($4 )4 "(4
( "(+"( 4(4(0-.+"$4+)('"-4 Vb "&##
]<!^ T*Rltx~4qdb|xnbKtdx1nz|xna|nts;_scattp-O-Kxdzz
xdbdndc ,RDR _sc
Kl1 cdjxddzns 4qdb|xnbKtdx Rz|drzfxtr
;_stn Usndxzn| tf Rbndsbd _sc Tdblstqtj
;URT Wnd|s_r ns && &&# _sc
xdzvdb|ndq ;dl_zadds_f_bq|rdradxtf
4qdb|xnb Ktdx Rz|drz 1dv| Rblttq tf
4qdb|xnb_q4sjnsddxnsj;URTznsbd;dnz
bxxds|q _ nzn|nsj zbltq_x ns |ld -ds|dx ftx
*c_sbdcKtdxRz|drz=4R4R8qtxnc_R|_|d
Usndxzn|UR* ;nzxdzd_xblns|dxdz|znsbqcd
vtdx zz|dr _s_qznz vtdx cnz|xna|nts
vq_ssnsj 1RD vtdx zz|dr w_qn| _sc
cnz|xna|dcjdsdx_|nts
\| p j|
112
eBook for You
P584O4F-4R
V. V. Thang
D. Q. Thong and B. Q. Khanh
Department of electric power systems

Thainguyen University of Technology (TNUT)
Thainguyen, Vietnam
thangvvhtd@tnut.edu.vn

Hanoi University of Science and Technology (HUST)
Hanoi, Vietnam
bq_khanh-htd@mail.hut.edu.vn
investment [3], increase the complexity in measurement and
relay protection and operation of DSs. By using mathematical
programming or heuristic programming, many authors have
already proposed planning models of DSs that have objective
function involving one or more objectives. A single-objective
planning model can aim at minimizing total active power loss
[4, 5] of DS including DG. Material [6] introduced two
objective models consisting of minimum loss of total active
power and allowed voltage drop limitation. Another planning
model having three objectives comprises the first that is the
construction and operation costs of DG as well as energy
expense from market, the second that is the cost of power loss
and the third, which is the cost of environmental pollution.
These three objectives are used to estimate the location,
capacity of DG in CEM [7]. Similarly, a different planning
model including total power cost, loss and minimum voltage
drop is given in [8].
Abstract - Recently, the restructuring of electricity market and

the development of technology have been enhancing the
application of distributed generators (DGs) and renewable
energy resources. Hence, the planning of distribution systems
(PDS) for competitive electricity markets (CEMs) has
encountered impacts which need to be investigated. This paper
proposes a novel approach for optimazing distribution system
planning in CEMs with the presence of DGs. The proposed model
can determine equipment sizing and timeframe required for
upgrading power network in order for utilities to purchase
electric energy from electricity markets. Besides, the problems of
DG system development (Optimal DG displacement, sizing,
technology selection and installation period) in distribution
planning to meet the demand growth will be solved. The model
uses the objective function that minimizes the total cost of
network (feeders and transformers) upgrading, new DGs
installation, distribution systems operating, and electric energy
purchasing from CEMs. The proposed model is tested using an
33 bus 22kV radial feeder. The calculation is programmed in
GAMS environment.
A significant change in DSs planning is the reconstruction

of electricity markets with source, price constraints in recent
years. Therefore, using DG is a new planning approach. Most
of researches mainly focus on one-objective models which
combine of construction and operation costs in order to
determine capacity, location and a new building investment
process or to upgrade current equipments by using popular
mathematical programming. A new model is proposed to plan
DSs in long-term in [9] when considering DG source planning
schemes. In this research, the objectives are the minimum
summary of investment and operation costs of DG, the
investing cost of feeder and substation transformers during
planning period, and goodness indices (incremental loss indices
and incremental feeder loading indices) are also represented.
The DG technology is not mentioned due to the assumption
that the costing functions and effects of DG in DSs planning
are the same, but they are impossible in reality. Another model,
which owns objectives consisting of the total investing and
operating costs of DG, feeders and substation transformers
upgrading costs, energy expenses and minimum interruptible
load costs is shown in [10]. In this model, effects of DG
technology are not mentioned in selecting variables.
Keyword:
Distributed Generator (DG), Planning of
Distribution System (DS), Competitive Electricity Markets (CEM)
I. INTRODUCTION
Recently, the issue of planning of DS has gained
remarkable achievements. Because of the construction of
competitive electricity market (CEM), technological
development and environmental pollutions, the development of
DG and renewable energy resources, in particular, is fostered
[1].
DGs connected directly to DSs or supplied straight to
customers [2], normally use new electric generating
technologies such as gas turbines, Combined Heat and Power
(CHP), Fuel Cells, solar energy, geothermal and wind energies.
The power of DGs can reach to 300MW depending on
particular technologies, but the power in use of DGs is
normally less than 5MW. The DG is installed close to loads so
that it gets some main advantages including the elimination of
transmission and distribution cost, enhancement of flexibility
and reliability of distribution systems, reduction of power loss,
and improvement of differential voltage at nodes as well as
reduction of environmental pollution because of using
renewable energy resources [2]. However, DGs require high
B mn H thng in - i hc Bch
Khoa IEEE
H Ni
978-1-4577-0365-2/11/$26.00
2011
Environmental pollution is one of the burning issues

worldwide nowadays. Traditional energy resources generate
immense impacts to environment, whereas high-tech DG and
631
113
eBook for You
A New Model Applied to the Planning

of Distribution Systems for Competitive Electricity
Markets
selecting technology of DG through economic and technical

indices. All costs during are calculated at the same time that is
the first time of planning period by using discount rate r. (1)
gives the objective function J of the proposed planning
problem in new conditions:
clean, renewable energy resources create a very small

pollution. Hence, in the planning of DSs comprising polluted
air index, [11] shows a two-stage planning model. The
minimum of total costs of upgrading feeders, substation
transformers and DGs construction, energy expenses purchased
from market and environmental pollution costs is applied as the
objective in this research. The heuristic programming method
illustrated in [12] with the objective is the minimization of total
costs of construction, upgrading and operation fees, feeder,
substation transformers and DG, energy expenses of DSs. This
method does not use binary variables so that the computation
burden is reduced significantly.
J =
t
NS
+ ( C SF + C SC .SiS,t ) +
i =1
NS
Effects of CEM, energy policies, environmental pollution

problem and development of technology emerge new
constraints that lead to using DG in DSs planning as a new
solution. This research proposes a new approach to DS
planning that considers DGs as an optimal selection. The
objective function is minimum amount of total costs including
investing and upgrading costs of feeder and substation
transformers, building cost of new DGs, and operating fees, as
well as fuel of DG and energy expenses purchased from market
via connected substation transformers. All these costs are
converted to the first year of planning period. In order to meet
the technical criterions, models constraints consist of nodal
power balance between supplies and load demands, DG power
limitation, efficiency of existing equipments, required nodal
voltage drop. The calculation tool to solve this proposed
planning problem is the GAMS program language.
+ Tmax .
NDG KDG
CkDG .SiDG
,k ,t
i =1 k =1
(1)
( PS .Pi S,t
QS .QiS,t
i =1
NDG KDG
+ Tmax .
i =1 k =1
( kPDG .Pi DG
,k ,t
kQDG .QiDG
,k ,t
) Min
1
calculated total cost at base year;
( 1 + r )t
Components in 1 are upgrading cost of feeders for year t with
fixed capital cost (CFF) and variable capital cost (CFC).
substation transformers upgrading costs in year t with fixed
capital cost (CSF) and variable capital cost (CSC) in 2 . 3 are
new investment costs of DGs at node i, year t with DG
technologies k. Electrical energy purchased cost from CEM in
4 and 5 are O&M and fuel costs of DG depending per
technology.
TABLE I. SETS, INDICES, VARIABLES AND PARAMETERS
No
Symbol
Definition
I. Sets and Indices
1
N
Set of buses in distribution system
2
i, j
Bus (i, j N)
3
NL
Set of load buses in distribution system
4
NS
Set of substation buses in distribution system
5
NDG
Set of DG buses in distribution system
6
T
Overall planning period, year
7
t
Planning period (t T)
8
KDG
Total number of DG's technology
9
k
Technology of DG (k KDG)
II. Variables
Active power purchased from CEM at node i, for year t
10
PSi,t
(MW)
Reactive power purchased from CEM at node i, for year t
11
QSi,t
(MW)
F
12
S i,j,t
Upgrading capacity of Feeder ij for year t (MVA)
13
SSi,t
Upgrading capacity for Substation i, at year t (MVA)
New investment capacity of DG node i for technology k,
14
SDGi,k,t
at year t (MVA)
Active power of DG node i, for technology k, at year t
15
PDGi,k,t
(MW)
Reactive power of DG node i, for technology k, at year t
DG
16
Q i,k,t
(MVAr)
17
Ui,t
Voltage for node i, at year t (pu)
Voltage angle at bus i, for year t (radian)
18
i,t
III. Parameters
19
r
discount rate (%)
20
CFF
Fixed capital cost of Feeder ($/km)
21
CFC
Variable capital cost of Feeder ($/MVA.km)
22
Li,j
Length of Feeder ij (km)
23
Yi,j
Magnitude of admittance matrix element (1/)
II.
PROPOSED A NOVEL DISTRIBUTION SYSTEM PLANNING

MODEL
In CEM, distribution systems are managed by distribution
companies (Discos). In order to meet load demands in future,
these companies can buy electrical energy from CEM via
power system connected substation transformers or to
coordinate to invest DGs. As DGs are chosen in DSs planning,
economic and technical indices of planning project are changed
which affects considerably to time, upgrading capacity and
improvements of feeders and substation transformers.
A. Objective Function
A medium-term planning of DS is proposed in this article.
The objective is to maximize profits of Discos or to minimize
total investing and operating costs. Therefore, the objective
function is minimum quantity of total costs of constructing and
upgrading investments for existing feeder and substation
transformers, new construction of DGs, operating and fuel
costs of DGs and energy expenses are bought from electricity
market during planning horizon. The proposed model is an
one-stage nonlinear programming model with decided
variables are continuous. For the purpose of calculation
reduction, these results are rounded to suitable values which
appropriate to existing equipments. This model also allows
632
114
eBook for You
Where:
The next parts of this paper are organized as follows.

Section II introduces a model of the proposed DS planning
problem with objective function and constraints. Section III
represents calculation results from the 33nodes, 22kV
distribution system. Conclusion is reported in Section IV.
1
N N
1
FF
FC
F
.
(
C
+
C
.
S
)
L
i , j ,t
i, j
( 1 + r )t i =1 j =1
TABLE I. Continued
Symbol
i,j
CSF
CSC
CDGk
SP
SQ
30
kPDG
31
kQDG
32
33
PDi,t
QDi,t
34
PDGmax,k
35
QDGmax,k
36
37
Umax
Umin
38
39
40
41
fSL
U min U i ,t U max
U i ,t = cons tan
DG
DG
DG
Pi DG
,k ,t = Pi ,k ,t 1 + P ,Qi ,k ,t = Qi ,k ,t 1 + Q
(6)
(7)
t 1,i NDG , k KDG
4) Constraint of ST capacity limits and Dynamic capacity

updutes
With the assumption that substation transformers are
ensuring electric supply for load demands of current DS to
make use of existing substation transformers capacities and to
satisfy annually upgrading power corresponding to equipment
parameters. This constraint is given as follows:
(2)
SiS,t f SL .SiS,1
j =1
i , j N
Constraint of nodal power balance in (2) is only used for
nodal loads, when DSs planning considers the use of DG with
different time and technologies, the previous formula is
rewritten as (3 and 4).
t 1, i NS
(8)
SiS,t SiS,t 1 + S t 1, i NS
(9)
The planning model from formula (1) to (9) is a nonlinear

programming model. The proposed investigation uses NLP or
MINLP solver in GAMS program language [13] to find out an
optimal solution.
III. RESULTS AND DISCUSSIONS
Distribution systems normally use popular radial diagram
that has advantages such as low-cost, easy operation, lower
power and electricity energy losses. However, the reliability is
lower since all loads would be interrupted if failure were
occurred at source side. Therefore, it should be used for loads
which require not so high electric quality and reliability. When
DG is connected to the DS, these disadvantages can be
S
,k ,t + Pi ,t PDi ,t =
PiDG
Yij .Ui,t .U j,t .cos(ij j,t i,t )
(3)
PiDG
,k ,t 0,i N,t T ,k KDG
(5)
i NS
DG
DG
DG
Pi DG
,k ,t Pmax,k , Qi ,k ,t Qmax,k i NDG , k KDG
i, j N,t T ,k KDG
i NL
3) Constraint of DG capacity limits and Dynamic capacity

updates
This constraint allows computed DG capacity at nodes in
limit of DG technology, and it ensures annually upgrading
power corresponding to equipment parameters (6 and 7).
j =1
j =1
(4)
voltage loss. Therefore, in order to meet technical requirements,

allowed voltage drop must be in range with full loads. Voltages
at substation nodes are assumed constants, constraint of limited
nodal voltage then is give as (5).
k =1
j =1
2) Constraint of limited nodal voltage

It is much important to voltage quality in DS using less
regulated voltage devices mainly provides to electric devices
directly. In such DS, voltage quality cannot be good due to large
PDi + Yij .U i .U j . cos( ij j i ) = 0
k =1
QiDG
,k ,t 0,i N ,t T ,k KDG
1) Contraint nodal power balance

In estimation of distribution grids, nodal power should be
balanced to make sure the balance of capacity of the whole
system. Nodal power balance in a grid for nodal loads [9] is
given as follows (2).
KDG
i, j N,t T , k KDG
B. The DS planning constraints

Optimal planning of DS with objective function satisfying
economic and technical requirements will be guaranteed when
all constraints including nodal power balance between supplies
and load demands, maximum DG power limitation, nodal
voltage drop, required and limited capacity of power system
connected substation transformers are matched.
QDi Yij .U i .U j . sin( ij j i ) = 0
KDG
S
,k ,t + Qi ,t QDi ,t = Yij .Ui ,t .U j ,t . sin(ij j ,t i ,t )
QiDG
Definition
Angles of admittance matrix elements (radian)
Fixed capital cost of Substation ($)
Variable capital cost of Substation ($/MVA)
New investment cost for DG technology k ($/MW)
Active power purchased cost from CEM ($/MWh)
Reactive power purchased cost from CEM ($/MVArh)
O&M cost and Fuel cost of DG for active energy
($/MWh)
O&M cost and Fuel cost of DG for reactive energy
($/MVArh)
Active power demand at bus i, for year t (MW)
Reactive power demand at bus i, for year t (MVAr)
Maximum DG capacity limit for active power with
technology k (MW)
Maximum DG capacity limit for reactive power with
technology k (MVAr)
Maximum voltage limit at bus (pu)
Minimum voltage limit at bus (pu)
Active power ramp-up limit for DG in planning year
(MW)
Reactive power ramp-up limit for DG in planning year
(MVAr)
Capacity ramp-up limit for Substation transformer in
planning year (MVA)
Load factor of Substation transformer base year
633
115
eBook for You
No
24
25
26
27
28
29
TABLE II. TOTAL COST OF FEEDERS AND TRANSFORMERS UPGRADING [3]
overcome. Hence, this structure is popular despite of increased

complexity of relay protection. The IEEE 33bus radial
structure is investigated in this research. Parameters are
changed to match the problem.
No
1
2
A. Diagram and Parameters of 33bus radial distribution

system
Figure 1 illustrates the IEEE 33bus radial structure that has
22kV with substation transformer node connected to 110kV
grid and 32 load nodes. The total active power and reactive
power at the base year are 10,675.0kW and 9,040.0kVAR,
respectively. Load data is in APPENDIX A. Total length of
feeders is 41.7km with detailed parameters in APPENDIX A.
22
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
DG technology is not mentioned in detail. The effects of

DG technology, however, are represented by investment,
operation and fuel costs. Two DG technologies namely,
solar photovoltaic (PV) and small gas turbine sources, are
used in this research with the corresponding capital costs to
be 5.0M$/MW and 0.5M$/MW. Average O&M and fuel
costs depend on used technology and they are shown in
Table III.
23
24
25
26
27
28
29
30
31
32
TABLE III. CAPITAL, O&M AND FUEL COST OF DGS [3]

No
1
2
DG technology
Gas turbine
Solar PV
Capital cost
(M$/MW)
0.5
5
O&M cost
($/MWh)
10
1
Fuel cost
($/MWh)
80
0
Constraint of limited load nodal voltage is changed from

0.9pu to 1.1pu, and it should be 1.05pu at substation node.
33
Since advanced DG technology is mature, integrated-DG

compact modules occupy small spaces and install in short
time. Moreover, installing areas at load locations are no
limit. Only one DG per technology is chosen at each load
location. However, it can be selected more than one DG
technology simultaneously at each load location. Each
planning year, additional power of DG should be 0.1MW.
Figure 1. Diagram of 33bus IEEE radial distribution system
B. Analysis and Discussions

To estimate and test the feasibility of the proposed model,
the above radial diagram is used to investigate with parameters
and assumptions as follows.
Upgrading areas of substation transformers and feeders are

not limited so that only existing equipments can be
upgraded. Each planning year, additional capacity of the
110kV transformer should be 16MVA corresponding to
devices in market. Upgrading feeder capacity is suitable to
cross-section of selected feeders.
1) Assumptions in analyis
This research utilizes some economic and technical
assumptions for the ease of computation:
Planning period is 5 years. Annual developing rate of load
demand is constant at 5% during planning time. Hence, the
total increased load demand is 25%.
Decided variables (feeder and ST upgrades, DG

investment) in the proposed model are continuous in order
to reduce the complexity of the model (not need to use
binary variable). Hence, they should be rounded to match
real equipments
The investing and operating costs are converted to the first

year of planning with discount rate is 10%, annually.
Current substation is capable of supplying to load demand
at base year. Hence, to make use of existing equipments,
minimum load fact of substation in this model is 0.9.
2) Analysis resuls of cases

The feasibility of the proposed model and efficiency of DG
are investigated by two cases in the 33bus radial diagram. Case
A, in which DG is not considered, decides an upgrading time
for feeders and substations. Case B is similar to Case A but DG
is mentioned in the researching model.
The constructing cost of 110kV substation includes fixed

costs (land clearance, design, tax) and variable costs
(equipment expenses) are 0.2M$ and 0.05M$/MVA,
respectively [3]. Similarly, the upgrading costs of 22kV
feeders consist of 0.15M$/km for fixed costs and
0.001M$/MVA.km for variable costs. These parameters are
shown in Table II.
Variable capital cost

50.000 $/MVA
1000 $/MVA
a) Case A
This case aims to estimate the working capacity of current
feeders and substations as load demands in future.
Furthermore, upgrading time and capacity of them will be
decided to ensure requirements within planning period.
634
116
eBook for You
19
20
21
Fixed capital cost

200.000 $
150.000 $/km
Energy expenses from CEM via substations are 100$/MWh

and 60$/MVARh corresponding to active and reactive
energy [3].
Substation
01
Resource
Substation
Feeder
power of DG can be selected by 2.0MW, PV is only capable of

providing active power while Gas turbine can supply both
active and reactive powers with 0.8 power factor. Parameters
of DGs in Table VII are used in computation of GAMS
environment.
Decision of an optimal planning model in term of 5-year

planning is illustrated in Table IV and V.
Substation Transformer
upgrading capacity in each
year (MVA)
1
2
3
4
5
16
-
Capacity
in
base year
(MVA)
Total capacity of
Substation
Transformer
(MVA)
16
32
TABLE VII. DATA OF DGS [3]

No
1
2
1
2
3
4
5
6
7
Capacity in
base year
(MVA)
No
Feeder
TABLE V. FEEDERS UPGRADING DECIDED - CASE A
1-2
2-3
3-4
4-5
5-6
8-9
9-10
16.96
16.96
12.57
12.57
12.57
4.95
4.95
Feeder section upgrading Feeder capacity upgrading

in each year (mm2)
in each year (MVA)
1
- 240 - 185
150 - 150 - 120 35
-
* Where: Cca - Capital cost of DG; Cpo, Cqo - O&M and Fuel cost of DG
Table VIII presents decisions optimal investment of

proposed planning model for DG. During planning time, it is
advisable to invest new DGs with 5.6MW in total equal to
52.46% of base years load demands. Investment of DG
focuses mainly on early years of planning period and chooses
both assumption technologies. PV is chosen for 18th and 33rd
nodes at the third year with 0.4MW and 0.2MW, respectively.
Gas turbine is decided immediately at the first year for 18th,
22nd and 33rd nodes with 1.2MW, 2.0MW and 0.6MW,
respectively. In the third year, 1.2MW is added at 18th node.
- 23.24 - 19.43 - 16.9 - 16.9 - 14.48 - 6.67 35

- 6.67
The above results show that substation transformer need to

upgrade a 16MVA capacity at the first year in order to meet the
5% increased demand. Similarly, feeders are also need to
upgrade depending planning time. Feeders which are close to
the substation transformer (from this substation to 6th node and
from 8th to 10th node) must be upgraded at the first years of
planning period (23.24MVA corresponding to 240mm2 is the
greatest upgrading capacity at the feeder from 1 to 2 node and
6.67MVA corresponding to 35mm2 is the smallest upgrading
capacity at 9-10 feeder). Upgrading time is lasted within the
planning period. In the first year, the 3-4 feeder must be added
16.9MVA and the capacity of 9-10 feeder then must be added
in the final year by 6.67MVA.
TABLE VIII. DG INVESTMENT DECIDED

DG
Bus
technology
Solar PV
Gas turbine
Total
1
2
3
Cost
Substation Transformer
upgrading
Feeder upgrading
O&M and Electrical
energy
Total
Investment, O&M and Energy

purchased from CEM
in each year (M$)
1
1.00
2
-
3
-
4
-
5
-
0.25
0.51
0.48
0.5
0.24
8.57
9.06
9.55
10.06 10.58
Total
cost
(M$)
1.00
1.2
2.0
0.6
3.8
0.0
0.4
0.2
1.2
1.8
0.0 0.0
Capacity in
Total capacity
base year
of DG (MVA)
(MVA)
0
0
0
0
0
0
0.4
0.2
2.4
2.0
0.6
5.6
In general, DGs owning high investments offer a cheap

energy price, PV source in particular. Hence, early investment
will achieve high effectiveness since cheap energy price is
utilized during final years of planning period. Therefore, all
DGs are invested early and they are mainly used at the first
year with 3.8MW.
1.98
47.82
50.80
Electricity energy purchased from CEM via substation

transformer and capacities of feeders have both been decreased
so that the upgrade of feeders and substation transformer can
be delayed as shown in Table IX. Consequently, compared to
case A, investing costs of feeders and substation transformer
have been reduced by 1.98M$ in this case.
In this case, costs for feeders and substation transformer

upgrading which are converted to the base year, are 2.98M$
and energy costs correspond to 47.82M$ shown in Table VI.
As a result, the total costs of case A are 50.80M$.
b) Case B
DG technology is utilized in this case. Solar PV energy (PV
type) and Gas turbine with opposed economic indices are used
in this investigation and they are shown in Table III. Maximum
18
33
18
22
33
DG capacity invested in
each year (MW)
The selected location of DG is far from the substation

transformer, namely, 18th and 33rd nodes so that high economic
and technical efficiencies are gained. A decreased transforming
capacity from the source to remote loads leads to a reduction of
power loss, electricity energy loss and operating cost. It also
improves voltage profile and reliability.
TABLE VI. TOTAL INVESTMENT, O&M AND ENERGY PURCHASED FROM

CEM COST OF CASE A
No
Cca
Cpo
Cqo
DG
Pmin Pmax
Qmin
Qmax
(M$/
($/
($/
technology (MW) (MW) (MVAr) (MVAr)
MW) MWh) MVArh)
Solar PV
0
2
0
0
5
1
0
Gas turbine
0
2
0
1.5
0.5
90
10
635
117
eBook for You
No
Substation
Transformer
TABLE IV. SUSTATION TRANSFORMER UPGRADING DECIDED - CASE A
TABLE IX. FEEDER UPGRADING DECIDED - CASE B
1
2
Loi thit b
Feeder, i-j
Substation
Transformer, 1
Compared to case A, the construction cost is raised by

2.52M$ since DGs investing cost is 5.5M$. However, O&M
and fuel expenses of DGs are low so the total costs of case B
are just only 46.24M$ as detailed in Table X.
b) Comparison of technical index

Total amount of electric energy bought from CEM in case
B is continuously reduced during the planning period as given
in Figure 2. In the fifth year, this amount has been dropped
most significantly with 8,780.0MWh. This should be important
to environmental pollution impact because of reduction of
traditional power energies. In this research, the planning time is
just only 5 years while life-span of electric devices normally
reaches to 20 years so that the efficiency of DGs will be much
higher.
TABLE X. TOTAL INVESTMENT, O&M AND ENERGY PURCHASED FROM CEM

COST OF CASE B
Cost
Total
cost
(M$)
2
3
Substation
Transformer
upgrading
Feeder upgrading
Investment DG
1.9
0.0
3.6
0.0
0.0
0
5.5
O&M and Electrical energy
7.66 8.10 8.12 8.34 8.52
Total
80.00
70.00
Electrical Energy (10e6kWh)
TT
Investment, O&M and

Energy purchased from CEM
in each year (M$)
40.74
46.24
3) Comparisions of economic and technical indices

The optimal decision of the proposed model, when the two
cases are tested, shows the optimal upgrading process (time
and capacity) of selected substation transformer and feeders.
Moreover, optimal location, technology and process of DGs
investment can be determined from this research. In case A,
when load demands are raised in future, feeders and substation
transformer will be overloaded. It hence must be necessary to
upgrade them to guarantee constraints of capacity limit and
voltage profile. However, in case B, DGs will support the
capacity of feeders and substation transformer. As a result, the
upgrade is not needed. This method offers excellent economic
and technical indices. Below comparisons will justify the
efficiency of DG in planning of distribution systems.
60.00
50.00
40.00
30.00
20.00
10.00
0.00
1
-10.00
-20.00
Case A
2
Case B
5
Planning Period (years)
Comparison Case B and Case B
Figure 2. Comparison of electrical energy purchases from electricity

markets between Case B and Case A
Active power and electricity energy losses are also one of

important indices that can be used to assess the efficiency of
the planning project [4]. Figure 3 presents active power losses
in two cases and the comparison between them. When DGs
investment is selected, active power loss is dropped during all
the time of planning. In the first year, this loss has been
decreased considerably by 1.11MW and it keeps dropping in
following years. 1.95MW is a significant drop in the last year.
This leads to a reduction of electric energy costs by
12,940.0MWh corresponding to 1.29M$.
a) Comparision of economic index

Economic indices between case B and case A are compared
in Table XI.
TABLE XI. COMPARISON OF I NVESTMENT, O&M AND ELECTRICAL ENERGY
COST BETWEEN CASE B AND A
3
2
Comparison cost
between Case B
Case B
and Case A
1
2
3
4
Cost
Substation
Transformer upgrading
Feeder upgrading
O&M and Electrical
energy
Investment DG
Total
Case A
1.00
0.00
-1
1.98
0.00
-1.98
47.82
40.74
-7.08
0.00
50.8
5.5
46.24
5.5
-4.56
Note
Total cots is
reduced -8.99%
TT
Active Power Loss (MW)
Total cost (M$)
1
0
1
-1
-2
-3
Case A
Case B
Comparison Case A and Case B Planning Period (years)
Figure 3. Comparison of active power loss between Case B and Case A
As can be seen from the Table XI, Case B holds a better

economic index. Costs of DGs investment and equipment
636
118
eBook for You
No
upgrades (feeders and substation) are more expensive those of

case A by 2.52M$ due to a very high cost of DGs investment.
For instance, capital cost of PV should be 5.0M$/MW.
Nevertheless, O&M and electric energy expenses have been
decreased by 7.08M$ because of very low O&M and fuel
expenses of DGs. For example, PV has zero cost of fuel.
Therefore, the efficiency gets higher as time reaches to final
years of planning period. What is more, total costs of case B
are cheaper than these of case A by 4.56M$, equal to 8.99%.
Capacity upgrading in each year (MVA)

1
2
3
4
5
-
Voltage comparison between two cases in the final year

points out supported voltages at all nodes, when DGs are
installed as give in Figure 7.
Voltages at load nodes in cases A and B are represented in

Figure 4 and 5, respectively. It can be seen that voltage profiles
in case A at remote nodes are low at the first year. Voltages of
17th and 18th nodes are less than 0.9pu. In following years,
voltages at all nodes are dropped significantly and most of load
nodes get voltages that are less than 0.9pu at final year. Voltage
of 18th node is smallest at 0.79pu.
20
15
Voltage Drop (%)
10
1.1
1.05
Voltage (p.u)
5
0
-5
11
13
15
17
19
21
23
25
27
29
31
33
-10
-15
0.95
-20
0.9
-25
Case A
Case B
Comparison Case B and Case A
Bus
0.85
Figure 7. Comparison voltage drops between Case B and Case A
0.8
It is assumed that substation transformer node voltage is

constant (1.05pu) and the length of feeders is short. Hence,
voltage drops at nodes, which are close to substation
transformer, are small. In case B, voltage drops of far load
nodes have been reduced sharply. At 33rd node, voltage drop
goes down moderately from 17% to 10% but this drop has a
significant reduction of 16% at 18th node.
0.75
0.7
1
5
7
Year 1
11 13
Year 2
15 17
Year 3
19 21
Year 4
23
25 27
Year 5
29
31
33
Bus
Figure 4. Bus voltage profiles in Case A

1.1
1.05
0.9
0.85
0.8
0.75
0.7
1
Year 1
11
13
Year 2
15
17
19
Year 3
21
23
Year 4
25
27
Year 5
29
31
33
Bus
Figure 5. Bus voltage profiles in Case B
Figure 5 shows that voltages at load nodes are enhanced

remarkably in case B. During planning time, voltage profiles at
all load nodes are greater than 0.9pu. This is because DGs
decrease voltage drops. In the fifth year, nodal voltages receive
the greatest support.
1
0.9
Voltage (p.u)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2
Case A
Case B
3
Comparison Case A and Case B
5
Planning Period (years)
Figure 6. Comparison voltage profile of bus 18 between Case B and

Case A
Load node having the biggest support is 18th node in which

optimal investing decision of the model selects the highest
power by 2.8MW. Figure 6 illustrates voltages profile of the
18th nodal in both cases within planning time. In case B,
voltage profile is improved in the first year immediately and is
still maintained at excellent values in following years. The
voltage improvement of case B compared to case A is raised
gradually from the first year to the fifth year and it gets the
maximum at 0.16pu.
637
IV. CONCLUSION
Planning of distribution systems have been change
significantly for recent years because of reconstruction process
of CEM, technological development and environmental
pollutions, promoted the development of DGs and renewable
energy resources, in particular. The new proposed model for
planning of distribution systems has allowed selecting DGs as
an optimum. The objective function is minimum quantity of
total costs of investing and upgrading costs of feeders and
substation transformers, building cost of new DGs, operating
fees as well as fuel cost of DGs and electricity energy expenses
purchase from CEM via connected substation transformers. All
these costs are converted to the first year of planning period.
Nodal power balance between supplies and load demands, DG
power limitation, efficiency of existing equipments, required
nodal voltage drop are constraints of this research model so
that they can guarantee technical requirements. An optimal
process can be estimated (time and power) to upgrade and
reconstruct feeders and substation transformers, and choose
optimal locations, power and investing process of DGs.
Particularly, DG technology can be selected in this model via
economic and technical indices. It can be seen from the results
that planning together with using DG usually provides better
economic and technical outcomes. Total investing and
operating costs of planning of distribution system, converted to
the base year, have been reduced, active power and electricity
energy losses have been decreased, voltage profiles have been
supported and upgrading time of existing feeders and
substation transformer has been delayed. Furthermore,
reduction of electricity energy purchased CEM has limited the
use of traditional energy resources, which contributes to the
decrease of environmental pollution.
119
eBook for You
Voltage (p.u)
1
0.95
APPENDIX A. DATA OF LOAD FOR 33 BUS RADIAL D ISTRIBUTION SYSTEM

Bus
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
QD0
(kVAr)
160
240
250
130
310
300
300
120
160
110
135
435
180
460
200
320
No
Bus
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Total
PD0
(kW)
490
190
290
190
390
390
420
220
660
560
360
420
300
550
310
460
10,675
27
28
29
30
31
32
QD0
(kVAr)
440
140
220
140
340
350
350
200
525
525
310
370
200
470
260
390
9,040
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
1.2
2.3
3.4
4.5
5.6
6.7
7.8
8.9
9.10
10.11
11.12
12.13
13.14
14.15
15.16
16.17
17.18
2.19
19.20
20.21
21.22
3.23
23.24
24.25
6.26
26.27
Fij
(mm2)
AC-150
AC-150
AC-95
AC-95
AC-95
AC-35
AC-35
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
AC-35
AC-35
Smax.ij
(MVA)
16.96
16.96
12.57
12.57
12.57
6.67
6.67
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
4.95
6.67
6.67
Lij
(km)
1.5
0.6
1.6
1.5
2.1
2.5
1.1
1.6
1.5
0.2
0.4
1.2
1.4
1.3
1.6
1.2
0.8
0.2
1.3
0.3
0.7
1.4
0.8
0.8
1.5
2.2
Rfij
()
0.2910
0.1164
0.5024
0.4710
0.6594
1.9325
0.8503
1.8336
1.7190
0.2292
0.4584
1.3752
1.6044
1.4898
1.8336
1.3752
0.9168
0.2292
1.4898
0.3438
0.8022
1.6044
0.9168
0.9168
1.1595
1.7006
4.95
4.95
4.95
4.95
4.95
4.95
1.7
1.8
1.3
2
2.2
1.4
1.9482
2.0628
1.4898
2.2920
2.5212
1.6044
0.7310
0.7740
0.5590
0.8600
0.9460
0.6020
REFERENCES
[1]
[2]
[3]
[4]
[5]
APPENDIX B. D ATA OF FEEDER PARAMETERS FOR 33 BUS RADIAL

DISTRIBUTION SYSTEM
Bus i - Bus j
AC-25
AC-25
AC-25
AC-25
AC-25
AC-25
* Where: Smax - Maximum capacity limit for Feeder
* Where: PD0, QD0 - active and reactive power demand at bus in base year of planning period
No
27.28
28.29
29.30
30.31
31.32
32.33
[6]
Xfij
()
0.5760
0.2304
0.6352
0.5955
0.8337
1.0725
0.4719
0.6880
0.6450
0.0860
0.1720
0.5160
0.6020
0.5590
0.6880
0.5160
0.3440
0.0860
0.5590
0.1290
0.3010
0.6020
0.3440
0.3440
0.6435
0.9438
[7]
[8]
[9]
[10]
[11]
[12]
[13]
638
P. Frase and S. Morita, Distributed Generation in liberalised electricity

markets, International Energy Agency, IEA Publications 2002. p.20.
Thomas Ackermann, Goran Andersson, Lennart Soder, Distributed
generation: a definition, Electric Power Systems Research 57, 2001.
S. Wong, K. Bhattacharya and J.D.Fuller, Comprehensive framework
for long-term distribution system planning, Proc. IEEE PES Annual
General Meeting, Tampa, USA, 2007.
N. Mithulananthan, Than Oo and Le Van Phu, Distributed Generator
Placemenitn Power Distribution System Using Genetic Algorithm to
Reduce Losses, Thammasat International Journal of Science and
Technology, 2004.
Amin Hajizadeh, Ehsan Hajizadeh, PSO-Based Planning of
Distribution Systems with Distributed Generations, International
Journal of Electrical, Computer, and Systems Engineering 2008.
Vinoth Kumar, K. and Selvan, M.P., Planning and Operation of
Distributed Generations in Distribution Systems for Improved Voltage
Profile, IEEE PES Power Systems Conference and Exposition (PSCE
2009), March 15-18, 2009, Washington, USA
A.Soroudi and M.Ehsan, Multi Objective Distributed Generation
Planning in Liberalized Electricity Markets, Transmission and
Distribution Conference and Exposition, 2008. IEEE/PES
Muhammad Ahmadi, Ashkan Yousefi, Alireza Soroudi, Mehdi Ehsan,
Multi Objective Distributed Generation Planning Using NSGA-II,
13th International Power Electronics and Motion Control Conference
(EPE-PEMC 2008).
Algarni, A.A.S.; Bhattacharya, K., A Novel Approach to Disco
Planning in Electricity Markets: Mathematical Model, Power Systems
Conference and Exposition, 2009. PSCE '09. IEEE/PES
El-Khattam, W.; Hegazy, Y.; Salama, M., An integrated distributed
generation optimization model for distribution system planning, Power
Engineering Society General Meeting, 2005. IEEE
S. Wong, K. Bhattacharya1and J.D. Fuller, Electric power distribution
system design and planning in a deregulated environment, IET
Generation, Transmission & Distribution, 2009.
El-Khattam, W.; Bhattacharya, K.; Hegazy, Y.; Salama, M.M.A.,
Optimal investment planning for distributed generation in a competitive
electricity market, Power Systems, IEEE Transactions on
Volume: 19, 2004.
GAMS 20.5, A Users Guide, GAMS Development Corporation, 1998.
120
eBook for You
No
PD0
(kW)
200
290
320
160
360
300
300
160
220
145
160
460
220
560
260
360
Analysis of DSMs Impacts on Electric Energy

Loss in Distribution System Using VPI Model
AbstractThis paper presents a novel proposal for

quantifying the influence of Demand Side Management (DSM)
objectives on electric energy loss in distribution system. DSM
objectives are modeled by VPI model and the testing results
prove its suitability, especially for analysis on energy loss in
distribution system. This paper uses IEEEs 34 Node Test
Feeder as the test system for energy loss calculation and the
analysis of DSMs influences. Network modeling is made using
PSS/ADEPT 5.0. The research uses typical daily load curves of
industrial and residential loads in Vietnam as a case study.
Results not only indicate the effectiveness of loss reduction in
DSM peak-clipping and load shifting, but also recommend
DSM strategies for the best effect on energy loss reduction
basing on modeling error testing.
Index TermsDSM, CLC, LDC, ILDC, peak clipping, load
shifting, power loss, energy loss, distribution system.
I. INTRODUCTION
EMAND side management has been defined as the

planning, implementation and monitoring of a set of
electric utility programs and actions designed to
influence energy demand in the ways that will produce
desired changes in utilitys electric load curve. Utilitys
DSM programs use a variety of different means including
load management, new uses, strategic conservation,
electrification customer generation, and adjustments in
market shares. It is generally the most convenient for
utilities to look at DSM in terms of broad load shaping
objectives. Six broad categories of load shaping objectives
can be distinguished as shown in Fig. 1 [1, 5].
Strategic
Conservation
Peak
Clipping
Valley
Filling
Strategic
Load Growth
DEMAND-SIDE
MANAGEMENT
Load
Shifting
Flexible
Load Shape
Fig. 1. DSMs six categories of load shaping objectives

This work is financially supported by the Vietnam Education
Foundation for Visiting Scholar Program for the year 2010-2011.
Bach Quoc Khanh is with Electric Power System Department,
Electrical Engineering Faculty, Hanoi University of Science and
978-1-4577-1002-5/11/$26.00
2011
IEEE
B
mn H thng in - i hc Bch Khoa
H Ni
With these load shaping objectives, it can be concluded that

DSM programs have the main objective of making the load
curve as flat as possible. These changes in utilitys load
shapes as the result of DSM programs implementation bring
to utilities with many different benefits from planning,
designing to operation of power networks. Because DSM
solutions are primarily seen as a potential generation, so-far
researches of DSMs impacts have been mainly related to
generation planning, not much research has considered
DSMs impacts on network operation, especially for the
distribution system. One of possible benefits in power
system operation is the DSMs influence on energy loss in
power systems and, this paper presents a new proposal for
analytical assessment of DSMs influence on technical loss
of energy in distribution system operation with regard to
load shaping objectives using VPI model [6].
This paper is arranged in the following manner. In
Section II, the test distribution network modeling and energy
loss calculation is covered. Section III analyses the methods
for modeling DSMs impacts and proposes the model for
this work. Section IV defines the problem and details
calculation results and analysis. Finally, conclusions are
given in Section V.
II. CALCULATION OF TECHNICAL LOSS OF ENERGY IN

DISTRIBUTION SYSTEM
This DSMs influential research is based on the
dependence of energy loss of power system on the demands.
So, we start it with considering the problem of energy loss
calculation.
A. The problem of calculating technical losses in
distribution system
The energy loss in distribution system is always the
utilitys concern in operation. Energy losses are commonly
classified into technical and non-technical losses. Technical
losses are those corresponding to Joule effect (I2R), corona
effect at high voltages and transformer core losses. In
distribution system, Joule effect on system components due
to the transport of electrical power energy and transformer
core magnetization account for almost energy losses. The
amount of energy loss mainly depends on the characteristics
of the distribution system, types of equipment in use and the
supplied demand. In order to calculate technical losses in
distribution system, it is required the topological data of
networks and the metered data of demand at all load buses.
From the data of demand, it is possible to obtain the
customers load curve for a period of time. Daily and annual
load curves are among the most commonly used types.
121
eBook for You
Bach Quoc Khanh
B. Network modeling
For analyzing DSMs impacts on energy losses in
distribution system, this research chooses the IEEE 34
Nodes Test Feeder as the test distribution system. Among
radial distribution test feeders proposed in IEEE Distribution
System Analysis Subcommittee Report [3], IEEE 34 Nodes
Test Feeder is characterized by long feeder with the highest
rate of power loss (13,37%) meanwhile it also displays a
wide variety of components and topological features
(multiple single-phase and three phase laterals, transformers,
voltage regulators, capacitor banks). In this work, the test
system is modeled by Siemens Power System Simulator
Advanced Distribution Engineering Productivity Tool
(PSS/ADEPT 5.0) as plotted in Fig. 2. This is an effective
tool for distribution system modeling. PSS/ADEPT 5.0 has
several useful functions for analyzing power system and
reporting results. In load flow analysis, reports on branch
power loss and total system power loss are included. That
helps fasten analyzing process concerning energy loss as
used for this research. Some remarks for this network
modeling are as follows
eBook for You
Power losses on components (distribution transformers,

lines) in distribution system at each instants of the load
curve will be obtained by executing three-phase unbalanced
load flow calculation. Once the component loss curves are
assembled, component energy losses and total network
energy loss can be easily obtained. Although the above
mentioned methodology is straight forward, it is always
challenged by how to adequately collect data of load curves
at all load buses throughout the distribution system. There
have been several researches proposing different methods
for estimating energy loss in the case of inadequate
monitored load data [15, 16, 17], but in general, the most
accurate energy loss can be obtained only if the load curves
of all load nodes in the interested system are available. With
only regard to the problem of assessing the DSMs
influence, this research assumes that load curves at all load
nodes are available and they have the same shape as the
typical load curve of the interested distribution system.
Different typical load curves for residential and industrial
related loads referred from the distribution system in
Vietnam as a case study (Part IV) are considered.
Fig. 2. Modeling 34 Nodes IEEE Test Feeder on PSS/ADEPT 5.0
- PSS/ADEPT only accepts sequence values for threephase line modeling, so the inherent unequal mutual
coupling in the phase impedance matrices for distribution
lines [2] is not considered. Normally, its only accepted for
the transposed lines. Since distribution lines are rarely if
ever transposed, mutual coupling between phases is not
equal, so it may cause errors. However, because coupling
inductances are much smaller than self inductances, the
error should be negligible.
- Distributed load lines are modeled using lumping load
modeling in which one-third of the load should be placed at
the end of the line branch and two-thirds of load placed at
one forth of the way from the source end [2]. Therefore, a
number of fictional load nodes labeling A are additionally
generated in the test system. For example, the line branch
802-806 is load distributed, and it is modeled by two
lumping loads. One-third of the load is allocated to node 806
and other two-thirds of load is placed at an additional load
node namely 806A at one forth of branch length from source
end (Node 802).
- The source transformer labeled XMF-1 is added in the
test system for power loss calculation. In order to have the
voltage level at node 800 equal 1.05 p.u. as specified by

IEEE results, the source voltage is set to 1.0 p.u. and the
XMF-1s Tap Setting is fixed at 1.075.
- Two voltage regulators are also modeled as the
transformers with voltage ratio equal unit. However,
compensation settings are complied with IEEEs data.
The results of PSS/ADEPT modeling generally agree with
IEEEs results [4]. Regarding the power loss calculation, the
comparison in Table I also proves the result agreement.
TABLE I
COMPARISON BETWEEN IEEE AND PSS/ADEPT RESULTS
OF POWER LOSS CALCULATION
Results
Total P (kW)
Total P loss (kW)
IEEE
2042.872
273.049
PSS/ADEPT
2047.240 (*)
270.819 (*)
Error (%)
0,21
-0.81
( )
* After excluding power loss on the source transformer XMF-1.
III. DSMS IMPACTS MODELING

A. Introduction
In order to analyze DSMs impacts in problems of power
systems, first of all, it is needed to model the impacts of
DSM activities on the system load curve. With regard to
122
P(MW)
P(MW)
800
800
600
600
400
400
200
200
6
12
18
24 t(h)
a) Load Curve
12
18
24 t(h)
b) Load Duration Curve (LDC)
t(h)
Pr
24
18
0.75
12
0.5
P(MW)
200
400
600
c) Inverted LDC
800
0.25
P(MW)
200
400
600
800
d) Normalized ILDC
Fig. 3. Probabilistic simulation of LDC
LDC is transformed into Inverted Load Duration Curve

(ILDC) where time and demand axes are swapped.
Furthermore, the time axe is normalized to denote the
probability P(X > Pload) that the demand exceeds a certain
level of load Pl. With normalized ILDC, many mathematic
models have been proposed.
For analyses in power generation problems, a very
important application of ILDC has been introduced as the
Equivalent Load Duration Curve (ELDC). This curve gives
the probabilistic description of the customer demand and
failure of generating units, and it is well known as the
Probabilistic Production Costing model (PPC) which is very
useful for production cost simulation and reliability analyses
for generation planning.
Pr
1
ILDC
0.75
ELDC
0.5
0.25
800
600
200
400
Fig. 4. ILDC and ELDC
P(MW)
ELDC also used in power system expansion planning [13]

as composite power system equivalent load duration curve
(CMELDC). Recently, other works [8,9,10] have
recommended further extended application of ELDC with
equivalent load frequency curves (ELFC) for analyses of
load-levelling benefits of pumped store units and DSM
planning. However, these models are mainly applied for
generation planning or DSMs impacts for generation parts,
because it depends on the production cost of different types
of generation sources
With regard to customer load curves, there are also a
couple of related researches. [12] introduces ILDC model as
a synthetic LDC with six fitted parameters (a, b, c, d, f, g)
and maximum demand, each parameter defines a part of
ILDC. The model is generated by probabilistically analyzing
samples of randomly chosen 332 customers by HydroQuebec and the resulting models are applicable for different
typical loads, each type corresponds to a set of those six
parameters. However, this model is not applicable for DSM
impact analysis because of its dependence on six parameters
that are only applicable for limited contexts. Further, only
peak demand is considered, therefore, it can not describe
entire DSM load shaping objectives.
Some researches as reviewed in [6] represents ILDC in
the polynomial form, where the coefficients of the
polynomial are closely related to two quantities - the ratio of
the minimum to the maximum load during the period, and
the ratio of the average load to the maximum load.
However, in order to represent the ILDC including the DSM
impacts, they have to go back to the numerical data to show
the changes in system load, regenerate the LDC and
calculate the ILDC again. This is a slow process and is
subject to errors.
Another work models ILDC using stochastic distribution
known as the asymmetric generalized Gaussian (AGG)
distribution [11]. This is an approximate model where,
parameters of AGG (right and left deviations and shape
parameter c) are calculated from limited data of ILDC.
However, this model is also inapplicable for DSM modeling
as AGG is generated basing on the entire data information
of load curve. Therefore, a change in load shape due to
DSM impact is difficult to simultaneously model for
different load levels.
B. VPI model for DSM impacts modeling
DSM induces changes in system load and consequently
generates new load shapes. It means that the expression for
new load shape has to be derived each time there is any
change in DSM parameters (peak load, base load and
supplied energy). Above reviewed models of LDC can not
totally consider changes in these DSM parameters in the
mathematical expression for ILDC. For the problem of
energy loss calculation, the accuracy depends significantly
123
eBook for You
DSM load shaping objects, DSM activities can be directly

characterized by peak load, base load and supplied energy.
So, once the expected impacts of demand side activities on
the system load are estimated, ways have to be found to
represent these in the load curve model. For example, under
DSM peak-clipping object, if 5% of the peak is clipped, how
the shape of load curve changes accordingly? Customer
demand is normally manifested by monitored data as the
chronological load curve (CLC). This format of load curve
is very difficult to model because of uncertain nature of
loading. However, in reality, various techno-economic
analyses of the power system do not necessarily use
complete load information in CLC format. Another more
predictable demonstration of load curve is well known as the
load duration curve (LDC). Such curves express the
relationship between time and demand, showing the percent
of time the demand is greater or equal to a certain level. The
LDC is obtained by a rearranging of the chronological loads
into descending order over a period of time. Because the
shape of LDC is much easier to model by mathematic tools,
it has been a basic tool used for the analysis of electric
power systems such as the operating cost of resource plans,
the planning of electricity generation.
There have been several mathematical models that try to
describe directly the shape of LDC. Some models [14]
transform the LDC in linear shape for different segments of
LDC. However, it may cause unacceptable error, especially
for highly unleveled load curves. It is noticeable that LDC is
also seen as the demand frequency distribution graph and
thus another trend to model LCD is based on probabilistic
simulations with the common transition procedure described
as the example of daily load curve in Fig. 3 [6,7].
on the data of LDC. To reduce the error, characteristics of

DSM activities should be built directly in LCD curve model.
And the model proposed by Rahman and Rinaldy [7] which
is also known as VPI model can handle these changes
directly and it is really a effective model for analyzing DSM
impact. VPI model manifests ILDC as a function of DSM
parameters as follows
X Pmin
X Pmin
(1)
exp C .
f ( X ) 1

P
P
P
P
max
min
min
max
same load pattern as the typical load curves. In other words,

system load diversity factor equals unit. We will discuss the
cases with diversity factor greater than unit basing on the
results with unit diversity factor later. This work proposes
two typical load curves (per unit) for residential and
industrial customers in Vietnam as case studies (Fig. 5 and
Appendix A) for analyzing the influence of origin load
shape on DSMs effects.
for Pmin < X < Pmax

for X < Pmin
= 1
= 0
for X > Pmax
where
Pmax: Peak load, Pmin : Base load, X: Load
C is a constant depending on Pmax, Pmin and E
n
E T . Pmin
0.5
max Pmin )
n 1
(2)
where
E: Supplied energy, T: Time period of LDC,
An: Coefficients of the inverse function of the following
function
m
ai . x i
i 1
(i 2)!.x
(3)
i 1
with m = 7 we have
A1 a11
A2 a13 a 2
A3 a15 (2a22 a1a3 )
A4
(4a)
(4b)
(4c)
a17 (5a1a2 a3 a12 a4 5a23 )

9
1
(4d)
A5
a (6a a 2 a 4 3a a 14a a a 21a a a ) (4e)
A6
a111 (7a13 a 2 a5 7a13 a3 a 4 84a1a 23 a3 a14 a6
2
1
2 2
1 3
4
2
3
1 5
2
1 2 3
28a12 a 2 a32 42a52 28a12 a 22 a 4 )
A7
13
1
(4f)
(8a a2 a6 8a a a 4a a 120a a a
4
1
4
1 3 5
4
1
2
4
2
1
3
2 4
180a12 a22 a32 132a26 a15 a7 36a13 a22 a5

72a13 a2 a3 a4 12a13 a33 330a1a24 a3 )
(4g)
1
for i=(17) (5)
(1 i )!
Comparative tests on using VPI model and similar models
shown that this model give more exact results in calculating
power system indices using ILDC and ELDC. This paper
proposes the application of VPI model for representing
DSM impacts in system load curve for analytical assessment
of DSM impacts of energy loss in distribution system.
and ai
IV. PROBLEM DEFINITION

A. Energy loss calculation and VPI model testing
For applying VPI model in this DSM problem, the paper
initially tests its performance in energy loss calculation for
the distribution system without DSM application. As
presented in Section II, energy loss calculation for
distribution system is based on typical load curves for the
test system. The calculation can be started for the critical
case which assumes all loads in the test system have the
Fig. 5. Typical daily load curves of distribution system

in Vietnam for 2009 [18]
In Fig. 5, dashed curve is typical for residential loads in

rural areas and solid curve is typical for loads in industrial
zones in the North of Vietnam. Daily load curve data is
given in every half an hour demands. For loss calculation
and analysis, the typical load curve is initially transformed
into LDC as Fig. 6. It is easy to find that industrial LDC is
more leveled.
As above assumed,
the load curve at all
load nodes of the test
system
will
be
calculated as follows
Pi (t ) Pi .P * (t ) (6)
Qi (t ) Qi .P * (t )
where
Pi, Qi: Load values at
load node i given by
IEEE test system
P*(t): The per-unit
peak demand at the
instant t in LDC (Fig. 6)
eBook for You
A T .( P
Fig. 6. LDC and VPI model LDC of

typical residential and industrial load
in Vietnam
At each instant t (30 minutes), with above (6) calculated

load nodes, test system load flow is performed using
PSS/ADEPT. The resulting branch power losses and total
system power loss 'P(t) are also reported by PSS/ADEPT.
By repeating this system power loss calculation for other
half-an-hour demands until all demands in the entire period
are taken into account,
we obtain system power
loss
duration
curve
which has the same
shape as LDC. Finally,
daily system energy loss
is determined as follows
'E
48
'P (t ).t
(7)
t 0
In calculating system
power loss, we also build
up the discrete relation
between system power
Fig. 7. Relation between power loss

and demand of IEEE 34 Nodes Test
System with unit diversity factor
124
m 1
24.Pmin t ( X i ).G P
EB E X
(10)
i 0
Pmax Pmin
and X i
m
m is an integer.
where G P
B i.G P (i=0,m)
t (h)
24
a) ILDC
EB
EX
P*
Pmax = 1
Pmin
t (h)
24
b) IPLDC
EB
EX
P(kW)
Pmin
Pmax
Fig. 8. VPI modeled ILDC (a) and IPLDC (b) built by

combining ILDC and P(P*)
The bigger the value of m is selected or the smaller the value

of G is, the more accurate the system energy loss is obtained.
In this work, GP=0.005 which means m=119 (for residential
LDC) or m=103 (for industrial LDC). Furthermore, by
combining ILDC with the relation P(P*), we also build up
the discrete Inverted Duration Power Loss Curve - IDPLC
(Fig. 8). From IDPLC, the system energy loss can be also
calculated as two areas as follows
(11)
'E 'E B 'E X
'EB
'E X
24.'Pmin
(12)
m 1
t ( X )>'P ( X
i
i 1
) 'P ( X i ) @
(13)
i 0
The system energy losses are summarized in Table II. The

results show that the differences (errors) of supplied energy
(E in per-unit-peak.hour) are very small between original
data and VPI modeled data. The differences of energy loss
amounts are also insignificant.
TABLE II
VPI MODELED LDC TESTING FOR ENERGY LOSS CALCULATION
Residential
Industrial
Load
E
E
'E (kWh)
'E (kWh)
Original
14.548
3203.087
17.780
4249.698
VPI modeled
14.628
3166.703
17.840
4194.661
Error (%)
-0.547
1.136
-0.337
1.295
Therefore, VPI modeled ILDC is suitable for analysis on

energy loss in power system with regard to DSMs impacts.
B. Modeling DSM objectives and analysis of DSM effects
on energy loss
Basing on VPI model, DSM load shaping objectives are
modeled. Because VPI model is directly built with
characterized outcomes of DSM activities (Pmax, Pmin and E),
DSMs impacts are easily modeled. However, in VPI model,
the constant C is approximately calculated basing on limited
7th series-reversion polynomial that ensures the equivalence
of supplied energy (energy amount must be almost the same
between the original LDC and the VPI modeled LDC) [6].
This approximation only results in acceptable error for
supplied energy E for the cases where DSM parameters
Pmax, Pmin and E vary within certain limits. Therefore, during
modeling DSMs impacts on the load curves, an effective
parameter for testing the modeling error is the error of
supplied energy GE.
- Load shifting: In this mode, DSM objective is to move
the demand from on-peak period to off-peak period.
Therefore, when the peak demand reduces, the base demand
likely increases and supplied energy is kept unchanged.
Depending on the LDC of each typical load, this paper
considers different levels of peak demand decrease. At each
peak decrease levels, different levels of base load increase
are considered. So, a two dimensions model of load shift is
proposed. For each case of Pmax, Pmin, and E, basing on the
combination of resulting VPI modeled ILDC and P(P*),
the corresponding IPLDC is obtained in similar manner as
we do for testing VPI model in Part A. From IPLDC, system
energy loss is calculated as formulas (11) to (13). Besides
that, from ILDC, error in supplied energy (GE=E-EVPI) is
tested in the same way as shown in Table II and the
conclusion should finally be made for this case of modeling.
The selection of the levels of peak load decrease and
levels of base load increase is based on the shape of LDC.
For the residential load curve, according to [14]
P Pmin means the shape of LDC is very
E
max
2
Pmin .T
sloping. Therefore, the decrease rate of peak load is selected
about twice time greater than increase rate of peak load. We
consider peak decrease levels of 0%, 5%,,30% whereas
the base increase levels are 0%, 2.5%,,20%. For the
E
P Pmin so we use the same
industrial load,
! max
Pmin .T
2
rate of variation (0%, 5%,,25%) for both peak clipping
and base filling. Furthermore, the error GE is very high for
the cases where deep peak clipping (20-25%) or high base
filling (25%), the chart is plotted for maximum of equal or
less than 3% error. The impacts of DSM load shifting on test
system energy loss are given in 3D chart in Fig.9 for
residential load and Fig. 10 for industrial load.
Fig. 9 shows that for cases with low error GE (deep peak-
125
eBook for You
loss and system load, P(P*), for the IEEE test system with
unit diversity factor as Fig. 7 and Appendix B. This curve
will be a useful tool for analysis of DSMs impact on
distribution system energy loss in Part B of this section.
In order to test VPI model for loss calculation, it needs to
build VPI modeled ILDC. From data of CLC, VPI modeled
ILDC for industrial and residential loads are given in the
equations (8) and (9).
- For residential loads
X 0.405
X 0.405
t 1
.24 (8)
exp 1.29253.
1 0.405
1 0.405
- For industrial loads
X 0.485
X 0.485
t 1
.24 (9)
exp 0.01954.
1
0
.
485

1 0.485
By converting them back to the form of LDC, VPI modeled

LDC for typical loads are also depicted on Fig. 6. On the
other hand, ILDC (8) and (9) have the shape as Fig. 8.
Therefore, the system supplied energy is calculated by two
areas EB and EX
load curve shape becomes, the lower the system energy loss
gets. However this effect of energy loss reduction depends
on the shape of LDC without DSM. For the industrial load,
because the original LDC is quite flat, the effect of loss
reduction is lower in comparison with residential load. In
Fig. 10, big errors appear either in the area where deeply
clipped peak goes with small base increase and the opposite
area where slightly clipped peak is paired to high raised base
load. If we also consider cases with GE < 0.5%, the best
possible strategy of DSM load shift is summarized in the
Table IV.
TABLE IV
OPTIMAL STRATEGY OF DSM LOAD SHIFT FOR THE BEST ENERGY LOSS
REDUCTION FOR TYPICAL INDUSTRIAL LOAD
Fig. 9. System energy loss variation and supplied energy error

of VPI model ILDC in load shift modeling for residential load
Fig. 10. System energy loss variation and supplied energy error
of VPI model ILDC in load shift modeling for industrial load
0
5
4151
1.035
0.338
5
5
4121
1.749
0.337
10
10
4064
3.119
0.338
15
15
4022
4.114
0.338
20
20
3996
4.732
0.338
25
25
3988
4.936
0.363
- Peak clipping: In this mode, DSM objective is to cut down

the peak demand. For this purpose, when the peak demand
reduces, the supplied energy can decrease, but the base load
remains unchanged. The modeling process for this DSM
objective is performed in similar way as DSM load shifting,
we can obtain the effects on system energy loss variation. It
is noticeable that for checking the error of supplied energy
GE, the E calculated from VPI modeled ILDC will be
compared with in-advance selected levels of saved energy E
under this DSM objective. The Fig. 12 depicts the energy
loss variation against DSM and errors of supplied energy for
residential load.
clipping and relative low base filling), system energy loss

reduce considerably. For example, if we only consider cases
with GE < 0.5%, the following DSM strategy (Table III) will
result in the greatest reduction of energy loss in which the
effect of energy loss variation is calculated as follows
G'E
'E 'EDSM
.100[%]
'E
(14)
where 'E and 'EDSM are system energy losses without and
with DSM application respectively.
TABLE III
OPTIMAL STRATEGY OF DSM LOAD SHIFT FOR THE BEST ENERGY LOSS
REDUCTION FOR TYPICAL RESIDENTIAL LOAD
Peak-clip (%)
Base-fill (%)
'E (kWh)
G'E (%)
Error GE(%)
5
0
3142
0.782
0.452
10
2.5
3091
2.41
0.439
Fig. 11. Using VPI modeled LDC for

modeling DSM load shift for the best
system loss reduction
15
7.5
3025
4.461
0.489
20
10
2993
5.497
0.471
25
12.5
2967
6.308
0.453
30
15
2948
6.892
0.436
Fig. 11 manifests the

change in VPI modeled
LDC under the impacts of
DSM load shift objective. If
we change the error limit,
among cases that meet the
new error limit, we can
choose another new DSM
strategy with the best
energy loss reduction. The
effect of energy loss
reduction by this DSM
objective is explained as the
non-linear relation P(P*).
Therefore, the flatter the
Fig. 12. System energy loss reduction and supplied energy error
of VPI model ILDC in peak-clip modeling for residential load
Because the LDC shape of the residential load is rather

sloping, normally, the decrease in peak load just results in
small amount of saved energy. So we consider the levels of
energy saved up to 20%. In the Fig. 12, the low error GE
corresponds to the area with small levels of energy saved.
This result totally meets the above assumption that DSM
peak-clipping normally results in slightly saved energy for
the case of very sloping shape of LDC. For GE < 0.5%, the
strategy that gives the best energy loss reduction is shown in
Table V and Fig. 13. For industrial load, due to its quite flat
126
eBook for You
Peak-clip (%)
Base-fill (%)
'E (kWh)
G'E (%)
Error GE(%)
TABLE V
OPTIMAL STRATEGY OF DSM PEAK-CLIPPING FOR THE BEST ENERGY LOSS
REDUCTION FOR TYPICAL RESIDENTIAL LOAD
5
0
3142
0.782
0.452
10
2.5
3012
4.895
0.44
Fig. 13. Using VPI modeled LDC

for modeling DSM peak-clip for
the best system loss reduction
15
7.5
2792
11.82
0.481
20
10
2682
15.3
0.461
25
12.5
2580
18.54
0.445
30
15
2484
21.55
0.431
shape of LDC, changes in

peak demand will result in
larger changes in saved
energy comparing to the case
of residential load. So, the
modeling
considers
the
amount of energy saved up to
25% with the corresponding
levels of peak clipped. The
affect of DSM peak-clipping
on energy loss for the
industrial load and the error
of supplied energy are
demonstrated in Fig. 14.
of VPI model ILDC in peak-clip modeling for industrial load
In Fig. 14, the area with slightly peak-clipping but large

energy saved cause very big errors GE, so the error GE is just
shown up to 5% for better demonstration. If we only
consider GE < 0.4%, the DSM strategy that gives the best
energy loss reduction is shown in Table VI.
TABLE VI
OPTIMAL STRATEGY OF DSM PEAK-CLIPPING FOR THE BEST ENERGY LOSS
REDUCTION FOR TYPICAL INDUSTRIAL LOAD
Peak-clip (%)
Energy saved (%)
'E (kWh)
G'E (%)
GE(%)
5
0
4153
1.002
0.337
10
5
3798
9.468
0.337
15
10
3470
17.27
0.338
20
15
3171
24.41
0.338
25
20
2900
30.86
0.341
The results indicate a better effect of energy loss

reduction resulted by DSM peak-clipping objective for
industrial load because under DSM impacts, saved energy
amount in peak-period is larger than for residential load at
the same level of peak clipping.
- Valley filling: In this mode, DSM objective is to increase
the base demand. Therefore, when the base demand grows,
the supplied energy amount possibly increases, but the peak
demand is fixed. For residential load, due to its low base
load, a high potential for valley filling is featured. When the
base load raises up to 30%, the energy growth can be
considered up to 45%. For industrial load, the potential of
valley filling is not much, the base raising and the energy
growth are both considered up to just 25%. The DSM effects
on energy loss for both two load types are depicted in Fig.
15 and Fig. 16. This mode features low errors corresponding
to a large area of Pmin-E pairing. With GE < 0.5%, DSM
strategies with max. and min. system energy loss increases
in percentage is summarized in Table VII. In Table VII,
energy loss increase is higher for residential load as the
result of high potential for valley filling.
of VPI model ILDC in valley-fill modeling for residential load
of VPI model ILDC in valley-fill modeling for industrial load
TABLE VII
SYSTEM ENERGY LOSS INCREASE IN DSM VALLEY FILLING OBJECTIVE
Residential
Industrial
Max 'E
Min 'E
Max 'E
Min 'E
B/E
B/E
B/E
B/E
G'E
G'E
G'E
G'E
5/30
46.7
5/10
12.99
5/10
16
5/0
-1.035
10/35 55.17 15/20 26.51 10/15 24.67
15/5
5.642
15/40 64.52 25/30 42.73 20/20
33.7
20/10
14.05
30/45 73.17 30/35 51.84
25/15
23.14
- Strategic conservation: Modeling the influence of this

DSM objective on energy loss should be similar to DSM
peak-clipping. However, because base load is also reducible
in this DSM objective, different levels of base load should
be considered.
- Strategic load growth: In this DSM objective, because
the peak load is increasable, the influence of this DSM on
energy loss should be analyzed basing on DSM valleyfilling objective for increasing levels of peak load.
- Flexible load shape: In this mode, because all three
DSM characteristics (Pmax, Pmin and E) are variable, effects
on energy loss is distribution system can be considered for
individual programs with assumed DSM characteristics.
Furthermore, for the load diversity factor less than unit, if
peak demand at all load nodes remains unchanged, system
peak demand will reduce and the base demand likely
increases. That means the system load shape is flatter and
energy loss will reduce accordingly. The impacts of DSM
on system energy loss also reduce. Therefore, the above
results should be seen as the critical case of DSM impacts
on system energy loss. If peak demand at all load nodes
changes, maybe the system peak demand also changes.
However, because the peak demand of load nodes can not
increase much for a specified system configuration due to
the initially designed limits of powering capacity of system
components, the assumption of unchanged peak demand for
a specified system is possibly accepted.
127
eBook for You
Peak-clip (%)
Energy saved (%)
'E (kWh)
G'E (%)
GE(%)
V. CONCLUSIONS
This paper presented a new proposal for analytical
assessment of DSMs impacts on technical energy loss in
distribution system using VPI model. The results show the
suitability of VPI model application for energy loss analysis
because the error in supplied energy always helps find the
set of practicable DSM strategies. The results also show the
quantitative effect of energy loss reduction for DSM peakclipping and load-shifting. That will be an additional
benefice for estimating the performance of DSM programs
in distribution system. The calculation and modeling in this
proposal will give more detailed results if the value of m in
(10) gets greater and more levels of DSM parameter
variation (peak, base, energy) are considered.
[7]
APPENDIX A
[8]
[2]
[3]
[4]
[5]
[6]
Data of Typical Load Curves [18]

h
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
8
Residential Industrial
0.435
0.53
0.43
0.52
0.415
0.505
0.405
0.5
0.405
0.5
0.405
0.5
0.41
0.5
0.42
0.485
0.46
0.495
0.51
0.51
0.555
0.55
0.585
0.6
0.575
0.67
0.55
0.74
0.52
0.83
0.495
0.92
h
8.5
9
9.5
10
10.5
11
11.5
12
12.5
13
13.5
14
14.5
15
15.5
16
h
0.5
0.98
16.5
0.54
0.995
17
0.595
1
17.5
0.65
1
18
0.655
0.945
18.5
0.625
0.85
19
0.585
0.815
19.5
0.55
0.8
20
0.52
0.84
20.5
0.485
0.9
21
0.465
0.93
21.5
0.475
0.96
22
0.49
0.975
22.5
0.52
0.985
23
0.58
0.96
23.5
0.69
0.925
24
0.8
0.88
0.88
0.825
0.95
0.8
0.98
0.795
1
0.79
0.95
0.78
0.91
0.775
0.88
0.72
0.85
0.685
0.815
0.655
0.755
0.645
0.69
0.645
0.625
0.63
0.56
0.61
0.5
0.58
0.465
0.55
[9]
[10]
[11]
[12]
[13]
APPENDIX B
Discrete System Power Loss vs Demand Relation P(P*) of IEEE 34
Nodes Test Feeder with Unit Diversity Factor modeled by PSS/ADEPT 5.0
P*
1
0.995
0.99
0.985
0.98
0.975
0.97
0.965
0.96
0.955
0.95
0.945
0.94
0.935
0.93
0.925
0.92
0.915
0.91
0.905
0.9
0.895
0.89
0.885
0.88
0.875
0.87
0.865
0.86
0.855
P(kWh)
288.969
286.09
283.23
280.389
277.567
274.764
271.979
269.217
266.471
263.746
261.039
258.352
255.683
253.034
250.405
247.794
245.203
242.631
240.079
237.544
235.031
232.536
230.06
227.604
225.167
222.749
220.351
217.972
215.612
213.272
P*
0.85
0.845
0.84
0.835
0.83
0.825
0.82
0.815
0.81
0.805
0.8
0.795
0.79
0.785
0.78
0.775
0.77
0.765
0.76
0.755
0.75
0.745
0.74
0.735
0.73
0.725
0.72
0.715
0.71
0.705
P(kWh)
210.951
208.649
206.367
204.104
201.861
199.637
197.432
195.247
193.082
190.936
188.809
186.702
184.614
182.546
180.497
178.468
176.459
174.469
172.499
170.548
168.617
166.706
164.814
162.942
161.09
159.257
157.444
155.651
153.878
152.124
P*
0.7
0.695
0.69
0.685
0.68
0.675
0.67
0.665
0.66
0.655
0.65
0.645
0.64
0.635
0.63
0.625
0.62
0.615
0.61
0.605
0.6
0.595
0.59
0.585
0.58
0.575
0.57
0.565
0.56
0.555
P(kWh)
150.39
148.676
146.982
145.308
143.653
142.019
140.404
138.809
137.234
135.68
134.145
132.63
131.136
129.661
128.206
126.771
125.357
123.963
122.588
121.235
119.901
118.587
117.294
116.021
114.768
113.536
112.324
111.133
109.961
108.81
P*
0.55
0.545
0.54
0.535
0.53
0.525
0.52
0.515
0.51
0.505
0.5
0.495
0.49
0.485
0.48
0.475
0.47
0.465
0.46
0.455
0.45
0.445
0.44
0.435
0.43
0.425
0.42
0.415
0.41
0.405
P(kWh)
107.68
106.57
105.48
104.411
103.363
102.335
101.328
100.341
99.375
98.43
97.506
96.602
95.719
94.856
94.015
93.194
92.394
91.616
90.858
90.121
89.405
88.71
88.036
87.383
86.751
86.14
85.551
84.983
84.435
83.91
ACKNOWLEDGMENT
It is acknowledged that this work received technical
supports from Prof. D.A. Cartes, Senior Member, IEEE, and
researchers with the Center for Advanced Power Systems,
Institute for Energy Systems, Economics and Sustainability,
Florida State University (FSU), USA as well as Dinh Cong
Bang (Ph.D candidate, College of Communication &
Information, FSU) for IT supports.
[14]
[15]
[16]
[17]
[18]
Clark W. Gellings, John H. Chamberlin, Demand Side Management:

Concepts and Methods, 2nd Edition, Fairmont Press, 1993.
W.H.Kersting, Distribution System Modeling and Analysis, CRC
Press LLC, 2002.
W. H. Kersting, Radial Distribution Test Feeders, Distribution
System Analysis Subcommittee Report, PES Summer Meeting, 2000.
J.A. Silva, H.B.Funmilayo, K.L. Butler-Purry, Impact of Distributed
Generation on the IEEE 34 Nodes Test Feeder with Overcurrent
Protection, Proceeding, 39th North American Power Symposium
(NAPS 2007), Sept. 30 2007-Oct. 2 2007.
Clark.W. Gellings, The Concept of Demand-Side Management For
Electric Utilities, Proceedings of The IEEE, Vol.73, No.10, Oct.
1985.
S. Rahman, Rinaldy, An Efficient Load Model for Analyzing
Demand Side Management Impacts, IEEE Transactions on Power
Systems, Vol. 8, No. 3, August 1993.
Mrs. Anjali Dharme, Dr. Ashok Ghatol, Demand Side Management
Quality Index for Assessment of DSM Programs, IEEE PSCE 2006,
pp. 1718-1721.
A. S. Malik, B.J Cory, An Application of Frequency and Duration
Approach in Generation Planning, IEEE Transactions on Power
System, Vo. 12, No. 3, August 1997.
A. S. Malik, Simulation of DSM Resources as Generating Units in
Probabilistic Production Costing Framework, IEEE Transactions on
Power Systems. Vol. 13, No. 4, November 1998.
A. S. Malik, Modeling and economic analysis of DSM programs in
generation planning, International Journal of Electrical Power &
Energy Systems, Pages 413-419, Volume 23, Issue 5, June 2001.
M.O.M Mahmoud, M. M.Jadane-Sadane, J. Souissi, N.Hizaoui,
Modeling of the Load Duration Curve Using the Asymmetric
Generalized Gaussian Distribution Case of the Tunisian Power
System, Proceeding, IEEE Power Engineering Society General
Meeting, 2007.
A. Poulin, M. Dostie, M. Fournier, S. Sansregret, Load duration
curve: A tool for technico-economic analysis of energy solutions,
ScienceDirect, Energy and Buildings, Pp29-35, Vol.40, Issue 1, 2008.
Choi, J., Tran, T., Kwon, J., Thomas, R., Mount, T., Billinton, R.,
Nodal probabilistic production cost simulation considering
transmission system unavailability, Generation, Transmission &
Distribution Journal, IET, Volume 2, Issue 1, pp 32 - 42 Jan. 2008.
Dang Quoc Thong, Bach Quoc Khanh, Preliminary estimate of
impact of load curve leveling on urban distribution network energy
loss, Journal of Science and Technology, No 25/26, 2000, Hanoi,
Vietnam, December 2000.
R. Cespedes, H. Duran, H. Hernandez, A. Rodriguez, Assessment Of
Electrical Energy Losses In The Colombia Power System, IEEE
Transactions on Power Apparatus and Systems, Vol. PAS-102, No.
11, November 1983
M.W.Gustafson, J.S.Baylor, Approximating the system losses
equation, IEEE Transaction on Power Systems, Vol.4, No.3,
Ausgust, 1989.
C.C.B.Oliveira N.Kagan A.Meffe S. Jonathan S.Caparroz J.
L.Cavaretti, A New Method For The Computation Of Technical
Losses in Electrical Power Distribution Systems, Proceeding,
ClRED2001, Conference Publication No. 482, IEEE 2001, 18-21 June
2001.
Analytical Assessment of loss reduction solutions for 110kV,
medium voltage and low voltage networks in Vietnam, FY 2009
Report, Technical and Operation Department, Vietnam Electricity.
Bach Quoc Khanh received B.S., M.S. and

Ph.D. degrees in power network and
systems from Hanoi University of Science
and Technology, Hanoi, Vietnam in 1994,
1997 and 2001 respectively. He has been a
faculty member of Electric Power Systems
Department, Electrical Engineering Faculty,
Hanoi University of Science and
Technology since 2002. He is currently a
visiting scholar in the Center for Advanced
Power Systems, IESES, Florida State
University, USA. His research interests
include power system analysis, DSM, power
distribution planning, power system quality, distributed generation.
128
eBook for You
REFERENCES
[1]
IEEE PEDS 2011, Singapore, 5 - 8 December 2011
A Novel Model to Determine Optimizing Power

and Capacity for Energy Storage Systems on
Competitive Electricity Markets
Abstract - Distributed generators (DGs) and renewable energy

resources
(Hydroelectric,
photovoltaic,
solar
thermal
technologies, and wind farms) have been applied widely on
planning and operating of distribution systems due to the
restructure of electricity market and the development of
technology. The energy storage systems (ESSs) can store energy
and then retrieve it in the other time, increase flexibility of
distribution systems. Hence, many researches and applications
have been conducted. Effects of ESSs on planning and operating
the wind generators therefore should be carefully investigated.
This paper proposes a novel approach to calculate maximum
profits of the wind generators and determine optimizing power
and capacity of ESSs. This model uses an objective function that
includes the total profits for electric energy sales, total costs
(investment, operation and maintenance) of ESSs, transmission
access fee, and upgrading cost of connected substation
transformers. The proposed model is applied to test two 1500kW
wind generators. The calculation is programmed by GAMS
environment.
Keyword: Competitive Electricity Markets (CEMs), Distributed
Generators (DGs), Energy Storage Systems (ESSs)
I.
INTRODUCTION
Recently, the environmental pollutions, technological

development and restructure of electricity markets have been
enhancing the application of DGs and renewable energy
resources in planning, designing and operating distribution
systems [1]. The DGs and wind generators have been applied
widely on the globe. In 2008, total power of the wind
generators have been installed in about 121.19GW and the
installing power of wind generators is forecasted by 2020
about 1500GW (increase 29 percent per year) [2]. However,
intermittent nature and the uncertainties associated of DGs
(wind generators, photovoltaic, hydroelectric) lead to low
stability, power quality and reliability of distribution systems
[3, 4]. ESSs can store energy and then use it in the other times
so they increase flexibility of distribution systems and DGs.
Hence, many researches of ESSs have been implemented in
last decade. Total power of the ESSs has been installed about
90GW in 2008, equivalent 2.6% of total electricity power on
the world [5, 6]. Peak shaving of distribution systems can be
done by ESSs (storing energy during low-demand periods
and releasing during high-demand periods) so they can defer
to upgrade network (feeders and transformers). Indices for
D. Q. Thong and B. Q. Khanh

Hanoi University of Science and Technology (HUST)
Hanoi, Vietnam
bq_khanh-htd@mail.hut.edu.vn
economy, power quality, reliability and stability of
distribution systems are improved [7]. Besides, in the
restructure of electricity market, ESSs store energy during
times of low price (low-demand) and retrieve during times of
high price (high-demand) so economic efficiency of
distribution companies and profits of DGs are expanded.
Therefore, many authors researched and applied the ESSs on
planning, operating distribution systems and the DGs for
CEMs. The ESSs have been applied to improve stability and
power quality of renewable energy resources [8, 9, 10], to
enhance power quality and reliability for distribution systems
[11, 12, 13, 14], and load demand balance [15]. Optimal size
of ESSs with objectives as above has been determined. In the
CEMs, volatility energy price depends purchase time from
market. ESSs have been used in distribution substations [16]
to load shifting, to decrease energy purchase cost on
electricity markets. This models objective is maximum
amount of total profits (reducing for energy purchase fees,
transmission access fees, and deferring facility investment
fees) and costs (increase capital, operating and maintenance
fee of ESSs) of distribution substations. The operation and
maintenance cost, environmental pollutions of DGs using
fuels (Diesel, Microtuabin, Combined Heat and Power, Fuel
Cells) are highly impacted by operation modes. Therefore,
application of ESSs in optimization operation for these DGs
has proposed in [17] with objective function including total
cost of investment and operation of ESSs, production saving
and emission cost of DGs, and distribution network saving
cost. However, applying of ESSs to improve operation
efficiency to wind generators has not been much published.
This paper proposes a novel approach to calculate maximum
profits of the wind generators and determine optimal power
and capacity of ESSs on planning and operating of wind
generators.
The next parts of this paper are organized as follows.
Section II introduces characteristics, model of wind
generators and ESSs. The proposed calculate model is
introduced Section III. Section IV represents calculation
results and discussions of proposed approach, and conclusion
is reported in Section V.
978-1-4577-0001-9/11/$26.00 2011 IEEE
129
322
eBook for You
V. V. Thang and N. H. Kong

Thainguyen University of Technology (TNUT)
Thainguyen, Vietnam
thangvvhtd@tnut.edu.vn
THE WIND GENERATORS AND ENERGY STORAGE

SYSTEMS
A. Wind generators
The wind energy is converted to electric energy by wind
turbines and generators. There are two basic configurations of
wind turbines, the horizontal axis and the vertical axis. The
vertical axis turbines are isotropic, receiving wind from any
direction. The gearbox and generator can be mounted at or
near ground so they have higher durability and lower
investment cost than horizontal axis ones. However, some
main disadvantages of this type consisting of more expensive
produce cost and smaller power ratio compared to horizontal
axis counterparts. Therefore, the only horizontal axis wind
turbines has been researched and applied widely on the
industry [18]. The change of wind speed depends strongly on
climatic conditions. These high variations are dependent
geographical location, both spatially and temporally. Hence,
wind generators power output are also naturally intermittent
and highly variable [19]. Wind generators cannot operate
when wind speed is lower than the Vin speed or higher than
the Vout speed. Power output of wind generators is
represented as follows [20]:
0
,Vi ,h < Viin
PGi ,h = ( i .Vi ,h i ).PGimax ,Viin Vi ,h Viout

(1)
max
out
,Vi ,h > Vi
PGi
B. Energy storage systems
ESSs have been used as a new solution to optimize
distribution systems expanded plan and operation. The ESSs
offer additional benefits to owners because they can improve
the economic efficiency, reliability, stability and flexibility of
both distribution systems and DGs. In practice, many possible
ESS technologies have been researched and developed. The
energy is stored by many types (potential energy, kinetic
energy, mechanical energy, chemical energy, thermal energy
and electrical energy) depending on technological
development. Energy can be stored by potential energy for
pumped hydro storage (PHS), kinetic or mechanical energy
for flywheel energy storage (FES) and thermal energy for
thermal energy storage (TES). The air is compressed within
an air reservoir and it is discharged to operate combustion gas
turbines for compressed air energy storage (CAES). These
technologies mentioned above are successfully applied in
France, America, Japan, Scotland and Wales [6]
Nevertheless, when distribution systems need meet the high
demand load in short-terms, primary stored energy types as
above usually do not satisfy. Storage technologies of electric
energy using super capacitors (SC) or superconducting
magnetic energy storage (SMES) have been researched and
developed as alternatives. These technologies offer large
power and high efficiency factor (about 95%) but small
storage energy and short operation duration so they are not
appropriate to load shifting, peak shaving or generator
powers characteristics changing of DGs. Chemical energy
based electric storage technologies commonly use battery
energy storage systems (BESS) with lead-acid batteries,
sodium-sulphur batteries (NaS), lithium-ion batteries (Li-ion)

and flow batteries (VRB, ZnBr). Although, the investment
cost of BESS is high, they offer low operation and
maintenance costs, high power and capacity, long operation
duration without harmful emissions or noise [6] As a result,
this technology has been used popularly. BESS technology
will be chosen to investigate in this research.
III.
THE PROPOSED ANALYZING MODEL
In competitive electricity markets, objectives of investors

are maximum profits of electrical energy sales benefits and
costs of investment and operation. Electric price changes
depend different sale times on CEMs and power output of
wind generators depends to climatic conditions. Therefore,
this research proposes to use BESS technology to increase
profits for investors. This means that electric energy of wind
generators is stored on low-price periods and sold to CEMs
on high-price periods. The proposed model uses an objective
function that includes the total profits of electric energy sales,
total costs (investment, operation and maintenance) of ESSs,
and upgrading cost of connected substation transformers.
Constraints of the model consist of power balance of
distribution systems, power and capacity balance of ESSs,
and power and capacity limits of ESSs.
A. Increased profits of electric energy sales
Electricity price and load demand for power systems are
mutually dependent as a linear function [2]. When load
demand is increased, electricity price is raised and vice-versa.
Assumption that, electricity price depends load demand as
formula (2).
h = Pr i .Ph + Pr i h H
(2)
The BESS charges on periods for low price, negative
BESSs power, so that power of wind generators selling to
markets is described by formula (3). The storage electric
energy is discharged at the high price with positive power and
power factor kF. The formula (4) expresses sale power for
wind generators in this case.
(3)
PS h = PGh PBh h H
PS h = PGh + k F .PBh h H
(4)
On overall planning period, therefore, total profits for
electric energy sales in case of changing generator powers
characteristics of wind generators by BESS are presented as
(5). These profits are converted to the first time of planning
via discount rate r.
T
H
1
h .PS h
BP =
.
365
.
( 1 + r )t
(5)
t = 1
h =1
t T ,h H
B. Increased cost of wind generators
All costs on model also are calculated by the first time of
planning using discount rate r.
1. Investment cost of BESS
Electric energy can be charged, discharged and stored by
BESS. Hence, investment cost of BESS is related to
130
323
eBook for You
II.
CB =
t =1 ( 1 + r )
. PB max .C PC + EB max .C EC
t T
(6)
2. Operating and maintenance cost of BESS

This operating and maintenance costs of BESS include
fixed cost depending on power, variable cost depending on
capacity and operation time.
T
COM =
t
t =1 ( 1 + r )
. PB max .C POM + EB max .C EOM
(7)
t T
where, CEOM is increased at operation time and it is
assumed to go up by 20% per year [17].
3. Electricity energy loss cost on BESS
The efficiency factor of BESS is about 65% to 85% [5, 6,
7] so electricity energy loss cost on itself is determined as
follows:
T
H
1
CF =
.
365
.
( 1 k F ).PBh . h (8)
( 1 + r )t
t = 1
h =1
t T ,h H
4. Transmission access fee
The transmission access fees in the competitive electricity
markets rely on power value and accessing times. During the
periods of high electric price, the accessing fees are raised
and vice-versa. Therefore, this cost is increased because
BESS delivers power at high electric price times.
T
H
1
CT =
.
365
.
PS h . F ,h
( 1 + r )t
(9)
t =1
h =1
t T ,h H
5. Upgrading cost of connected substation transformers
The transmission power through substation transformers
can be increased when BESS discharges at high electric price
times. At the same, generator power of wind generators is
also high. Therefore, upgrading cost of connected substation
transformers is calculated as follows.
T
CS =
t =1 ( 1 + r )
PS max S
.C S t T
Cos
(10)
C. The proposed objective function

The proposed model uses an objective function that
includes the total profits for electric energy sales, total costs
(investment, operation and maintenance) of BESSs, and
upgrading cost of connected transformers. This objective
function is combined equations that are defined in previous
sections, from (5) to (10).
J = BP CB COM CF CT CS
=
h =1
( 1 + r )t . 365. h .PS h ( PB max .C PC + EB max .C EC )

t =1
( 1 k F ).PBh . h
( PB max .C POM + EB max .C EOM ) 365.
(11)
h =1
H
PS h . F ,h
365.
h =1
PS
.C S Max
Cos
max
t T ,h H
D. Constraints of the proposed model

The proposed model has determined optimizing size of
BESS by previous economic requirements. Technical
requirements must be guaranteed simultaneously by
following constraints.
TABLE I. SETS, INDICES, VARIABLES AND PARAMETERS
No Symbol
Definition
I. Sets and Indices
Set of wind generators (i N)
1
i
Total number of wind generators
2
N
Set of calculation time, hour
3
h
Total of time on a calculation period, hour
4
H
Total charge times of BESS, hour
5
H1
Total discharge times of BESS
6
H2
Planning period, year (t T)
7
t
Overall planning period, year
8
T
II. Variables
Active power of wind generators sale to CEM at the
9
PSh
time h, kW
Active power of BESS at the time h (kW)
10
PBh
Capacity of BESS, kWh
11
EB
III. Parameters
Discount rate, %
12
r
Capital cost at active power of BESS, $/kW
13
CPC
Capital cost at capacity of BESS, $/kW
14
CEC
Operating and maintenance cost at active power of
15
CPOM
BESS, $/kW
Operating and maintenance cost at capacity of BESS,
16
CEOM
$/kW
Estimation parameters for electrical price model
17
Pri, Pri
Estimation parameters for wind generators power
18
i, i
output model
Electrical energy price for CEMs at time h, $/MWh
19 h
Transmission access fee at time h, $/kW.month
20 F.h
Power factor of BESS:
- BESS charge, kF = 1
21
kF
- BESS discharge, 0 < kF < 1
Minimum operation wind speed of wind generators,
22
Viin
m/s
Maximum operation wind speed of wind generators,
23
Viout
m/s
Wind speed of wind generators at period h, h
24
Vi,h
Active power for wind generators at time h, MW
25
PGh
Active power of system at time h, MW
26
P,h
Maximum wind generators active power limit, kW
27
PGimax
Active power for wind generators at time h, kW
28
PGi,h
Maximum active power limit of BESS, kW
29
PBmax
Maximum capacity limit of BESS, kWh
30
EBmax
Maximum transmission power through connected
max
31
PS
substation transformer, kW
Upgrading capital cost of substation transformer,
32
CSS
$/kW
Power factor
33 Cos
Load demand at time h, kW
34
PD,h
Active power of wind generators i sale to CEM at the
35
PGi,h
time h, kW
131
324
eBook for You
maximum power and capacity. This capital cost is presented

as follows:
WGs
1
Sub
2
i =1
BESS
(12)
i N ,h H
2. Constraint of BESS capacity balance

The power output of wind generators depends on climatic
conditions so operation cycle is assumed by one day (24h). In
operation cycle, total charge and discharge electric energy of
BESS must be balanced as follows:
H2
PBh + k F .PBh = 0
h =1
h =1
A. Calculation parameters of proposed model

(13)
h H , H 1 + H 2 = H
3. Constraint of BESS power and capacity limits
With the assumption that operation cycle of BESS is one
day. The BESS power and capacity limits depend on
generator powers characteristics of wind generators.
PGh PBh PGh
EB EB max =
1. The wind generator parameters

The average wind speed in Ninh Thuan, Vietnam has been
investigated in May 2011 and was regulated at operation
position of wind generators (height 80m corresponding to
correction factor kW = 1.4) on figure 2.
7.0
6.0
5.0
PBh
(14)
h =1
h H
4.
h H
5. Constraints of BESS power limit on planning period
During
planning
duration,
generator
powers
characteristics of wind generators in an operation cycle are
assumed constant. This constraint is given as follows.
PBh ,t = PBh ,t 1 h H ,t T
(16)
The proposed model including equations from (1) to (16)
is a nonlinear programming model. In order to investigate this
model, Nonlinear Programming with Discontinuous
Derivatives (DNLP) solver in GAMS program language [1,
22] is used to find out an optimal solution.
IV.
RESULTS AND DISCUSSIONS
The proposed model is applied to test two 1500 kW wind

generators in Ninh Thuan, Vietnam that have been connected
to distribution systems by substation transformers as figure 1.
4.0
3.0
2.0
1.0
0.0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time (h)
Figure 2. Data of average wind speed on a day

(source: http://www.foreca.com/Vietnam/Phan_Thiet
Wind turbines and generators in Tuy Phong, Ninh Thuan

use FL-MD77 turbine type and JFRA 580 generator type
(Fuhrlaender-German). The nominal capacity of one wind
generator is 1500 kW, nominal voltage is 0.69 kV, maximum
and minimum operation wind speed limits are 20m/s and
3m/s, respectively. Generator power on one day of two wind
generators has been calculated by (1) with assumption that
values for and corresponding to 0.2973 and 1.0378 as
illustrated in figure 3.
2500
2000
Power
(kW)
g
( )
Constraints of BESS power limits and dynamic

power updates
This constraint allows decreasing computing time and it
ensures upgrading power corresponding to equipment
parameters.
PBh PBh 1 + PB
PBh PBh 1 PB
(15)
Wind speed (m/s)
H1
Figure 1. Diagram connected wind generators, BESS and distribution systems
1500
1000
500
0
1
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time (h)
Figure 3. Wind Generators Power
132
325
eBook for You
PGi ,h + k F .PBh + Ph = PDh
Distribution
systems
1. Constraint of distribution systems power balance

The total power output of wind generators and BESS
through substation transformers must satisfy the load demand
in every period of operating time. Thus, the below equation is
presented as the system power balance.
140
12000
120
10000
100
8000
80
6000
60
4000
40
2000
20
Electric energy Price ($/MWh)
Demand of Viettnam Power Systems (MW)
14000
0
1
Demand (MW)
3. The parameters for BESS

The investment, operation and maintenance costs of BESS
are represented on table II.
TABLE II. INVESTMENT, O&M COST OF BESS [6, 16]
1
2
Power ($/kW)
Capacity ($/kWh)
Investment costs
CP C, CEC
500
100
Operation and
Maintenance costs
CP OM, CEOM
10
0.5
4. Transmission access fees

The restructure of electricity market has separated
independently
transmission
companies,
distribution
companies and generation companies. Consequently,
distribution companies that purchase electric energy from
market must pay transmission access fees. Accessing fees
depend on access time and capacity as on table III.
TABLE III. TRANSMISSION ACCESS FEE [16]
No
1
2
3
4
Periods (h)
1-6
7-9
10-22
23-24
Capital, operation and maintenance costs of wind

generators in two cases are the same so they can be
ignored in analysis.
Sale electric energy on CEMs via substation

transformers is assumed to increase 2% annually.
Total profits are converted to present time (Net

Present Value - NPV) with discount rate is 10%,
annually. The calculating service lifetime of BESS is
assumed 20 years.
Upgrading area of substation transformers and

building area for BESS are not limited.
Wind speed in DGs operation location is assumed

stability, little change in every calculation cycles.
Decided variables (power and capacity of BESS) in

the proposed model are continuous in order to reduce
the complexity of the model (not need to use binary
variable). Hence, they should be rounded to match real
equipments on markets.
Time (h)
Electric energy price ($/MWh)
Parameters
Market demand is much higher than operation power

of wind generators. Power and capacity of BESS are
not limited in all calculation period. This assumption
allows using maximum wind generators capacity and
efficiency of BESS. Hence, it is suitable in reality
because the penetration of DGs is low in distribution
systems.
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Figure 4. Power system demand and electric energy price in

electricity market
No
Fees ($/kW.month)
2.0
2.17
3.55
2.0
Note
The parameters and upgrading cost for connected

transformers
With the assumption, that a 3200kVA capacity of existing
substation transformers can guarantee electric transmission of
two wind generators. Upgrading fixed and varied costs of
connected substation transformers are 200,000$ and
50,000$/MVA, respectively.
With these above assumptions, calculation and test results

of proposed model are shown as follows.
2.
Analyzing results
a) Case A
In this case, power of wind generators is sold out to
market as the characteristics in figure 3. Total electric energy
value that is provided to market on calculation period is
7,700,059$ and transmission access fees are 265,731$.
Therefore, total profits are 7,434,327$ as in table IV.
TABLE IV. TOTAL PROFITS, COSTS FOR CASE A AND CASE B
5.
B. The calculation results and discussions

The feasibility of the proposed model and efficiency of
BESS are tested in two cases with parameters on section A.
Sale duration of electric energy for wind generators is
estimated also in both cases. In case A, BESS is not
considered. In case B, BESS is mentioned in the researching
model.
No
1
2
3
4
Profits and Costs

Profits for electric
energy sale
Transmission access
cost
Investment cost of
BESS
Operation
and
Maintenance costs
of BESS
Upgrading cost of
connected
transformers
Total
Case A
($)
Comparison
cost between
Case B and
Case A ($)
Case B
($)
7,700,059.0
8,508,204.0
808,145.0
265,731.0
273,879.0
-8,148.0
0.0
634,082.0
-634,082.0
0.0
15,261.0
-15,261.0
0.0
0.0
0.0
7,434,327.0
7,584,981.0
150,654.0
133
326
eBook for You
1. Assumptions in analysis
This research utilizes some economic and technical
assumptions for the ease of computation:
2. Electrical energy price

The demand of power system in Vietnam is investigated
in December 2008. Electric energy price in electric market is
calculated in (2) with two assuming values of Pri and Pri
corresponding to 0.02 and -115.19. These parameters are
expressed in the figure 4.
2000
120
1500
90
1000
60
500
30
0
1
Power for W Gs
0
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time (h)
Power to sell on market )
6
Figure 5. Power for WGs and sale on market
Figure 5 illustrates that market price of electric energy at

durations from 0h to 5h is very low with minimum price is
24.41$/MWh at 4th hour. In these durations, BESS stores all
energy of wind generators so that electric power sold to
market is 0kW. At 6th hour, BESS only stores a part of wind
generators power of 630.08kW, corresponding to 83.20%
because the electric energy price on market start increasing
sharply. The price of electric energy at 10th to 11th hour and
16th to 20th hour is extremely high, maximum price is
133.95$/MWh at 18th hour. This is perfect times for BESS to
discharge and sell energy to markets. The maximum power is
sold by 451.56kW at 11th hour and 16th to 20th hour. After the
21st hour, electric energy price is lessened. This leads to the
reduction of selling power of BESS. At the same time, it
resumes storing energy that is estimated by 69.61kW
corresponding to 52.35% at 21st hour. In the remaining
periods including 7th to 9th and 12th to 15th hours, the electric
energy price is fairly high and there are power and energy
losses in BESS. Therefore, wind generators sell directly
electric energy to market through substation transformers.
180
500
150
400
120
300
90
200
60
100
30
Power (MW)
-100
2 3
4 5
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
-200
-300
-400
-500
V.
REFERENCES
[2]
[3]
-600
-700
Power for BESS )
Time (h)
Figure 6. Power to chanrge and rechanrge for BESS
Table IV shows that the investment and production costs

of BESS are 634,082.0$ and 15,261.0$, respectively. The
CONCLUSION
Planning of distribution systems have been changed

significantly for recent years because of the application of
DGs and renewable energy resources. However, power of
these sources is highly variable due to weather, so that
economic (electric energy sale) and technical (power quality,
reliability, and stability) indicators of distribution systems are
decreased. These problems can be solved by ESSs because
ESSs change the generator powers characteristics of wind
generators. This research has proposed a new approach to
determine optimizing size of ESSs and calculate maximum
profits of the wind generators within economic and technical
constraints. The testing results of this research show that
BESSs investments in planning and operating wind
generators have been improved indicated by economic and
technical outcomes (for both wind generators and distribution
systems). BESSs charges electric energy during periods of
low electric energy price (low demand) and recharges during
periods of high electric energy price (high demand). As a
result, total profits for electric energy sale are increased.
Moreover, BESSs can shift loading characteristics that result
to a decrease of peak demand of distribution systems as a
solution for Demand Side Management (DSM). Therefore,
upgrading cost of distribution systems (feeders,
transformers) is also decreased. At the same time, power
quality, reliability and stability of distribution systems are
improved and this solution contributes to the decrease of
environmental pollution as well.
[1]
600
electric energy is sold to markets at high load demand

durations (high electric energy price) so that transmission
access fees are raised corresponding by 8,148.0$. Total costs
on this case are 923,222.0$ and are higher than those of case
A by 657,491.0$. Nevertheless, profits by electric energy sale
on the calculation period have been increased by 808,145.0$
compared to case A, corresponding to 1.96%. Hence, case B
offers better economic efficiency.
As can be seen in the figure 5, maximum powers
transmitted through substation transformers are equal in the
both cases. Therefore, connected substation transformers need
not to upgrade.
[4]
V.V.Thang, D.Q.Thong, B.Q.Khanh, A New Model Applied to the

Planning of Distribution Systems for Competitive Electricity Markets,
The Fourth International Conference on Electric Utility Deregulation
and Restructuring and Power Technologies (DRPT) 2011, Weihai,
Shandong, China, 6-9 July, 2011
World Wind Energy Association WWEA, World Wind Energy Report
2008, 8th World Wind Energy Conference & Exhibition Wind Power
for Islands - Offshore and Onshore, Jeju Island, South Korea, 23-25
June 2009
Masato Oshiro, Tomonobu Senjyu, Atsushi Yona, Naomitsu Urasaki,
Toshihisa Funabashi, Optimal Operation Strategy by Battery Energy
Storage Systems in Distribution System, IPEC, 2010 Conference
Proceedings, Date:27-29 Oct. 2010
Cristina L. Archer and Mark Z. Jacobson, Supplying Baseload Power
and Reducing Transmission Requirements by Interconnecting Wind
Farms, Volume 46, Journal of Applied Meteorology and Climatology,
November 2007
134
327
eBook for You
150
Power (MW)
g
( )
2500
b) Case B
The optimal decision of the proposed model shows the
optimal selling power of wind generators through substation
transformers in a calculation cycle depicted in figure 5. In this
case, BESS is considered as an optimal selection to improve
efficiency of wind generators. Although, high investment
cost, electric energy sale profits are high. Thus, operation
power of BESS in a calculation cycle is selected as in figure
6. The optimal power and capacity of BESS are decided to
invest by 632.43kW and 3178.66kWh, respectively.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
System, International Conference on Electrical Machines and Systems

(ICEMS2010), 2010
P. Hu R. Karki R. Billinton, Reliability evaluation of generating
systems containing wind power and energy storage, IET Generation,
Transmission & Distribution, 2009
Srgio Faias, Jorge Sousa1, Rui Castro, Contribution of Energy
Storage Systems for Power Generation and Demand Balancing with
Increasing Integration of Renewable Sources: Application to the
Portuguese Power System, European Conference on Power
Electronics and Applications, 2007
Rong-Ceng Leou, An Economic Analysis Model for the Energy
Storage Systems in a Deregulated Market, IEEE International
Conference on Sustainable Energy Technologies. ICSET 2008
S. Chakraborty, T. Senjyu, H. Toyama1 A.Y. Saber, T. Funabashi,
Determination methodology for optimising the energy storage size for
power system, IET Generation, Transmission & Distribution, 2009
Tracy Dahl, Wind Power Systems, Technology white paper,
PolarPower.org
M. H. Albadi, and E. F. El-Saadany, Impacts of Wind Power
Variability on Generation Costs: an Overview, International
Conference on Communication and Power 2009 (ICCCP'09)
Ayed A. S. Algarni, and Kankar Bhattacharya, Disco Operation
Considering DG Units and Their Goodness Factors, IEEE
Transactions on Power Systems, Vol. 24, No. 4, 2009
Ayed A. S. Algarni, and Kankar Bhattacharya, A Generic Operations
Framework for Discos in Retail Electricity Markets, IEEE
Transactions on Power Systems, Vol. 24, No. 1, 2009
Richard E. Rosenthal, GAMS - A User's Guide, GAMS Development
Corporation, Washington, USA, 2010
eBook for You
[13]
Electric Energy Storage, Climate TechBook - Pew Center on Global

Climate Change, 2009
H. Ibrahim, A. Ilinca, J. Perron, Energy storage systems Characteristics and Comparisons, Renewable and Sustainable Energy
Reviews, PP 1221-1250, 2008
Dr. Rahul Walawalkar and Dr. Jay Apt, Market Analysis of Emerging
Electric Energy Storage Systems, National Energy Technology
Laboratory, www.netl.doe.gov, 2008
Jie Zeng, Buhan Zhang, Chengxiong Mao and Yunling Wang, Use of
Battery Energy Storage System to Improve the Power Quality and
Stability of Wind Farms, International Conference on Power System
Technology, 2006
Kejun Qian, Chengke Zhou, Zhaohui Li, Yue Yuan, Benefits of
Energy Storage in Power Systems with High Level of Intermittent
Generation, 20th International Conference on Electricity Distribution,
Prague, 8-11 June 2009
Yusuke Hida, Yuki Ito, Ryuichi Yokoyama, Kenji lba, A Study of
Optimal Capacity of PV and Battery Energy Storage System
Distributed in Demand Side, 45th International Universities Power
Engineering Conference (UPEC2010), 2010
D. Casadei, Member, IEEE, G. Grandi, Member, G. Serra, C. Rossi,
Power Quality Improvement and Uninterruptible Power Supply Using
a Power Conditioning System with Energy Storage Capability, Power
Tech, IEEE Russia, 2005
D. Sutanto, Energy Storage System to Improve Power Quality and
System Reliability, 2002 Student Conference on Research and
Development Proceedings. Shah Alam. Malaysia, 2002
Gyeong-Hun Kim,, and Sangjin Lee, Application of a Battery
Energy Storage System for Power Quality Improvement of Jeju Power
135
328
Mitigating Voltage Sags Due to DOL Starting

of Three Phase Asynchronous Motors Using
Dynamic Voltage Restorer (DVR)
Bach Q. Khanh, J. Lian, B. Ramachandran, S. Srivastava, D. Cartes
Index TermsVoltage sags, Dynamic voltage restorer, Custom

power devices, Power quality.
I. INTRODUCTION
OLTAGE sag (dip) is the power quality phenomena that

is defined by IEEE-1159 (1995) as a decrease in RMS
voltage to between 0.1 to 0.9 of nominal voltage at
power frequency for duration of 0.5 cycle to 1 minute. The
theory of voltage sags for electrical networks is thoroughly
described in [1], [2]. It is summarized that voltage sags are
caused by faults in electrical grids, the starting of large loads
such as motors or transformer energizing. It is a common
reason for failures or malfunctions of power electronics based
equipments that are used widely in modern day power
systems.
A typical shape of voltage sag due to starting a three phase
This work is supported by Vietnam Education Foundation for the Visiting
Scholar Program 2010-2011.
Bach Quoc Khanh is a faculty member with Electric Power Systems
Department, Electrical Engineering Faculty, Hanoi University of Science and
Jainming Lian (email: jainming.lian@gmail.com), Bhuvaneswari
Ramachandran (e-mail: ramachan@caps.fsu.edu) and Sanjeev Srivastava (email: sanjeev@caps.fsu.edu) are assistant scholar scientists with Center for
Advanced Power Systems, Institute for Energy Systems, Economics and
Sustainability, Florida State University, 2000 Levy Ave, Tallahassee, FL32310, USA.
David A. Cartes is an associate professor with Center for Advanced Power
Systems, Institute for Energy Systems, Economics and Sustainability, Florida
State University, 2000 Levy Ave, Tallahassee, FL-32310, USA (e-mail:
dave@ieses.fsu.edu).

978-1-4673-1935-5/12/$31.00
2012 IEEE
Fig. 1. Voltage sag due to starting of a three phase induction motor

simulated in the test system in Section IV
induction motor is shown in Fig. 1. The characteristics of

these voltage sags depends on various factors, of which the
motor rating, rotor structure, the method of starting motor and
systems power supply capacity are seen as critical impacts.
In industrial power distribution systems, the application of
asynchronous motor (induction motor) perhaps accounts for
the largest part of system load due to its low cost and rugged
construction. However, voltage sags in supply lines caused by
the starting of asynchronous motor are generally the most
severe because many of its applications are started using DOL
starter that normally draw high current (6 to 8 times of fullload current) during the starting period. These voltage sags
may severely affect the working of other equipment supplied
from the same feeder, causing loss of production due to
unexpected tripping and malfunction of process control
equipment that may result in huge economical losses. The
high frequency of motor starting in industrial electric
networks makes the problems even more pressing.
Many different researches and applications have been
developed for mitigating voltage sags in general as well as
that caused by motor starting in particular. The development
of power electronic application in recent decades has made it
possible for introducing new effective compensation methods
in power systems using custom power devices. Among these,
the distribution static compensator (D-STATCOM) and the
dynamic voltage restorer (DVR) which operate based on
voltage source converter (VSC) principle [3] are most
effective devices. In general, there is no standard solution that
works for every site, instead each mitigation action is
carefully designed and evaluated. Recent solutions for voltage
sag mitigation using custom power devices have been
analyzed and proposed through many publications [7]-[10].
However, they mainly consider the mitigation of voltage sag
136
eBook for You
AbstractVoltage sags caused by starting large asynchronous

motors, especially in industrial distribution system have huge
impacts on system power quality. The paper analyses and
proposes an extended application of custom power devices using
dynamic voltage restorer for mitigating voltage sags due to the
direct-on-line (DOL) starting of asynchronous motors in
industrial distribution system. In the paper, different control
strategies used for custom power devices are analyzed and an
effective method is proposed. A test system is simulated in
Matlab/Simulink with different DVR configurations that proves
the effectiveness and robustness of the proposed control strategy
with dynamic voltage restorer for voltage sag mitigation. The
results show that voltage sags are totally compensated.
dynamically controlled voltage in series to the bus voltage by

means of a booster transformer for voltage restoration and
regulation. The amplitude of the injected phase voltage is
controlled such as to eliminate any detrimental effects of
power quality disturbance to the voltage of a certain location
on the network. The location is normally load buses. Fig. 2 is
a principle diagram of DVR operation for compensating
voltage sag due to fault in AC supply to the sensitive load.
Regarding DVR application for voltage sag mitigation,
various discussions in recent publications have concerned
only with mitigation voltage sag due to faults in power
system. DVR acts as an additional voltage to the supply
voltage in order to keep the load voltage with its permissible
tolerance [7]-[10]. So,
VL = VS + VDVR
(1)
where
Voltage of protected load
V L:
System voltage
VS:
VDVR: DVRs compensating voltage
For mitigating voltage sags due to starting of large loads like
motors, there are two contexts for DVR application. The first
one regards the starting of induction motor as a source of
voltage sags [13], [14], [17] and DVR is used to protect only
one or a couple of sensitive loads. So the motor is connected
to the source side of DVR as Fig. 3. The way of voltage
compensation remains as (1). This case is normally used when
not many sensitive loads exist in the system.
ZS
DVR
II. DVR MODEL

A. Definition of DVR Application
DVR is a series static compensator. The general structure of
a DVR consists of a booster transformer, a harmonic filter, a
voltage source converter (VSC), a energy storage and a
control system. Main functions of DVR include voltage
regulation in steady state operation and voltage restoration in
transient operation. In either operating modes, DVR injects a
dynamically VS
VL
Booster
Transformer
Sensitive
Load
Supply
VDVR
Filter
Control
DVR
Voltage
Controlled
Bus
VSC
Energy
Storage
Fig. 2. Principle of operation of the dynamic voltage restorer
VDVR
IM
Sensitive
Load
VPCC
VIM
Fig. 3. DVR application to protect individual sensitive loads
The other context regards the bus between the started

induction motor and the system as the objective of protection
and the DVRs operation principle is changed as illustrated in
Fig. 4. The bus of protection is the PCC and DVR will
compensate the missing voltage due to the starting of motor
such as to keep the voltage at the point of common coupling
(PCC) VPCC unchanged. So, (1) is modified as follows
VIM = VPCC VDVR
(2)
where
VIM:
VPCC:
Motor terminal voltage

Voltage at the point of common coupling to other
sensitive load.
VDVR: DVRs compensating voltage
Fig. 4 also includes voltage vector diagrams for this
application of DVR. This application works more effectively
in the case of a large number of sensitive loads operating in
the system. It can reduce the number of required DVRs.
To avoid DVRs impacts on system parameters in the
steady state condition, (phasor shifting, losses, harmonics),
DVR is normally active during event duration and inactive in
steady state operation. When activated, DVR will compensate
137
eBook for You
due to faults in power system. With regard to mitigating

voltage sag due to motor starting, some researches [15], [16]
introduced interesting solutions using D-STATCOM.
However, the effectiveness of these solutions is limited
because they can only reduce voltage sag duration or increase
the remaining voltage level during a voltage sag event. Also,
they are not flexible solutions because the compensation
effectiveness can be reduced when motor rating changes. [17]
introduced a study using DVR for mitigating voltage sag due
to starting of induction motor. However, this research focuses
on control solutions for eliminating factors influencing on
DVRs performance as well as DVRs impacts on system
voltage in steady state condition. Therefore, it is the
motivation for this paper to introduce a new solution for
mitigating voltage sag due to starting of three phase
asynchronous motor using DVR which can completely
compensate the missing voltage and maintain its
compensation effectiveness consistently as motor rating
changes.
This paper is arranged in the following manner. In Section
II, the background of compensation problem using DVR is
reviewed and a model DVR structure is selected. In Section
III, different control strategies for custom power devices are
discussed, and then, a suitable control strategy is proposed for
mitigating voltage sag due to starting of a three phase
induction motor. The simulation model and implementation of
the control strategy in Matlab/simulink is also presented in
this section. Section IV presents discussions on the simulation
results. Finally, conclusions are given in Section V.
as (2). After the motor speed reaches near the rated speed,
terminal voltage of the induction motor VIM increases close to
1 p.u, DVR is switched to stand-by mode (VDVR = 0).
Starting
Motor
ZS
DVR
IM
III. DVR CONTROL
VIM
VDVR
VPCC
a)
Sensitive
Load
I.ZS
VS
VS
I.Zs
VPCC
VPCC
VDVR
VIM
VPCC = VIM
b)
c)
d)
Fig. 4. DVR application for mitigating voltage sag due to starting of induction
motor. a) System scheme, b) Phasor diagram before starting motor IM, c)
Phasor diagram at the point of starting motor IM without DVR, d) Similar to
c) with DVR. I supply current during starting motor.
B. Topologies of DVR
There are different topologies for DVR application as
discussed in [6]. For the problem under our consideration, the
paper only considers two typical system configurations for
DVR application as shown in Fig. 5.
Supply
IM
Starting
Motor
+
_
Series
Converte
a) DVR with energy storage and with variable dc-link

IM
Supply
Starting
Motor
Shunt
Rectifier
+
_
Series
Converte
b) DVR with no energy storage and supply-side-connected rectifier

Fig. 5. System configuration of DVR applications for mitigation voltage sag
due to starting induction motor
The topology of a DVR with energy storage is shown in Fig.

5a. This case assumes that the independent DC supply is
strong enough to keep DC bus voltage constant. Several
methods of DC energy storage are applicable, such as SMES,
batteries, or supercapacitors, which can be used in a DVR by
adding a separate high-power-rating converter to the system
[6]. The topology of a DVR without energy storage is shown
in Fig. 5b. A supply side connected passive converter has an
uncontrollable dc-link voltage and the passive converter will
charge the dc-link capacitor to the actual state of the supply
A. General Requirement and Method Reviews

In any compensation problems, satisfactory performance,
fast response, flexibility and easy implementation are among
key objectives. The control strategies for controlling custom
power devices basically implement three following steps
Step 1. Measurements of system variables and signal
conditioning.
Step 2. Extraction of reference compensating signals.
Step 3. Generation of firing angles for switching
devices.
Among above steps, the second is the core of the control
strategy that defines the method of compensation. There have
been several control theories developed so far for extraction of
reference signals like instantaneous power theory,
synchronous reference frame theory, etc [4], [12] that are
based on transformation of three phase quantities to other two
phase frame for better calculation and control of instantaneous
power quantities for compensation. However, these
publications only consider DSTATCOM controls and
depending on the controlled objectives, control techniques can
be introduced accordingly.
For voltage sag mitigation, a well-known method namely
phase shift control is proposed in [7], [11] that is easy to
implement and needs less signal measurement. However, it
works only for mitigating three phase balanced sags due to
faults in power system. For better control performance, [8],
[11] introduced the method of indirect decoupled current
control where self supporting DC bus is recommended.
However, it requires intensive computation, making the
operation quite complex. Also, for large sags due to starting of
an induction motor, that absorbs either active and reactive
power, not having energy storage can result in partial
compensation only. [15] and [16] proposed simple control
strategy using DSTATCOM for mitigation of voltage sag due
to induction motor starting, but the firing angle for pulse
generator is fixed and that results in partial compensation like
the reduction in voltage sag duration or sag depth. This partial
compensating effect is due to the fact that the control of shunt
compensators is normally based on the load current or load
power, but it is difficult to exactly determine the missing
power that causes the voltage sag during starting of induction
motors.
It is also noticeable that DVR is a very effective tool for
problems that requires voltage compensation only [5]. The
control strategy of series compensators such as DVR, which is
based only on voltage measurement is robust and easy to
implement. Mitigating voltage sag due to starting three phase
138
eBook for You
VS
voltage. This option of DC supply does not need a strong

energy storage system, but the dc-link voltage is variable
during sag. In order to reduce the variable voltage, the
capacitor added in dc-link should be sized greater or in other
words, small energy storage is still needed. The results of
simulation in Section IV will illustrate this feature.
induction motor belongs to the said problem of voltage

compensation and thus DVR is proposed in this paper. [13],
[14] and [17] proposed some control strategies corresponding
to (1) and assume that DVR is permanently connected in all
system operating condition. Therefore, their designs are rather
complicated for mitigating the impacts of DVR on systems in
steady state conditions. This paper proposes the application of
feed-forward control technique [10] using instantaneous
power theory for extracting reference compensating voltage.
Block 1
Va.PCC
--0
Vb.PCC
a-b-c
V0
B. Proposed Control Method

The control scheme is shown in Fig. 6.
Block 2
d-q-0
Vd
Vc.PCC
This control strategy is much simpler and result in very good

compensation performance during voltage sag period. As said
in the Part II.A, the DVRs impacts on system in steady state
conditions can be eliminated when DVR is controlled to be
active during voltage sag period. To do so, a voltage sag
detection block is included in DVRs control system.
Vq
--0
Block 6
_
+
--0
_
+
Three
phase
PLL
Block 7
a-b-c
V
V0
d-q-0
Va.Con
Vb.Con
Vc.Con
--0
Block 8
PWM
Generator
Block 5
V
--0
Vb.Ref
Vc.Ref
a-b-c
V
V0
Block 3
PWM1-6
switch
t
d-q-0
--0
Vd.Ref
Voltage Sag
Detection
Vq.Ref
DVR
Breaker
Block 4
Fig. 6. Control Scheme for mitigating voltage sag due to starting of induction motor
Input voltages measured at the PCC bus (VPCC) is

transformed (Block 1) from three phase frame to two phase
frame using Clarks transform (3).
V cos
V =
sin
Va
1 2 1 2
(3)
Vb
3 2 3 2
Vc
Block 2 transforms measured voltages from - frame to
synchronously rotating d-q frame using Parks transform (4).
Vd cos sin V
(4)
V =

q sin cos V
Similarly, Block 3 and Block 4 converts the reference
voltage to two phase d-q frame. The reference voltage VRef is
three-phase symmetrical fundamental frequency, sinusoidal
component of the source voltage.
Block 5 - phase locked loop (PLL) tracks continuously the
fundamental frequency of the measured supply voltages and
generates a unit sinusoidal wave in phase with system voltage.
Measured supply voltages and reference voltage are compared
to each other
Vdq.Con = Vdq.PCC Vdq. Re f
(5)
Va
V =
b
Vc
V
V =

2 1
3 0
and the differential voltage (VCon) is reversely transformed to

three phase control voltage (reference voltage) by Blocks 6
and 7 as the following equations (6) and (7).
sin Vd

cos Vq
(6)
0
1
V
2
(7)
3 2
1 2
V
3
1 2 3 2
This straight-forward comparison allows to obtain the
desired extract voltage (compensating voltage) without any
further control techniques (PI control or feedback loops). So,
the control strategy is simplified and consistent.
Block 8 is the PWM generator that generates firing signals
to control three phase voltage source converter for injecting
the compensating voltage in series with motor voltage. Block
voltage sag detection is used to control the operation of DVR
between active and inactive modes. The voltage sag is
detected when Root Mean Square (rms) value of motor
terminal voltage drops under 0.9 p.u. During voltage sag
duration, DVR is activated by opening the DVR breaker. In
steady state operation, DVR is short-circuited by closing the
DVR breaker.
IV. SIMULATION AND RESULT ANALYSIS
A. Test System
The test system simulation in the Matlab/Simulink is
printed out as Fig.7. The parameters of the main components
are listed in the Table I.
139
eBook for You
Va.Ref
eBook for You
oltages
Fig. 7. Matlab/Simulink model of the test system

TABLE I
TEST SYSTEM PARAMETERS
Component
Ratings
AC phase-to-phase source voltage

Source resistance
Asynchronous motor general ratings
Motor stator resistance
Motor stator inductance
Motor rotor resistance referred to stator
Motor rotor inductance referred to stator
Mutual inductance
Inertia Torque
Boost Transformer
Filter inductance
Filter resistance-capacitance
Voltage source converter
PWM generator, switching frequency
DC bus battery voltage,
DC bus storage capacitance
Rectifier
Rectifier transformer
400 V, 50 Hz
1
7.5 kW, 400V, 1445rpm, 2P
0.7384
3.045mH
0.7402
3.045mH
124.1mH
0.0343kg/m2
10kVA, 400/400V
8mH
10, 20F
IGBT based
6 pulses, 1080Hz
300V
2000F
Diode based
1kVA, 400/230V
In the test system, DVR is simulated to work in two cases

of DC energy supply to voltage source converter as proposed
in Section II.B. DC supply selector and transfer switch blocks
make it possible for selecting the energy storage type. DVR
activate block is created for testing the system operation with
or without DVR. Furthermore, because the DVR is designed
to work only during the voltage sag period, an input of the

motor terminal voltage to DVR activate block is made for
detecting voltage sag duration and controlling the duration of
DVR operation. The sag is detected using the rms method in
[9]. The voltage compensation block with inputs of three
phase voltages measured from the point of common coupling
is modeled as shown in Fig. 6 to generate firing pulses for
controlling the converter. Voltages and currents are measured
at three AC buses (PCC bus, motor bus and DVR bus) to
verify the compensation performance of DVR.
B. Simulation Result and Analysis
The test system is initially simulated in the case of starting
induction motor without DVR as the base case. The main
system operating variables are shown in Fig. 8. In this
operation mode, the induction motor is started at 0.5s. DVR
breaker is in close position to bypass DVR. During the motor
starting operation, the motor absorbs a large reactive power,
so starting current (Im) rises highly (about 10 times steadystate current). Increased electro-magnetic torque (Te)
accelerates the motor speed (nm) to the rated value in very
short time (about 5 cycles). As the result, the voltage at PCC
(Vpcc) experiences a non-rectangular voltage sag down to 0.7
p.u. for 5 cycles as Fig 8 or a close-up as Fig 1.
140
Vpcc
(p.u)
VApcc
(p.u)
Vpcc
(p.urms)
VApcc
(p.urms)
Im(A)
VAdvr
(p.u)
Im
(Arms)
VAdvr
(p.urms)
VAm
(p.u)
nm
(rpm)
VAm
(p.urms)
time(s)
Fig. 8. Test system parameters when motor starts without DVR
The next operating condition is tested where the system

works with the battery DC supplied DVR. DC supply selector
is switched to Battery position. Before running the simulation,
DVR breaker still remains in close position. When motor
starts (also at 0.5s), a voltage sag is detected if motor terminal
voltage Vm drops below 0.8 p.u. and DVR is activated
accordingly by opening DVR breaker. At the end of starting
period, Vm is restored above 0.8 p.u. and DVR breaker is
switched back to the close position and the compensation
process terminates. The results of simulation are given in Fig.
9 and 10.
Vpcc
(p.u)
Vpcc
(p.urms)
Im(A)
DVR
deactivated
Im
(Arms)
nm
(rpm)
Te
(N.m)
time(s)
Fig. 9. Test system parameters when motor starts with battery supplied DVR
time(s)
Fig. 10. Bus voltages when motor starts with battery supplied DVR
The result shows that the voltage sag at PCC is almost

compensated. Besides, the starting current is considerably
reduced (just 2.5 times of steady-state current) and Te torque
is also much lower. The starting period lasts longer (about
0.5s).
Some important points are drawn from the results are
a) Compensation control will extract the control voltage
which is the differential voltage between the supply
voltage (Vpcc) and the reference voltage (Vref). The
control voltage is transformed to DVRs output that
offsets totally the motor terminal voltage during the
starting period. Fig. 10 depicts phase A voltages at all
concerned location (PCC, DVR output and motor
terminals) for illustrating the voltage compensating
effect. Therefore, the proposed control method works
very well for voltage compensation using DVR in the
case of dynamic sags caused by the starting of
induction motor that was partly sorted out previously
by D-STATCOM [13], [14].
b) The threshold for voltage sag detection is set low (0.8
p.u) for having the capability to deactivate DVR (close
DVR breaker) when motor terminal voltage restores at
the end of starting period. In fact, the voltage sag has
triangular form and at the end of starting period, a
stable voltage divider is created between DVR output
voltage (Vdvr) and motor terminal voltage (Vm). This
voltage divider makes value Vm impossible to reach the
voltage sag detection threshold in order to deactivate
DVR. Fig. 11 shows Vdvr and Vm when voltage sag
detection threshold is set to 0.9 p.u. Therefore, by
reducing this threshold to 0.8 p.u, it is possible to
capture the restored motor terminal voltage at the end
of motor starting period and thus, DVR can be
successfully deactivated. However, setting a low
141
eBook for You
Te
(N.m)
voltage sag detection threshold causes a small surge of

motor current and torque when DVR is deactivated
(Fig.9). However, they are much lower than the
corresponding values when the system is without DVR.
QS
(VAr)
Vdvr
(p.u)
PDVR
(W)
Vm
(p.u)
time(s)
Fig. 11. DVRs output voltage and motor terminal voltage when voltage sag
detection threshold is set to 0.9.
Fig. 12 depicts DC current supplied to DVRs

converter. DC voltage is assumed stable at 300V as DC
supply is a strong battery set. DC current experiences
ripples during DVR operation. These monitored values
are compared with those in the case of rectifier DC
supply in Fig. 14.
QDVR
(VAr)
PM
(W)
QM
(VAr)
time(s)
Fig. 13. Power flow when motor starts with battery supplied DVR
Idc (A)
Vpcc
(p.urms)
Vdc
(V)
eBook for You
c)
PS (W)
Im
(Arms)
time(s)
Fig. 12. DC bus voltage and current when DVR is supplied by battery
d) Fig. 13 demonstrates the system power flows. From (2)

we have
PS = PIM + PDVR
(8)
QS = QIM + QDVR
(9)
where PS, QS: Supply power, PIM, QIM: Motor starting

power, PDVR, QDVR: DVRs outcome power
With the existence of DVR, the motor starting power
does not increase quickly. Therefore, the high starting
current and torque are avoided.
The same operating condition that was tested with a battery
energized DVR is now tested with the case where DVRs
converter is energized from the rectifier. In Fig. 6, DC supply
selector is switched to Rectifier position. The rest of test
system remains the same as the case it works with battery DC
supply. The remarkably resulting variables of the test system
are reported in Fig. 14.
In this case, the voltage sag is also totally compensated. The
effectiveness of reducing starting current and torque is as
good as the case of using battery DC supply. The only
problem is that the DC bus voltage varies largely during
compensation duration and DC current also has higher ripples.
Vdc increases during the compensating period because of
capacitors discharging. If the capacitance is small, the
discharging is faster, and Vdc raises high faster before voltage
sag terminates. To reduce this voltage boost, the capacitance
nm
(rpm)
Te
(N.m)
Idc
(A)
Vdc
(V)
time(s)
Fig. 14. Test system parameters when DVR is energized through rectifier
should be sized greater and thus the discharging is slower.

Fig. 15 shows different voltage boosts with different
capacitances (2, 5 and 10mF).
Vdc
(V)
time(s)
Fig. 15. DC bus voltage for different capacitances
142
V. CONCLUSION
This paper presents a new solution for mitigating voltage
sag due to the starting of three phase induction motors in
industrial electric networks. For the problem of voltage
compensation like the case under papers consideration, using
the series compensator like DVR proves it a suitable solution.
The feed-forward control strategy also proves its reliability
and suitability for DVR application. The simulation results
shows that this compensation solution gives a better
performance in comparison with previous researches using
DSTATCOM because the voltage sag is totally compensated
(not partly compensated). Apart from the above said power
quality improvement, a better motor starting condition
(reducing starting current and electro-magnetic torque) is
obtained by this proposed solution as well. The
Matlab/Simulink simulation tested two DVR configurations conventional DC supply using battery and DVR using rectifier
DC supply and both prove the same performance. This work
can be further developed with regards to further different
optional energy storage applications and the combination with
DSTATCOM for including other compensation objectives in
industrial distribution systems.
VI. REFERENCES
M.H.J. Bollen, Understanding power quality problems - voltage sags
and interruptions, IEEE Press, 2000.
[2] R.C. Dugan, M.F. Granaghan, S. Santoso, H.W. Beaty, Electrical Power
Systems Quality, 2nd Edition, McGraw Hill, 2004.
[3] A. Ghosh, G. Ledwich, Power Quality Enhancement Using Custom
Power Devices, Kluwer Academic Publishers, 2002.
[4] H. Akagi, E. H. Watanabe, M. Aredes, Instantaneous Power Theory and
Applications to Power Conditioning, IEEE Press, 2007.
[5] A. Ghosh, G. Ledwich, Compensation of Distribution System Voltage
Using DVR, IEEE Trans. Power Delivery, vol. 17, no. 4, pp. 10301036, Oct. 2002.
[6] J.G. Nielsen, F. Blaabjerg, A Detailed Comparison of System
Topologies for Dynamic Voltage Restorers, IEEE Trans. Industry
Application, vol. 41, no. 5, pp 1272-1280, Sep/Oct 2005.
[7] T. Devaraju, V.C. Veera Reddy, V. Kumar, Modeling and Simulation
of Custom Power Devices to Mitigate Power Quality Problems,
International Journal of Engineering Science and Technology, vol. 2(6),
pp. 1880-1885, 2010,
[8] B. Singh, P. Jayaprakash, D. Kothari, A. Chandra, Kamal-Al-Haddad,
Indirect control of capacitor supported DVR for power quality
improvement in distribution system, Presented in Power & Energy
Society General Meeting, 2008.
[9] Mahmoud A. El-Gammal, Amr Y. Abou-Ghazala, and Tarek I. ElShennawy, Dynamic Voltage Restorer (DVR) for Voltage Sag
Mitigation, International Journal on Electrical Engineering and
Informatics, vol. 3, no. 1, 2011.
[10] R. Omar, N. A. Rahim, M. Sulaiman, Modeling and Simulation for
Voltage Sags/Swell Mitigation Using Dynamic Voltage Restorer (DVR),
Journal of Theoretical and Applied Information Technology, pp. 464470, 2005-2009.
[11] D. Masand, S. Jain, G. Agnihotri, Control Strategies for Distribution
Static Compensator for Power Quality Improvement, IETE Journal of
Research, Vol. 54, Issue 6, pp. 421-28, Nov/Dec. 2008.
[1]
[12] B. Singh, J. Solanki, A Comparison of Control Algorithms for

DSTATCOM, IEEE Trans. Industrial Electronics, vol. 56, no. 7, pp.
2738 2745, Jul. 2009.
[13] H. Kim, S.K. Sul, Compensation Voltage Control in Dynamic Voltage
Restorers by Use of Feed-forward and State Feedback Scheme, IEEE
Trans. Power Electron., vol. 20, no. 5, pp. 1169-1177, Sep. 2005.
[14] Jing Zhou, Hui Zhou, Zhiping Qi, The Study on a Dual-feed-forward
Control of DVR to Mitigate the Impact of Voltage Sags Caused by
Induction Motor Starting, Proc., International Conference on Electrical
Machines and Systems (ICEMS 2008), Wuhan, China, Oct.17-20, 2008.
[15] L.S. Patil, A.G. Thosar, Application of D-STATCOM to Mitigate
Voltage Sag Due to DOL Starting of Three Phase Induction Motor, in
Proc., Int. Conf. on Control, Automation, Communication and Energy
Conservation (INCACEC-2009), Jun. 4-6, Perundurai, India, 2009.
[16] A. F . Huweg, S. M. Bashi- and N. Mariun, A STATCOM Simulation
Model to Improve Voltage Sag Due to Starting of High Power Induction
Motor, Proc., National Power and Energy Conference (PECon 2004),
Kuala Lumpur, Malaysia, Nov. 29-30, 2004.
[17] F.M. Mahdianpoor, R.A. Hooshmand, M. Ataei, A New Approach to
Multifunctional Dynamic Voltage Restorer Implementation for
Emergency Control in Distribution Systems, IEEE Trans. Power
Delivery, vol. 26, no. 2, pp 882-889, April, 2011.
Bach Quoc Khanh received B.S., M.S. and Ph.D. degrees in Electric Power
Systems from Hanoi University of Science and Technology (HUST), Vietnam
in 1994, 1997 and 2001 respectively. He has been a faculty member of
Electric Power Systems Department, School of Electrical Engineering, HUST
since 2002. He is a visiting scholar at the Center for Advanced Power
Systems, IESES, Florida State University, USA in 2010-2011. His research
interests include power systems planning and analysis, power quality. He has
been a member of IEEE/PES since 2011.
Jianming Lian was born at Fuzhou, Fujian China in 1982. He attended the
University of Science and Technology of China, Hefei, Anhui China in 2000.
After he received the B.S. degree in 2004, he attended Purdue University,
West Lafayette, in USA for his graduate study, where he received the M.S.
degree in 2007 and the Ph.D. degree in 2009. He is now a postdoct research
associate in Center for Advanced Power Systems at Florida State University.
His research interests include power system stability analysis and real-time
control, power quality analysis and improvement, and nonlinear system
analysis and control, especially, adaptive control and decentralized control.
Dr.Bhuvaneswari Ramachandran (M09) is working as Assistant scholar
Scientist at the Center for Advanced Power Systems, Florida State University .
She completed her PhD in 2007 and was sponsored by the Government of
India under the BOYSCAST fellowship scheme to conduct advanced research
at the Center for Advanced Power Systems. Florida State University , USA .
She is interested in Power System operation and control, Electric machine
design and Power market and auction strategies in microgrid economics. She
has published around 50 research papers in referred National and International
journals. She is a member of IEEE.
Sanjeev Srivastava received Ph.D degree in Texas A&M University in 2003.
He currently works as Assistant scholar Scientist at the Center for Advanced
Power Systems, Florida State University. His research interests include fault
detection, diagnosis, prognosis, and reconfiguration of power systems; expert
system and multi-agent applications to power systems and controls. He is a
member of IEEE and IASTED.
David A. Cartes received the Ph.D. degree in engineering science from
Dartmouth College, Hanover, NH. He is an Associate Professor of the
Department of the Mechanical Engineering at Florida State University (FSU),
Tallahassee, FL from January 2001. He directs the Institute for Energy
Systems, Economics and Sustainability (IESES) at FSU. His research interests
include energy management systems, energy policy and energy business
practices, having published over 100 peer reviewed papers. In 1994, he
completed a 20-year U.S. Navy career with experience in operation,
conversion, overhaul, and repair of complex marine propulsion systems.
143
eBook for You
The case of using rectifier is more practical as no external

energy storage is needed. DVR can be combined with
DSTATCOM in Unified Power Quality Conditioner (UPQC)
for other compensations in power networks.
eBook for You
144
eBook for You
145
eBook for You
146
eBook for You
147
An Analytical Approach for Stochastic Estimation of

Voltage Sags in Power Systems
Le Viet Tien and Thavatchai Tayjasanant
Abstract- This paper presents a method to determine the area

of vulnerability (AOV) of a sensitive load and to estimate the
expected sags frequency (ESF) in power systems. Monte Carlo
simulation-based approach taking into account the impact of fault
positions along lines is used. Voltage sags are analyzed by using
short-circuit analysis for balanced and unbalanced faults. In this
paper, a generalized algorithm is developed and tested with IEEE
30-bus test system to verify the technique and illustrate its
applications.
Keywords- Voltage sags, area of vulnerability, expected sag
frequency, Monte Carlo simulation, short-circuit analysis.
I.
INTRODUCTION
Voltage sag is a short-duration reduction in rms voltage

between 0.1 and 0.9 p.u. with duration from 0.5 cycles to 1
min [1]. Voltage sags that affect sensitive loads are usually
caused by faults somewhere in power systems [1-7]. Voltage
magnitude and duration are essential characteristics of voltage
sag. The magnitude of voltage sags depends on various factors
such as the fault location, fault type, pre-fault voltage,
transformer connection, and fault impedance. The sag
magnitude - the remaining voltage in percent or per unit - can
be calculated by short-circuit analysis. The sag duration can
normally be determined from the protection schemes.
Sensitive equipment is widely used in modern industrial
processes such as computers and adjustable speed drives.
Malfunctioning or failure of this equipment can be caused by
voltage sags that lead to work or production losses. As a result,
it is essential to have information of the sensitivity of the
equipment. If the magnitude and duration of voltage sag
exceed the equipment sensitivity threshold, the equipment can
be damaged, and such damage can affect an entire automatic
process, resulting in high cost. Therefore, the characteristic of
equipment must be provided by the manufacturer or obtained
by tests. System performance, expressed by the expected sag
frequency (ESF), can be estimated through the monitoring of
the system supply or by stochastic prediction methods.
The region of the network that includes buses and lines
where the occurrence of the faults will lead to voltage lower
than the sensitivity threshold of a load is called the area of
vulnerability (AOV) [8]. The minimum rms voltage for a
certain duration that a piece of equipment can withstand
without misoperation or failure determines the sensitivity
threshold [6]. To determine the AOV, critical points (or fault
locations) which lead to fault voltages lower than the

sensitivity threshold of a load need to be found.
Stochastic prediction can be used to estimate the ESF for a
sensitive load. The ESF can be obtained by using both AOV
and the failure rate based on the network model and the
historical fault data. Several stochastic prediction methods
have been proposed [2-5]. The method of critical distances and
the method of fault positions are two major stochastic
prediction methods of voltage sags [8]. The method of critical
distances is an uncomplicated technique of voltage sag
prediction based on the voltage divider model. The method of
critical distances is suitable for radial system, but not to
meshed systems. For the method of fault position, many
possible faults are simulated by short-circuit analysis. Fault
positions and the number of simulated faults will determine the
method accuracy. The method of fault positions is not efficient
for large power system because many fault positions have to be
simulated in order to obtain more accurate sag prediction.
Consequently, an efficient method for finding the AOV and
ESF in a large meshed network is needed.
This paper presents a method to determine the AOV of a
sensitive load and to estimate ESF in power systems by using
Monte Carlo simulation-based approach. The estimation of
ESF also takes equipment sensitivity into account. The method
is verified by IEEE 30-bus test system.
II. ALGORITHM
The primary goal of the proposed method is to analyze the
AOV in power systems and determine ESF using Monte Carlo
simulation (MCS). The typical electrical system with -buses
is shown in Fig. 1.
Figure 1. Electrical power system.
148
eBook for You
Department of Electrical Engineering, Faculty of Engineering, Chulalongkorn University, Bangkok, 10330, Thailand
Email: tien_lv@yahoo.com and tayjasanant@yahoo.ca
A. Fault at Buses
When faults take place at a bus of a power system, the sag
magnitude at bus can be calculated by the Z-bus matrix.
Therefore, the voltage at bus when a fault occur at bus is
given by
(1)
where are phase (zero, positive, and negative) sequence

voltage phases at .
are prefault phase sequence voltages.
are the transfer complex impedances between

and in the Z-bus matrix.
are phase sequence fault currents at .
B. Fault on Lines
This section considers a fault on line at a fictitious
position , defined as the ratio of length between bus and
location ( ) to the length of the line ( or
. The voltage at bus , when a fault occurs at position
between and , can be calculated from
(2)
It can be seen that the form of (2) is similar to (1) by

replacing bus with a fictitious bus . The voltage equation in
(2) is in terms of the transfer bus impedance

between
bus m and a fictitious bus .
The value of matrix

is
(3)
is calculated based on the type of faults (see Table I). The
to be used in (6) to (9) can be calculated

value of matrix
from
TABLE I
THE ZERO, POSITIVE AND NEGATIVE SEQUENCE FAULT CURRENT PHASE AT
BUS
Type of
fault
SLGF
LLF
3PF
(7)
DLGF
(6)
(8)
(9)
C. Voltage Threshold
Information Technology Industry Council (ITIC) provides
the ITIC curve which defines the withstand capability of
sensitive loads and devices for protection from power quality
variations. Each type of sensitive equipment has different
curve. The equipment sensitive limit can be evaluated from the
curve.
An example of ITIC curve represents magnitude and
duration of the event is shown in Fig.2 [3]. Points outside the
envelope are presumed to cause damage to the equipment. The
lower envelope is used to determine the acceptable sag
magnitude and duration level. In this paper, ITIC curve is
defined as the voltage threshold of a sensitive load.

(4)

is a diagonal 3x3 matrix of impedance of transmission
line .
The pre-fault voltage at the fault position is
are the prefault voltages at buses and
where
and
, respectively.
(5)
Figure 2. ITIC curve.
149
eBook for You
In this system, bus is selected as the bus where the

sensitive equipment is connected. The voltage sag at bus
caused by four types of faults at bus or along the line
will be discussed next.
E. Calculation of Estimation Sag Frequency

The ESF can be estimated by using failure rate of power
components which are in AOV. It is based on failure rates of
buses ( ) and lines ( ).
The ESF for the three unbalanced faults is calculated from

The ESF for the three-phase balanced fault is calculated

from
Probability
0.80
0.05
0.10
0.05
It requires more historic performance of the system

information for the stochastic variables in MCS process.
Simulation steps can be summarized as follows:
Step 1: Set all initial data
Step 2: Set an observation bus .
Step 3: Set an experience simulation for each line.
Step 4: For each experience, type of fault is a random
number. Remaining voltage is calculated at bus .
Step 5: After the program simulates for all lines, go to the
next step. If not, go back Step 4.
Step 6: Calculate ESF based on the AOV.
III. APPLICATION AND RESULTS
The method is applied to the IEEE 30-bus test system [9]
(see Fig. 3). This system consists of 30 buses, 4 transformers,
37 lines, 6 generators and 21 loads. Sag durations and failure
rate for buses and lines are shown in Table III and IV,
respectively [3].
TABLE III
SAG DURATION FOR VOLTAGE LEVEL
Voltage level (kV)

132
33
Durations (ms)
100
200
TABLE IV
FAILURE RATE FOR BUSES AND LINES
Type of
Fault
SLGF
LLF
DLGF
3PF
Bus failure rate

(Event/Year)
0.064
0.004
0.008
0.003
Line Failure Rate

(Event/100km/Year)
2.000
0.125
0.300
0.100
Fault type
Single Line-to-Ground fault (SLGF)
Line-to-line fault (LLF)
Double Line-to-Ground fault (DLGF)
Three-Phase fault (3PF)
where is the phase

and are the total numbers of buses and lines.
and are the bus and line inside AOV, respectively.
And is the length of line inside AOV.
TABLE II
FAULT TYPE DISTRIBUTION
F. Algorithm and Simulation Parameters

This algorithm is implemented to perform stochastic
analysis of voltage sags for a sensitive bus by Monte Carlo
simulation (MCS). The MCS is a stochastic method for solving
problem by using random variables. The MCS is simulated by
using the fault probability in Table II in order to determine
AOV [5].
The pre-fault voltages are calculated by using power flow

program. Bus 18 is assumed to be a sensitive load bus with the
voltage threshold is 0.8 p.u. and 0.7 p.u. at 33 kV and 132 kV,
respectively [3]. Figure 4 shows the remaining voltage (phase
a) at the sensitive bus 18 when a random fault occurs at bus 6,
its value is 0.49 p.u. The AOV of the sensitive load (bus 18)
calculated by using MCS is showed in Fig. 3 (thick line). The
value of ESF is estimated to be 9.59 sags/year from 10,000
simulations. Figure 5 shows the ESF of this scenario. Values of
ESF for other buses (excluding generator buses) are
summarized in Table V.
150
eBook for You
D. Area of Vulnerability
The area of vulnerability (AOV) of a sensitive load is
essential to estimate the expected number of voltage sags from
the utility systems. The sag magnitudes at a sensitive load bus
can be approximated to quadratic curve for .
Therefore, there are four cases to determine the AOV by using
voltage threshold:
Case 1: , line is outside the AOV.
Case 2: , line is inside the AOV.
Case 3: If the voltage threshold is a value between the
remaining voltage at and , there is one
part of line inside the AOV.
Case 4: If the voltage threshold is greater than the remaining
voltage at and , and lower than the
maximum remaining voltage, there are two parts
of line inside the AOV.
TABLE V
VALUES OF ESF AT VARIOUS BUSES
Bus
3
4
6
7
9
10
12
14
ESF
2.10
2.61
3.80
1.32
5.47
6.15
3.90
7.09
Bus
15
16
17
18
19
20
21
22
ESF
9.27
7.44
7.22
9.59
9.10
8.75
6.75
6.45
Bus
23
24
25
26
27
28
29
30
ESF
9.69
8.49
6.51
7.55
3.73
1.32
4.91
5.49
ESF values at buses near the generator buses are lower than
buses further away. The area which covers buses with ESF
value above 8.3 sags/year is drawn with the dotted line in
Fig.3. Buses in this area need more attention if sensitive loads
are about to connected to.
Figure 3. The area of vulnerability for sensitive
load at bus 18 by using MCS
The Monte Carlo simulation-based method is shown in this

paper to determine the area of vulnerability and to estimate the
expected sag frequency in power systems. The method applies
a random type of balanced and unbalanced faults for each line
section in the system. The method takes the equipment
sensitivity into account by using ITIC curve. Simulation results
from the test system show good results.
REFERENCES
[1]
[2]
Figure 4. The remaining voltage (phase a) at bus 18 when

a random fault occurs at bus 6.
[3]
[4]
[5]
[6]
[7]
[8]
[9]
IEEE Standard 1159-1995: Recommended Practice for Monitoring

Electric Power Quality.
C.H, Park and G. Jang, Fast Method to Determine an Area of
Vulnerability for Stochastic Prediction of Voltage Sags, IEE Proc.
Gener. Transm. Distrib., vol. 152, no. 6, November 2005.
C.H. Park, G. Jang, Stochastic Estimation of Voltage Sags in a Large
Mesh Network, IEEE Transaction on Power Delivery, Vol. 22, No. 3,
July 2007.
E.E. Juarez and A. Hernandez, Analytical Approach for Stochastic
Assessment of Balanced and Unbalanced Voltage Sags in Large
Systems, IEEE Transaction on Power Delivery, Vol. 21, No. 3, July
2006.
G. Olguin, M. Aedo, M. Arias, A. Ortiz, A Monte Carlo Simulation
Approach to the Method of Fault Position for Stochastic Assessment of
Voltage Dips (Sags), Transmission and Distribution Conference and
Exhibition: Asia and Pacific 2005 IEEE/PES, pp. 1-6, 2005.
R.C. Dungan, M.F. McGranaghan, and H.W. Beaty, Electrical Power
Systems Quality, McGraw Hill, 1996.
Y.S. Lim and G. Strbac, Analytical Approach to Probabilistic Prediction
of Voltage Sags on Transmission Networks, IEEE Proc.-Gener.
Transmi. Distrib, Vol. 149, No. 1, January 2002.
M.H.J. Bollen, Understanding Power Quality Problems: Voltage Sags
and Interruption, New York: IEEE Press, 2000.
IEEE
30-bus
Test
Case,
[Online],
Available:
http://www.ee.washington.edu/research/pstca/pf30/pg_tca30bus.html
Figure 5. Estimated sag frequency by using MCS.
151
eBook for You
IV. CONCLUSION
Impacts of Voltage Sags and Protection

Coordination on Sensitive Equipment in
Distribution Systems
Abstract Nowadays, a lot of sensitive electronic equipment is

widely used in modern power systems such as power converters
and adjustable speed drivers. Voltage sag has gained more
interest due to its consequence on the performance of sensitive
equipment (SE). Malfunction or failure of the equipment that
leads to work or production losses can be caused by voltage sags.
As a result, it is essential to have information on equipment sensitivity. If the magnitude and duration of voltage sag exceed the
equipment sensitivity threshold, the equipment can malfunction,
and such a consequence can affect an entire automatic process,
resulting in high economical losses. Reclosers and fuses are the
main overcurrent protection devices in distribution systems. Poor
coordination may adversely impact on the sensitive equipment.
This paper presents a method to analyze the impacts of voltage
sags and protection coordination on sensitive equipment. A
fault position method and mathematical equations for protective
devices are used to set up the protection setting and to calculate
voltage sags. Voltage tolerance thresholds and protective device
characteristics are used to analyze protection and sensitive
equipment coordination. Based on the results, new settings for
protective devices can be adjusted to consider sensitive equipment
in distribution systems. The Roy Billinton Test System (RBTS)
bus 2 is used to analyze the impacts of voltage sags and protection
coordination system on the sensitive equipment in distribution
systems.
Index Terms sensitive equipment, voltage sag, protection
coordination.
I. I NTRODUCTION
Voltage sag is a short-duration reduction in RMS voltage
between 0.1 and 0.9 p.u. with duration from 0.5 cycles to 1
min [1-3]. Voltage sags that affect sensitive load are usually
caused by faults somewhere on transmission and distribution
systems.
Voltage magnitude and duration are essential characteristics of voltage sag. The magnitude of voltage sags mainly
depends on the fault location, fault type and some other
factors such as the pre-fault voltage, transformer connection,
and fault impedance [2-3]. The voltage sag magnitude, which
is expressed in percent or per unit, is calculated by shortcircuit analysis. The voltage sag duration is dened as the
ow duration of the fault current in a network. Therefore, the
duration is determined by the characteristics of the system
protection devices such as overcurrent relays, circuit breakers
and fuses. Generally, the duration is calculated by adding the
* The Department of Electrical Engineering, Chulalongkorn University,
Bangkok 10330, Thailand.
** The University of Tokyo, School of Frontier Science, Department of
Advanced Energy, Hongo 7-3-1, Bunkyo-ku, Tokyo 113-8656, Japan.
intentional time delay considering protection coordination to

the fault clearing time of each device.
Much sensitive equipment are used in modern industrial
such as computers, programmable logic controllers, adjustable
speed drives, and robotics. Many industrial customers using
sensitive equipment suffer from voltage sags. Malfunctioning
or failure of this equipment can be caused by voltage sags that
lead to work or production stops. To analyze these cases, it is
essential to have information of equipment sensitivity. If the
magnitude and duration of voltage sag exceed the threshold of
equipment, the equipment is damaged, and such damage can
effect an entire process at the customer site associate cost.
Therefore, characteristics of the equipment sensitivity must be
provided by the manufacturer or obtained by tests. System
performance, which can be expressed by the expected sag
frequency in the site, can be estimated through the monitoring
of the supply or stochastic prediction methods.
In a distribution system, protection devices are circuit breakers, reclosers and fuses [4]. The coordination of protection
devices are presented as fault clearing process. Poor coordination adversely impacts the overall power quality especially
from the momentary voltage interruption and voltage sags [56]. For example, improper coordination between a midline
recloser and downstream fuses in a fuse-saving scheme can
cause unnecessary momentary interruptions and voltage sags
downstream from the recloser. In practice, the recloser in
fast mode should operate for a temporary fault and give the
fault a chance to clear and operate faster than the fuse. For
permanent fault, a lateral fuse should be opened to clear the
fault. However, the duration of recloser and fuse setting may
be longer than an allowed duration of sensitive equipment.
Therefore, recloser-fuse system could not protect the sensitive
equipment.
This paper presents a novel approach to detect which
protective device will be operated to clear a short-circuit
condition and will protect sensitive equipment. Coordination
of protective devices is used to help incorrect settings of
overcurrent relays and power quality on sensitive equipment.
II. P ROTECTION C OORDINATION
Figure 1 shows traditional fuse-recloser coordination in
distribution systems [6]. In recloser function, there is an
interval between each operation when the recloser remains
open. If the fault is temporary, recloser will clear before the
fuse. If the fault persists after the recloser closes then the fault
152
eBook for You
Le Viet Tien*, Thavatchai Tayjasanant*, Akihiko Yokoyama**, and Bundhit Eua-Arporn*
Fig. 2.
Recloserfuse coordination range
has to be a permanent one and hence fuse must operate to cut

it off.
The general coordination is that the fuse should only operate
for a permanent fault on the load feeder. However if the fault
is a temporary fault or the fault occurs behind the recloser, the
recloser should disconnect the circuit with fast operation and
give the fault a chance to clear. Recloser also provides back
up function when a fuse fails to blow up. In order to have a
correct operation, the fuse must be coordinated with recloser
on the main feeder.
In Fig. 1, the TC curve of the fuse is below the slow
curve of recloser in coordination range. Therefore, for a
permanent fault, fuse will open before recloser will back it
up by operating in slow mode and nally locking out. The
coordination curves of recloser and fuse have to be modied.
The area between If aultmax and If aultmin shows as the
recloser-fuse coordination range. Therefore, as long as the
fault current values for faults on lateral feeder are within
coordination range, the fuse-recloser coordination is accepted.
We can see that the fast characteristic of the recloser lies
below the MM characteristic of fuse between If aultmax and
If aultmin . So, in coordination range the recloser operates in
less time than the time sufcient to damage the fuse.
III. VOLTAGE S AGS ON S ENSITIVE E QUIPMENT
dened as the voltage threshold of a sensitive load. Voltage

threshold can be determined by given sag duration.
B. Area of Vulnerability and Fault Distribution Along Line
The region of the network that includes a part of line or
whole line where the occurrence of the faults will lead to
voltage sag at a sensitive load is called the area of vulnerability
(AOV). The area of vulnerability responding to the sensitive
load is essential to estimate the expected number of voltage
sags from the utility systems as shown in Fig. 3. The fault
distribution along line will cause changes the expected number
of sag frequency at the sensitive load.
Fig. 3.
Area of vulnerability and fault distribution along lines.
A. Voltage Tolerance Curves

Information Technology Industry Council (ITIC) curve is
useful for understanding the limited of voltage sag on sensitive
equipment [1]. This curve denes the withstand capability of
curve. An example of ITIC curve represents magnitude and
duration of the event as shown in Fig.2. Points (outside
envelop) are presumed to cause to damage the equipment.
The lower envelop will use to determine the acceptable sag
magnitude and duration level. In this paper, ITIC curve is
C. Voltage Sag Estimation

Frequently, adverse environmental and weather conditions
can lead to sections of lines exposed to higher fault rates than
others. This can have a considerable impact on the expected
number of sags at the bus of interest. Fault rates along the
line can be distributed uniformly (same faults probability along
whole line) or, on the opposite, some locations can experience
higher fault rates than others. Therefore, taking into account
the random nature of , different fault probability distribution
153
eBook for You
Fig. 1.
Information Technology Industrial Council (ITIC) curve
can be considered (uniform distribution, exponential distribution, etc.).

The line failure rate is normally expressed as the number of
faults per year per unit. However, because of the short length
of line, the line failure rate is calculated for the whole line i
as follows:
li
(1)
LF Ri = Nf ault n
lk
B. Recloser-Fuse Operations
k=1
where Nf ault is number of fault in the test system; li is length

of the line i; n is total number of line in the test system.
After taking into account the line failure rates, the expected
sag frequency due to the faults at the sensitive bus can be
calculated as following equation,
m
equipment tf aultSE . These two parameters will be compared

to fuse and recloser characteristic curves. The point (If ault ,
tf aultSE ) is dened as the protective coordination point
which must be in the coordination range and above fuses
minimum melting time in the case of a fuse operation, or the
recloser fast or slow curve in the case of recloser operation
(see Fig. 1).
When a permanent or temporary fault occurs, a recloser

or fuse should blow to isolate the short-circuit. Times corresponding to If ault on the fuse time-current curves (TCC)
and the recloser TCC are determined as tf aultf useM M and
tf aultf useT C , tf aultrecf ast and tf aultrecslow respectively as shown in Fig. 4.
Lij LF Ri
N SF =
(2)
i=1 j=1
where N SF is the numbers sag frequency; Lij is length of

line segment j of line i inside the area of vulnerability; L is
the number of line segment of line i; m is the total number
of line inside the area of vulnerability.
Assuming that all phases have equal probability of fault
statistics, the expected number of sags frequency for a singlephase can be calculated as following
1
3
N SFF T,P
(3)
eBook for You
N SFSP =
F T =1P =1
where F T is type of fault (i.e. single-line-ground fault; lineto-line fault, double-line-ground fault and three-phase fault);
P is the number of phases.
Fig. 4.
Protection coordination
IV. P ROBLEM D EFINITION AND S OLUTION

A. Case Study Denition
Normally, sensitive equipment is analyzed by using voltage
threshold [7-8]. The voltage threshold is determined by xed
duration of fault current. Therefore, if the duration of fault
current is greater than xed duration, it will impact on sensitive
equipment. This paper will present an approach to determine
vulnerability time tf aultSE based on remaining voltage at
sensitive equipment bus.
Sensitive equipment will be impacted by voltage sags Vf ault
when a fault occurs in system. Then tf aultSE will be
determined by using sensitive equipment characteristics. If a
fault duration is less than tf aultSE , sensitive equipment can
withstand. Therefore, the fault time tf aultSE should be input
to protection coordination for considering impacts of voltage
sags on sensitive equipment.
The problem addressed in this paper can be stated as
follows: Assuming a fault and calculating the fault current
and voltage in the system, it is determined which protective
device interrupts the fault. Moreover, this paper concentrates
on fuse and recloser operations based on sensitive equipment
characteristics. Two parameters are estimated and inputed
to the algorithm as follows: 1) the magnitude of the fault
current If ault , 2) the vulnerability fault duration of sensitive
For considering sensitive equipment, the time tf aultSE

when the fault occurs must be greater than tf aultf useM M
or tf aultf useT C
tf ault-f use-M M tf ault-SE or
tf ault-f use-T C tf ault-SE
(4)
Any fuses whose corresponding tf aultf useM M and

tf aultf useT C satisfy the above requirement indicate an
operation of the fuse.
Moreover, coordination of fuse and recloser is based on the
recloser TCC. The fuse TCC must be located between the
recloser fast and delayed curves.
tf ault-rec-f ast tf ault-f use-M M tf ault-rec-slow
(5)
tf ault-rec-f ast tf ault-f use-T C tf ault-rec-slow
(6)
V. A PPLICATION AND R ESULTS

The proposed method is applied to the RBTS bus 2 (see Fig.
5) [9]. The fault position method [7, 8] is used to calculate the
fault current and voltage. Bus 2 is assumed to be a sensitive
bus. In the simulation, the operation range of the recloser and
fuses are set to be 200 to 6000A. The recloser has fast and slow
154
10
100K Fuse
Recloser S
10
Breaker
Time(s)
10
Recloser F
10
Fuse TC
10
Fuse MM
2
10
10
10
Current(A)
10
10
80K Fuse
10
RBTS bus 2 test system
Breaker
10
Time(s)
Fig. 5.
Recloser S
Recloser F
10
10
Fuse MM
Fuse TC
10
10
10
Current(A)
10
Fig. 6.
Voltage sag ride-through capacity curve from 0 to 100 s
pickup currents of 184A and 452A respectively. For recloser

and breaker, the standard extremely inverse trip characteristic
is used with the parameters A, B and p of TCC are 28.2,
0.127 and 2, respectively. Two types of fuse link are used to
analyze in this paper are 100K fuse link and 80K fuse link.
Let us make the following assumptions: 1) timecurrent
characteristic curves of fuses and reclosers used in the feeder
and 2) sensitive equipment characteristics are available [7] (see
Fig. 6). Each line is divided into ve equal line segments.
500 faults occurring in the test system are assumed. Besides
that, the details of system fault statistics are also assumed as
follows:
Single-Line-Ground fault (SLGF): 85%
Line-to-Line fault (LLF): 8%
Double-Line-Ground fault (DLGF): 5%
Three-Phase fault: 2%
Based on the list of percentages of different fault types [10],
number of SLGF, LLF, DLGF and 3PF are 425, 40, 25 and
10 faults respectively.
A. Simulation of RecloserFuse Coordination with Consideration of Sensitive Equipment
Assume that the distribution network are protected by a
100K fuse link and recloser on the main feeder upstream
from the fuse. Three-phase fault occurs at bus 18 in the test

system. Voltage sag measured at the sensitive equipment is
analyzed by using the fault position method. In this case,
the duration and magnitude of the fault current are estimated
to be tf aultSE = 0.5s and If ault = 2.45kA based on
the approach described before. Besides that, 80K fuses are
also analyzed with the test system. The results indicate that
the 100K and 80K fuses coordinate well with the recloser
considering sensitive equipment, respectively. Figure 7 shows
the protection coordination point is inside the recloser-fuse
coordination range.
B. Sag Frequency without Protection Coordination Consideration
In this case, sag frequency is considered with an assumption
of 0.7 p.u. of voltage threshold of sensitive equipment. Voltage
threshold is used to determine the area of vulnerability based
on the fault position method. Table I shows the number of line
segments inside the area of vulnerability without protection
coordination consideration. As shown in Fig. 8, the area of
vulnerability for sensitive load at bus 2 are illustrated.
The inuence of two kinds of fault distributions along line
are considered in order to calculate expected sags frequency.
Uniform and normal distribution of fault along line are considered. The resulting sag frequency for three phases for sensitive
equipment in this case is obtained by using the numbers of line
segment inside the area of vulnerability in Table II. Therefore,
the expected sag frequencies for a single phase with uniform
and normal distribution fault along lines are 130 and 130.29
sags, respectively.
155
eBook for You
Fig. 7. 100K fuse and 80K fuse coordination with the recloser to clear
threephase fault at bus 18
(b)
(c)
Fig. 8. The areas of vulnerability for sensitive load at bus 2 without protection
coordination. (a), (b), and (c) are areas of vulnerability due to SLGF and 3PF,
LLF, DLGF respectively.
C. Sag Frequency with Protection Coordination Consideration

Protection coordination is considered in order to show
impacts on sensitive equipment in this case. Two kinds of fault
distributions along line are also used to analyze the expected
sag frequency.
Based on the sensitive equipment characteristics and protection coordination, the number of line segments inside the
area of vulnerability are shown in Table III. The areas of
vulnerability for sensitive load at bus 2 are shown in Fig.
9. Table III shows the expected sag frequencies for three
(a)
(b)
eBook for You
(a)
(c)
Fig. 9. The areas of vulnerability for sensitive load at bus 2 with protection
coordination. (a), (b), and (c) are areas of vulnerability due to SLGF and 3PF,
LLF, DLGF respectively.
phases for all type of faults in the test system. The tables
clearly show that the number of sag frequency decreases
with protection coordination consideration. In this case, the
numbers of sag frequency for a single phase with uniform and
normal distribution along lines are 115.85 and 117.73 sags,
respectively.
VI. C ONCLUSION
This paper presents a new method to detect fuse and
recloser operations with consideration of impacts on sensitive
156
TABLE I
T HE NUMBER OF LINE SEGMENTS INSIDE THE AREA OF VULNERABILITY
FOR SENSITIVE EQUIPMENT AT BUS 2 WITHOUT PROTECTION
COORDINATION CONSIDERATION
Type of fault
SLGF (A)
LLF (B,C)
DLGF (B,C)
3PF
Phase A
124
0
0
131
Phase B
0
34
109
131
Phase C
0
145
143
131
TABLE II
T HE NUMBER OF SAG FREQUENCY FOR SENSITIVE EQUIPMENT AT BUS 2
WITHOUT PROTECTION COORDINATION CONSIDERATION
Type
of fault
SLGF
LLF
DLGF
3PF
Uniform distribution
PhaseA PhaseB PhaseC
292.52
0
0
0
7.62
32.26
0
15.16
20.72
7.27
7.27
7.27
Normal distribution
293.42
0
0
0
7.29
32.75
0
14.68
20.46
7.42
7.42
7.42
TABLE III
T HE NUMBER OF LINE SEGMENTS INSIDE THE AREA OF VULNERABILITY
FOR SENSITIVE EQUIPMENT AT BUS 2 WITH PROTECTION COORDINATION
R EFERENCES
[1] IEEE Std 1159-1995: IEEE Recommended Practice for Monitoring Electric Power Quality.
[2] M.H.J. Bollen, Understanding Power Quality Problems: Voltage Sags and
Interruption, New York: IEEE Press (2000)
[3] R.C. Dungan, M.F. McGranaghan, and H.W. Beaty, Electrical Power
Systems Quality, McGraw Hill (1996)
[4] J.M. Gers and E.J. Holmes, Protection of Electricity Distribution Networks. London, U.K.: Inst. Elect. Eng. (2004)
[5] L. Comassetto, D.P. Bernardon, L.N. Canha and A.R. Abaide: Automated
Coordination and Optimization Tool of Protection Devices for Distribution System , POWERENG2007, pp.388-393 (2007)
[6] S. Chaitusaney and A. Yokoyama: Prevention of Reliability Degradation
from Recloser-Fuse Miscoodination Due To Distribution Generation,
IEEE Transactions on Power Delivery, Vol. 23, No. 4, pp.2545-2554
(2008)
[7] C.H. Park and G. Jang: Stochastic Estimation of Voltage Sags in a Large
pp. 1655-1664 (2007)
[8] Le Viet Tien and Thavatchai Tayjasanant: An Analytical Approach for
Stochastic Estimation of Voltage Sags in Power Systems, ECTI-CON
2008, Vol. 2, pp. 1029-1032 (2008)
[9] R.N. Allan, R. Billiton, I. Sjarief, L. Goel, and K.S. So, A Reliability
Test System for Education Purposes Basic Distribution System Data and
Results, IEEE Transactions on Power System, Vol. 6, No. 2, pp. 813-820
(1991)
[10] J.V. Milanovic, Myo Thu Aung, C.P. Gupta, The Inuence of Fault
Distribution on Stochastic Prediction of Voltage Sags. IEEE Transactions
on Power System, Vol. 20, No. 1, pp. 278-285 (2005).
CONSIDERATION
Phase A
117
0
0
123
Phase B
0
20
99
123
Phase C
0
62
143
123
eBook for You
Type of fault
SLGF (A)
LLF (B,C)
DLGF (B,C)
3PF
TABLE IV
T HE NUMBER OF SAG FREQUENCY FOR SENSITIVE EQUIPMENT AT BUS 2
WITH PROTECTION COORDINATION CONSIDERATION
Type
of fault
SLGF
LLF
DLGF
3PF
274.90
0
0
0
4.60
13.86
0
13.74
19.89
6.84
6.84
6.84
Normal distribution
279.87
0
0
0
4.60
14.19
0
13.56
19.95
7.01
7.01
7.01
equipment in a distribution system. The method requires sensitive equipment data, and time-current characteristic curves
of the protective devices. Two types of fuse were used to
evaluate performance of coordination of protective devices
with sensitive equipment.
Further, two kinds of fault distributions along lines are
also simulated to show the numbers and characteristics of sag
frequency. The fault position method is used to determine the
number of line segments inside the area of vulnerability of the
sensitive equipment load. The effectiveness of the technique is
tested for impacts of voltage sags and protection coordination
on sensitive equipment. It has been make clear that sag
frequency taking the protection coordination into account is
lower than sag frequency without the protection coordination
consideration.
157

6B03
Paper
Impacts of Voltage Sags and Protection Coordination on Sensitive

Equipment in Distribution Systems
Student Member
Non-member
Member
Non-member
Nowadays, a lot of sensitive electronic equipment is widely used in modern power systems such as power converters
and adjustable speed drivers. Voltage sags have gained more interest due to their consequences on the performance of
sensitive equipment (SE). Malfunction or failure of the equipment that leads to work or production losses can be caused
by voltage sags. As a result, it is essential to have information on equipment sensitivity. If the magnitude and duration
of voltage sag exceed the equipment sensitivity threshold, the equipment can malfunction, and such a consequence can
aect an entire automatic process, resulting in high economical losses. Reclosers and fuses are the main overcurrent
protection devices in distribution systems. Poor coordination may adversely impact on the sensitive equipment. This
paper presents a method to analyze impacts of voltage sags and protection coordination on sensitive equipment. A
fault position method and mathematical equations describing characteristics of protection devices are used to calculate
voltage sags and the protection settings. Voltage tolerance thresholds and protective device characteristics are used to
analyze protection and sensitive equipment coordination. Based on the results, new settings for protective devices can
be adjusted for sensitive equipment in distribution systems. The Roy Billinton Test System (RBTS) bus 2 is used to
analyze impacts of voltage sags and protection coordination.
Keywords: sensitive equipment, voltage sag, protection coordination.
1.
customers using sensitive equipment suer from voltage

sags. Malfunctioning or failure of this equipment can be
caused by voltage sags that lead to work or production stops.
To analyze these cases, it is essential to have information of
equipment sensitivity. If the magnitude and duration of voltage sag exceed the threshold of equipment, the equipment
is disrupted or malfunctioned, and such a consequence can
cause an entire process at the customer site an associated cost.
Therefore, characteristics of the equipment sensitivity must
be provided by the manufacturer or obtained by tests. System performance, which can be expressed by the expected
sag frequency at the site, can be estimated through the monitoring of the supply or stochastic prediction methods.
In a distribution system, protection devices are circuit
breakers, reclosers and fuses (4) . The coordination of protection devices are presented as fault clearing process. Poor
coordination adversely impacts the overall power quality especially from the momentary voltage interruption and voltage sags (5) (6) . For example, improper coordination between
a midline recloser and downstream fuses in a fuse-saving
scheme can cause unnecessary momentary interruptions and
voltage sags downstream from the recloser. In practice, the
recloser in fast mode should operate for a temporary fault
and give the fault a chance to clear and operate faster than
the fuse. For permanent fault, a lateral fuse should be opened
to clear the fault. However, the duration of recloser and fuse
setting may be longer than an allowed duration of sensitive
equipment. Therefore, recloser-fuse system could not protect the sensitive equipment.
Introduction
Voltage sag is a short-duration reduction in RMS voltage

between 0.1 and 0.9 p.u. with duration from 0.5 cycles to
1 min (1)(3) . Voltage sags that aect sensitive load are usually
caused by faults somewhere on transmission and distribution
systems.
Voltage magnitude and duration are essential characteristics of voltage sag. The magnitude of voltage sags mainly
depends on the fault location, fault type and some other factors such as the pre-fault voltage, transformer connection,
and fault impedance (2) (3) . The voltage sag magnitude, which
is the remaining magnitude expressed in percent or per unit,
is calculated by short-circuit analysis. The voltage sag duration is dened as the ow duration of the fault current in a
network. Therefore, the duration is determined by the characteristics of the system protection devices such as overcurrent
relays, circuit breakers and fuses. Generally, the duration is
calculated by adding the intentional time delay considering
protection coordination to the fault clearing time of each device.
Much sensitive equipment are used in modern industrial such as computers, programmable logic controllers,
adjustable speed drives, and robotics. Many industrial
The Department of Electrical Engineering, Chulalongkorn University

Bangkok, 10330, Thailand
Department of Advanced Energy, School of Frontier Sciences,
The University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo 113-8656
c 2010 The Institute of Electrical Engineers of Japan.

1
158
eBook for You
Le Viet Tien
Thavatchai Tayjasanant
Akihiko Yokoyama
Bundhit Eua-Arporn
This paper presents a novel approach to detect which protective device will operate to clear a short-circuit and protect
sensitive equipment. Proper protection coordination can provide correct relay settings and increase power quality.
Protection Coordination
2.1 Protection Devices and Characteristics

The devices frequently used in a distribution system protection are
overcurrent relays, reclosers, and fuses. Fuses are one of the
most common forms of protection used to deal with excessive
currents. A fuse has two characteristics: the minimum melting (MM) and the total clearing (TC). The minimum melting
is the relationship between the magnitude of the current and
the time large enough to cause the fuse to melt. The total
clearing time is the total time required from the beginning of
the fuse element to melt to the nal circuit interruption; i.e.
TC time is the minimum melting time plus the subsequent
arcing time. S and F are two characteristics of the recloser
which are slow and fast operation modes, respectively (4) .
Fuses contain inverse-time overcurrent characteristics. The
straight line I 2 t log-log plot is usually expressed for the minimum melting and total clearing time for fuses (6) . From the
fuse characteristic on the log-log curve, it is better to approximate by the second order polynomial function. The general
equation describing the fuse characteristics curve can be expressed as the following equation:
Fig. 1. A typical distribution feeder
log(t) = a log(I) + b (1)
Fig. 2. RecloserFuse coordination range
where t and I are the associated time and current; a and b can
be known from the curve tting.
Circuit breakers (CB) and reclosers usually located at the
beginning and middle of main feeders. The general characteristics of these devices can be shown as the following equation:
A
t(I) = p
+ B (2)
M 1
where t is the operating time of inverse-time overcurrent device; I is fault current seen by the device; M is the ratio of
I/I pickup (I pickup is relay current set point); A, B, p are constants for selected curve characteristics.
For the protection settings, this paper uses the mathematical equations for overcurrent relays and the straight line I 2 t
log-log curve to formulate protection coordination (5) (6) . The
protection settings are done in the initial or existing condition. CBs and reclosers characteristics are assumed to be
equipped with the extremely inverse characteristic of overcurrent relays. A typical radial distribution system is shown
with protection devices in Fig. 1. In a typical distribution system, all demand loads are supplied from the main feeder.
2.2 Recloser-Fuse Coordination
Figure 2 shows traditional recloser-fuse coordination in distribution systems (6) .
A recloser has two characteristic curves which are fast and
slow (buck up) operation. In recloser function, there is an
interval between each operation when the recloser remains
open. If the fault is temporary, recloser will clear before the
fuse. If the fault persists after the recloser closes then the
fault has to be a permanent one and hence fuse must operate
to cut it o.
The general coordination is that the fuse should only operate for a permanent fault on the load feeder. However if
the fault is a temporary fault or the fault occurs behind the

recloser (downstream fault), the recloser should disconnect
the circuit with fast operation and give the fault a chance to
clear. Recloser also provides back up function when a fuse
fails to blow up. In order to have a correct operation, the fuse
must be coordinated with recloser on the main feeder. On the
other hand, if the fault occurs before the recloser (upstream
fault), the fuse should isolate and coordinate with a sensitive
equipment.
In Fig. 2, the TC curve of the fuse is below the slow curve
of recloser in coordination range. Therefore, for a permanent fault, fuse will open before recloser will back it up by
operating in slow mode and nally locking out. The coordination curves of recloser and fuse have to be modied. The
area between I f aultmax and I f aultmin shows as the recloserfuse coordination range. Therefore, as long as the fault current values for faults on lateral feeder are within coordination
range, the recloser-fuse coordination is accepted. We can see
that the fast characteristic of the recloser lies below the MM
characteristic of fuse between I f aultmax and I f aultmin . So, in
coordination range the recloser operates in less time than the
time sucient to damage the fuse.
3.
3.1 Voltage Tolerance Curves

Information Technology Industry Council curve (7) and SEMI F47 (8) provide
guidelines and dene the withstand capability of sensitive loads and devices for protection from power quality
variations.
Information Technology Industry Council (ITIC) curve
is useful for understanding the limited of voltage sag on
2
Voltage Sags on Sensitive Equipment
IEEJ Trans. PE, Vol.130, No.6, 2010

159
eBook for You
2.
Impacts of Voltage Sags and Protection Coordination
bus of interest. Fault rates along the line can be distributed

uniformly (same faults probability along whole line) or, on
the opposite, some locations can experience higher fault rates
than others. Therefore, taking into account the random nature
of dierent fault probability distribution can be considered.
Fault distribution on lines are likely to be normal distribution
in practice (7) (10) .
The line failure rate is normally expressed as the number
of faults per year per unit length. However, because of the
short length of line in a distribution system, the line failure
rate (LFR) is calculated for the whole line i as follows
LFRi = N f ault
li
(3)
lk
Fig. 3. Information Technology Industrial Council

(ITIC) and SEMI curves
k=1
where is uniform, normal or exponential distributions of

the fault on the line i; N f ault is number of fault in the test system; li is length of the line i; n is total number of line in the
test system.
After taking into account the line failure rates, the expected
sag frequency due to the faults at the sensitive bus can be calculated from
m
NS F =
Li j LFRi (4)
where NS F is the numbers of sag frequency; Li j is length of

line segment j of line i inside the area of vulnerability; L is
the number of line segment of line i; m is the total number of
line inside the area of vulnerability.
Assuming that all phases have equal probability of fault
statistics, the expected number of sags frequency for one
phase can be calculated as following
Fig. 4. Area of vulnerability and fault distribution along

lines
sensitive equipment (1) (7) . This curve denes the withstand capability of sensitive loads and devices for protection from
power quality variations. Each type of sensitive equipment
has dierent curve. An example of ITIC curve represents
magnitude and duration of the event as shown in Fig. 3.
SEMI F47 is a standard that denes the voltage threshold that
semiconductor processing must be controlled through without interruption during conditions identied in the area above
the standard line (Fig. 3). Points above the threshold are presumed to cause damage to the equipment. The lower region
will use to determine the acceptable sag magnitude and duration level.
In this paper, ITIC curve is dened as the voltage threshold
of a sensitive load. Voltage threshold can be determined by a
given sag duration.
3.2 Area of Vulnerability and Fault Distribution
Along Line
The region of the network that includes a
part of line or whole line where the occurrence of the faults
will lead to voltage sag at a sensitive load is called the area
of vulnerability (AOV). The area of vulnerability responding to the sensitive load is essential to estimate the expected
number of voltage sags from the utility systems as shown in
Fig. 4. The fault distribution along line will cause changes in
the expected number of sag frequency at the sensitive load.
3.3 Voltage Sag Estimation
Frequently, adverse
environmental and weather conditions can lead to sections of
lines exposed to higher fault rates than others. This can have
a considerable impact on the expected number of sags at the
B130 6 2010
NS FS P =
1
NS F FT,P (5)
3 FT =1 P=1
where FT is type of fault (i.e. single-line-ground; line-toline, double-line-ground and three-phase faults); P is the
number of phases.
4.
Problem Denition and Solution
4.1 Problem Denition

Normally, voltage tolerance
of sensitive equipment is analyzed by using voltage threshold (9)(11) . The voltage threshold is determined by this xed
duration of fault current. Therefore, if the duration of fault
current is greater than this xed duration, it will impact on
sensitive equipment. This paper will present an approach
to determine vulnerability time t f aultS E based on remaining
voltage at sensitive equipment bus.
Sensitive equipment will be impacted by voltage sags
V f ault when a fault occurs in system. Then t f aultS E will be
determined by using sensitive equipment characteristics. If a
fault duration is less than t f aultS E , sensitive equipment can
withstand. Therefore, the fault time t f aultS E should be input
to protection coordination for considering impacts of voltage
sags on sensitive equipment. The problem addressed in this
paper can be stated as follows: Assuming a fault and calculating the fault current and voltage in the system, it is determined which protective device interrupts the fault. Moreover,
3
160
eBook for You
i=1 j=1
Fig. 5. Protection coordination

Fig. 6. RBTS bus 2 test system
this paper concentrates on fuse and recloser operations based

on sensitive equipment characteristics. Two parameters are
estimated and inputed to the algorithm as follows: (1) the
magnitude of the fault current I f ault , (2) the fault duration
of sensitive equipment t f aultS E . These two parameters will
be compared to fuse and recloser characteristic curves. The
point (I f ault , t f aultS E ) is dened as the protective coordination point which must be in the coordination range and above
fuses minimum melting time in the case of a fuse operation,
or the recloser fast or slow curve in the case of recloser operation (see Fig. 5).
4.2 Recloser-Fuse Operations
When a permanent or
temporary fault occurs, a recloser or fuse should blow to isolate the short-circuit. Times corresponding to I f ault on the
fuse time-current curves (TCC) and the recloser TCC are determined as t f ault f useMM and t f ault f useTC , t f aultrec f ast and
t f aultrecslow respectively as shown in Fig. 5.
For considering sensitive equipment, the time t f aultS E
when the fault occurs must be greater than t f ault f useMM or
t f ault f useTC . Therefore,
Table 1. Coecient a and b for fuse setting
fuse link (80 K and 100 K) are used as examples in this paper.
Parameters a and b of the straight line I 2 t of (1) are shown in
Table 1.
Let us make the following assumptions: (1) timecurrent
and (2) sensitive equipment characteristics are available (9)
(see Fig. 7). Each line is divided into ve equal line segments.
500 faults occurring in the test system are assumed. Besides
that, the details of system fault statistics are also assumed as
follows (11)
Single-Line-Ground fault (SLGF): 85%
Line-to-Line fault (LLF): 8%
Double-Line-Ground fault (DLGF): 5%
Three-Phase fault: 2%
Based on the list of percentages of dierent fault types,
number of SLGF, LLF, DLGF and 3PF are 425, 40, 25 and
10 faults, respectively.
5.1 Simulation of RecloserFuse Coordination with
Consideration of Sensitive Equipment
Assume that the
distribution network is protected by a 100 K fuse link and a
recloser on the main feeder upstream from the fuse. Threephase fault occurs at bus 18 in the test system. Voltage sag
Any fuses whose corresponding t f ault f useMM and

t f ault f useTC satisfy the above requirement indicate an operation of the fuse.
recloser TCC. The fuse TCC must be located between the recloser fast and slow curves.
t f ault-rec- f ast t f ault- f use- MM t f ault-rec- slow (7)
t f ault-rec- f ast t f ault- f use-TC t f ault-rec-slow (8)
Application and Results
The proposed method is applied to the RBTS bus 2 (see

Fig. 6) (12) . The fault position method (9)(11) is used to calculate
the fault current and voltage. Bus 2 is assumed to be a sensitive bus. In the simulation, the operation range of the recloser
and fuses are set to be 200 to 6000 A. The recloser has fast
and slow pickup currents of 184 A and 452 A respectively.
For recloser and breaker, the standard extremely inverse trip
characteristic is used with the parameters A, B and p of TCC
to be 28.2, 0.127 and 2, respectively. However, two types of
4
Total clearing
aTC
bTC
-2.5132 -2.9199
-2.6951 8.4269
Fig. 7. Voltage sag ride-through capacity curve from 0

to 100 s
t f ault- f use- MM t f ault-S E or

t f ault- f use-TC t f ault-S E (6)
5.
Minimum melting
a MM
b MM
-2.9199 9.4467
-2.8378 8.5925
eBook for You
Type of
fuse
100 K
80 K

161
eBook for You
Fig. 8. 100 K fuse and 80 K fuse coordination with the recloser to clear threephase fault at bus 18
Fig. 9. Protection coordination with 100 K fuse and 80 K fuse when fault occurs along line from bus 5 to bus 18
Table 2. The number of line segments inside the area

of vulnerability for sensitive equipment at bus 2 without
protection coordination consideration
measured at the sensitive equipment is analyzed by using the

fault position method. In this case, the duration and magnitude of the fault current are estimated to be t f aultS E = 0.5s
and I f ault =2.45 kA based on the approach described before.
Besides that, 80 K fuses are also analyzed with the test system. The results indicate that the 100 K and 80 K fuses coordinate well with the recloser considering sensitive equipment.
Figure 8 shows the protection coordination point is inside the
recloser-fuse coordination range.
5.2 Simulation With Fault Along Line
Based on
the fault location method, a three-phase fault occurs along
line from bus 5 to bus 18 on the feeder 1 as shown in Fig. 9.
It is shown that 100 K fuse can coordinate with recloser to
protect sensitive equipment if the fault occurs. The estimated
fault current owing in the fuse were 2.33 kA, 2.62 kA and
2.45 kA corresponding with a fault at the beginning, mid-line
and end of the line, respectively. The duration is 0.5 s based
on the sensitive equipment characteristic. Figure 9 shows that
if a fault occurs at mid-line, 80 K fuse cannot protect the sensitive equipment, but 100 K fuse can protect with the protection coordination.
B130 6 2010
Type of fault
SLGF (A)
LLF (B,C)
DLGF (B,C)
3PF
Phase A
124
0
0
131
Phase B
0
34
109
131
Phase C
0
145
143
131
5.3 Sag Frequency without Protection Coordination

Consideration
In this case, sag frequency is considered
with an assumption of 0.7 p.u. of voltage threshold of sensitive equipment. Voltage threshold is used to determine
the area of vulnerability based on the fault position method.
Coordination of protection devices and sensitive equipment
characteristics curve are not considered in this case. Table 2
shows the number of line segments inside the area of vulnerability without protection coordination consideration. As
shown in Fig. 10, the area of vulnerability for sensitive load
at bus 2 are illustrated. The inuence of two kinds of fault
distributions along line are considered in order to calculate
expected sags frequency. Uniform and normal distribution of
5
162
Table 3. The number of sag frequency for sensitive equipment

at bus 2 without protection coordination consideration
fault along line are considered. The resulting sag frequency

for three phases for sensitive equipment in this case is obtained by using the numbers of line segment inside the area
of vulnerability in Table 3. Therefore, the expected sag frequencies for one phase with uniform and normal distribution
fault along lines are 130 and 130.29 sags, respectively using
Eq. (5).
Type
of fault
SLGF
LLF
DLGF
3PF
Phase A Phase B Phase C
292.52
0
0
0
7.62
32.26
0
15.16
20.72
7.27
7.27
7.27
Normal distribution
293.42
0
0
0
7.29
32.75
0
14.68
20.46
7.42
7.42
7.42
(a)
eBook for You
(a)
(b)
(b)
(c)
(c)
Fig. 10. The areas of vulnerability for sensitive load at

bus 2 without protection coordination: (a) SLGF and 3PF,
(b) LLF, and (c) DLGF
Fig. 11. The areas of vulnerability for sensitive load at

bus 2 with protection coordination: (a) SLGF and 3PF,
(b) LLF, and (c) DLGF
6

163

Table 4. The number of line segments inside the area of
vulnerability for sensitive equipment at bus 2 with protection coordination consideration
Phase A
117
0
0
123
Phase B
0
20
99
123
(1)
Phase C
0
62
143
123
(2)
(3)
(4)
(5)
Table 5. The number of sag frequency for sensitive

equipment at bus 2 with protection coordination consideration
Type
of fault
SLGF
LLF
DLGF
3PF
274.90
0
0
0
4.60
13.86
0
13.74
19.89
6.84
6.84
6.84
(6)
Normal distribution
279.87
0
0
0
4.60
14.19
0
13.56
19.95
7.01
7.01
7.01
(7)
(8)
(9)
( 10 )
5.4 Sag Frequency with Protection Coordination

Consideration
Protection coordination is considered in
order to show impacts on sensitive equipment in this case.
Sensitive equipment characteristics curve (ITIC or SEMI
47 F) are used to analyze fault duration with protection
coordination. It means that the proposed method considers
not only the magnitude of voltage, but also its duration. Two
kinds of fault distributions along line are also used to analyze
the expected sag frequency.
Based on the sensitive equipment characteristics and protection coordination, the number of line segments inside the
area of vulnerability are shown in Table 4. The areas of
vulnerability for sensitive load at bus 2 are shown in Fig. 11.
Table 5 shows the expected sag frequencies for three phases
for all type of faults in the test system. The tables clearly
show that the number of sag frequency decreases with protection coordination consideration. In this case, the numbers
of sag frequency for one phase with uniform and normal distribution along lines are 115.85 and 117.73 sags, respectively.
6.
( 11 )
( 12 )
Appendix
When a fault concurs along line k j at a fault position
p which is dened as the ratio of length between bus k and
fault location (Lkp ) to the length of the line k j (Lk j ) or
p = Lkp /Lk j . The voltage at bus m, when a fault occurs at
fault position p between k and j, can be calculated from
012 012
Vm012 = Vm012p f Zmp
I p (A1)
012
is the transfer bus impedance beIn equation (A1), Zmp
tween bus m and a fault position p on the line k j and can
be shown as following
012
012
[Zmp
] = (1 p)[Zmk
] + p[Zm012j ] (A2)
Conclusion
The pre-fault voltage at the fault position p is also calculated

as following
This paper presents a new method to detect fuse and recloser operations with consideration of impacts on sensitive
equipment in a distribution system. The method requires sensitive equipment data, and time-current characteristic curves
of the protective devices. Two types of fuse were used to
evaluate performance of coordination of protective devices
with sensitive equipment.
Further, two kinds of fault distributions along lines are also
simulated to show the numbers and characteristics of sag
frequency. The fault position method is used to determine
the number of line segments inside the area of vulnerability of the sensitive equipment load. The eectiveness of the
technique is tested for impacts of voltage sags and protection coordination on sensitive equipment. It is clear that sag
frequency taking the protection coordination into account is
lower than sag frequency without the protection coordination
consideration.
(Manuscript received Sep. 14, 2009,
revised Jan. 13, 2010)
B130 6 2010
IEEE Std 1159-1995: IEEE Recommended Practice for Monitoring Electric

Power Quality
M.H.J. Bollen: Understanding Power Quality Problems: Voltage Sags and
Interruption, New York: IEEE Press (2000)
R.C. Dungan, M.F. McGranaghan, and H.W. Beaty: Electrical Power Systems Quality, McGraw Hill (1996)
J.M. Gers and E.J. Holmes: Protection of Electricity Distribution Networks,
Inst. Elect. Eng., London, U.K. (2004)
L. Comassetto, D.P. Bernardon, L.N. Canha, and A.R. Abaide: Automated
Coordination and Optimization Tool of Protection Devices for Distribution
System, POWERENG2007, pp.388393 (2007)
S. Chaitusaney and A. Yokoyama: Prevention of Reliability Degradation
from Recloser-Fuse Miscoodination Due To Distribution Generation, IEEE
Trans. Power Delivery, Vol.23, No.4, pp.25452554 (2008)
The Information Technology Industry Council (ITI)[online]: www.itic.org
Semi Voltage Sags Tolerance [online]: www.semi.org
C.H. Park and G. Jang: Stochastic Estimation of Voltage Sags in a Large
Mesh Network, IEEE Trans. Power Delivery, Vol.22, No.3, pp.16551664
(2007)
L.V. Tien and T. Tayjasanant: An Analytical Approach for Stochastic
Estimation of Voltage Sags in Power Systems, ECTI-CON 2008, Vol.2,
pp.10291032 (2008)
J.V. Milanovic, M.T. Aung, and C.P. Gupta: The Inuence of Fault Distribution on Stochastic Prediction of Voltage Sags, IEEE Trans. Power Syst.,
Vol.20, No.1, pp.278285 (2005)
R.N. Allan, R. Billiton, I. Sjarief, L. Goel, and K.S. So: A Reliability Test
System for Education Purposes Basic Distribution System Data and Results,
IEEE Trans. Power Syst., Vol.6, No.2, pp.813820 (1991)
V ppre f = Vkpre f + (Vipre f Vkpre f )p (A3)

When a short-circuit fault occurs along line k j, the fault
current can be calculated and shown as following.
1. Single-Line-Ground Fault (SLGF)
The sequence fault currents for a phase-a-line-to-ground
fault are
I 0p = I 1p = I 2p =
V p1p f
Z 0pp + Z 1pp + Z 2pp + 3z f
(A4)
2. Line-to-Line Fault (LLF)

For this type fault, the sequence fault currents when the
fault between phases b and c are
I 1p = I 2p =
V p1p f
(A5)
Z 1pp + Z 2pp + 3z f
and I 0p = 0.
7
164
eBook for You
Type of fault
SLGF (A)
LLF (B,C)
DLGF (B,C)
3PF
References
3. Double-Line-Ground Fault (DLGF)

The sequence fault currents for phase-b-to-phase-c-toground fault are
I 1p =
V p1p f
Z 1pp +
I 2p = I 1p
I 0p = I 1p
Z 2pp Z 0pp +3z f

Z 0pp +Z 2pp +3z f
(A6)
Z 0pp + 3z f
Z 0pp + Z 2pp + 3z f
Z 2pp
Z 0pp
Z 2pp
Thavatchai Tayjasanant (Non-member) received the B.Eng. degree

from Chulalongkorn University, Bangkok, Thailand,
in 1994, the M.Sc. degree from the University of
Manchester, Manchester, U.K. (formerly UMIST) in
1996, and the Ph.D. degree from the University of Alberta, Canada, in 2005, all are in electrical engineering. Currently, he is an assistant professor at Chulalongkorn University where he has been since 1998.
His research interest is power quality.
+ 3z f
(A7)
Akihiko Yokoyama (Member) was born in Osaka, Japan, on October
9, 1956. He received the B.S., M.S. and D.Eng. degrees from The University of Tokyo, Tokyo, Japan, in
1979, 1981, and 1984, respectively. He has been with
the Department of Electrical Engineering, The University of Tokyo, since 1984 and is currently a Professor in charge of Power System Engineering. He is
a member of IEEJ and CIGRE.
(A8)
4. Three-Phase Fault (3PF)

For a balanced fault, the sequence fault currents are
I 1p =
V p1p f
(A9)
Z 1pp + z f
and I 0p = I 2p = 0.
The value of matrix [Z 012
pp ] to be used in (A4) to (A9) can
be calculated from
Bundhit Eua-Arporn (Non-member) received both Bachelor and

Master degrees from Elec. Eng. Dept., Faculty of Engineering, Chulalongkorn University. In 1992, he received Ph.D. degree from Imperial College of Science
Technology and Medicine, London, UK. He is currently a Professor at the Elec. Eng. Dept., and deputy
director for Research aair at the Energy Research
Institute, Chulalongkorn University.
2 012
2 012
[Z 012
pp ] = (1 p) [Zkk ] + p [Z j j ] +
eBook for You
012
+ 2p(1 p)[Zk012
j ] + p(1 p)[zk j ]
(A10)
where [z012
k j ] is a diagonal 3x3 matrix of impedance of distribution line k j.
Le Viet Tien (Student Member) was born in Vietnam, in 1980. He
received the B.Eng. in electrical engineering from
Hanoi University of Technology, Hanoi, Vietnam in
2003. He is currently pursuing the Ph.D. degree under the AUN/SEED-Net Scholarship from JICA at the
Power System Research Lab., Department of Electrical Engineering, Faculty of Engineering, Chulalongkorn University. He is now staying at the University of Tokyo for short-study program of JICA scholarship. He is a student member of IEEJ.
8

165

on Sensitive Equipment in Distribution System
Abstract--Nowadays, more sensitive electronic equipment is
widely used in modern power systems such as power converters
and adjustable speed drivers. Power quality has been a greater
interest in voltage sags due to their impacts on the performance
of sensitive equipment (SE). Malfunction or failure of the equipment that leads to work or production losses can be caused by
voltage sags. As a result, it is essential to have information on
equipment sensitivity. If the magnitude and duration of voltage
sag exceed the equipment sensitivity threshold, the equipment
can be malfunctioned, and such a consequence can affect an entire automatic process, resulting in high economical losses. Reclosers and fuses are the main overcurrent protection devices in
distribution systems. Poor coordination could adversely impact
on the sensitive equipment. This paper presents a method to analyze the impacts of voltage sags and protection coordination on
sensitive equipment. A fault position method and mathematical
equations for protective devices are used to set up the protection
setting and to calculate voltage sags. Voltage tolerance thresholds
and protective device characteristics are used to analyze protection and sensitive equipment coordination. Based on the results,
new settings for protective devices are done to consider sensitive
equipment in distribution systems. The Roy Billinton Test System (RBTS) bus 2 is used to analyze the impact of voltage sags
and protection coordination system on the sensitive equipment in
distribution systems.
Index Terms-- sensitive equipment, voltage sag, protection coordination.
I. INTRODUCTION
OLTAGE sag is a short-duration reduction in rms voltage

between 0.1 and 0.9 p.u. with duration from 0.5 cycles to
1 min [1-3]. Voltage sags that affect sensitive load are
usually caused by faults somewhere on transmission and
distribution systems.
Voltage magnitude and duration are essential characteristics of voltage sag. The magnitude of voltage sags mainly depends on the fault location and fault type and some other factors such as the pre-fault voltage, transformer connection, and
fault impedance [2-3]. The voltage sag magnitude, which is
expressed in percent or per unit, is calculated by short-circuit
analysis. The voltage sag duration is defined as the flow duraLe Viet Tien is with the Department of Electrical Engineering,
Chulalonkorn
University,
Bangkok,
10330,
Thailand
(e-mail:
tien_lv@yahoo.com).
Akihiko Yokoyama is with The University of Tokyo, School of Frontier
Sciences, Department of Advanced Energy, Hongo 7-3-1, Bunkyo-ku, Tokyo
113-8656, Japan (e-mail: yokoyama@syl.t.u-tokyo.ac.jp).
Thavatchai Tayjasanant is with the Department of Electrical Engineering,
Chulalongkorn University, Bangkok, 10330, Thailand (e-mail:
tayjasanant@yahoo.ca).
Bundhit Eua-Arporn is with the Department of Electrical Engineering,
Chulalongkorn
University,
Bangkok,
10330,
Thailand
(e-mail:
Bundhit.E@chula.ac.th).
tion of the fault duration of the fault current in a network.

Therefore, the duration is determined by the characteristics of
the system protection devices such as overcurrent relays, circuit breakers and fuses. Generally, the duration is calculated
by adding the intentional time delay considering protection
coordination to the fault clearing time of each device.
Much sensitive equipment are used in modern industrial
such as computers, programmable logic controllers, adjustable
speed drives, and robotics. Many industrial customers using
sensitive equipment suffer from voltage sags. Malfunctioning
or failure of this equipment can be caused by voltage sags that
lead to work or production stops. To analyze these cases, it is
essential to have information of the sensitivity of the equipment. If the magnitude and duration of voltage sag exceed the
equipment threshold sensitivity, the equipment is damaged,
and such damage can effect an entire process at the customer
site associate cost. Therefore, characteristics of the sensitivity
equipment must be provided by the manufacturer or obtained
by tests. System performance, which can be expressed by the
expected sag frequency in the site, can be estimated through
the monitoring of the supply or stochastic prediction methods.
In distribution system, protection devices are circuit breakers, reclosers and fuses [4]. The coordination of protection
devices are presented as fault clearing process. Poor coordination adversely impacts the overall power quality especially
from the momentary voltage interruption and voltage sags [56]. For example, improper coordination between a midline
recloser and downstream fuses in a fuse-saving scheme can
cause unnecessary momentary interruptions and voltage sags
downstream from the recloser. In practice, the recloser in fast
mode should operate for a temporary fault and give the fault a
chance to clear and operate faster than the fuse. For permanent
fault, a lateral fuse should be opened to clear the fault. However, duration of recloser and fuse setting may be longer than
an allowed duration of sensitive equipment. Therefore, recloser-fuse system could not protect the sensitive equipment.
This paper presents a novel approach to detect which protective device will be operated to clear a short-circuit condition and will protect sensitive equipment. Coordination of
protective devices is used to help incorrect settings of overcurrent relays and power quality on sensitive equipment.
II. PROTECTION COORDINATION
A. Protection Devices and Characteristics
The devices most used for distribution system protection
are overcurrent relays, reclosers, and fuses. Fuses are one of
the most common forms of protection used to deal with excessive currents. A fuse has two characteristics: Minimum Melting (MM) and Total Clearing (TC). S and F are two characteristics of the recloser which are slow and fast operation modes,
respectively [4].
166
eBook for You
Le Viet Tien, Akihiko Yokoyama, Thavatchai Tayjasanant and Bundhit Eua-Arporn
A, B, p : constant for selected curve characteristics.

For the protection settings, this paper uses the mathematical equations for overcurrent relays and the straight line I 2 t
log-log curve to formulate protection coordination [5-6]. The
protection settings are done in the initial or existing condition.
The CBs and reclosers characteristics are assumed to be
equipped with the extremely inverse characteristic of overcurrent relays.
Breaker
Recloser
Fuse
Fuse
Main feeder
Fuse
Fuse
Lateral feeders
Fig. 1. A typical distribution feeder.
A typical radial distribution system is shown with protection devices in Fig. 1. In typical distribution system, all demand loads are supplied from a bulk.
B. Protection Coordination
Fig. 2 shows traditional fuse-recloser coordination in distribution systems [6]. In recloser function, there is an interval
between each operation when the recloser remains open. If the
fault is temporary, recloser will clear before the fuse. If the
fault persists after the recloser closes then the fault has to be a
permanent one and hence fuse must operate to cut it off.
The general coordination is that the fuse should only operate for a permanent fault on the load feeder. However if the
fault is a temporary fault or the fault occurs behind the recloser, the recloser should disconnect the circuit with fast operation and give the fault a chance to clear. Recloser also provides back up function when a fuse fails to blow up. In order
to have a correct operation, the fuse must be coordinated with
upstream recloser on the main feeder.
In Fig. 2, the TC curve of the fuse is below slow curve of
recloser in coordination range. Therefore, for a permanent
fault, fuse will open before recloser will back it up by operating in slow mode and finally locking out. The coordination
curves of recloser and fuse have to be modified. The fuserecloser coordination range graph between I fault max
and I fault min . Therefore, as long as the fault current values for
faults on lateral feeder are within coordination range, the fuserecloser coordination is accepted. We can see that the fast
characteristic of the recloser lies below the MM characteristic
of fuse between I fault max and I fault min . So, in coordination
range the recloser operates in less time than the time sufficient
to damage the fuse.
Fig. 2. Recloser-fuse coordination range.
III. VOLTAGE SAGS ON SENSITIVE EQUIPMENT

A. Voltage Sags
Information Technology Industry Council (ITIC) curve is
useful for understanding the limited voltage sag of sensitive
equipment [1]. This curve defines the withstand capability of
curve. An example of ITIC curve represents magnitude and
duration of the event as shown in Fig.3. Points (outside envelop) are presumed to cause to damage the equipment.
Fig. 3. ITIC curve
The lower envelop will use to determine the acceptable sag

magnitude and duration level. In this paper, ITIC curve is defined as the voltage threshold of a sensitive load. Voltage threshold can be determined by given sag duration.
167
eBook for You
Fuses contain inverse-time overcurrent characteristics. The

straight line I 2 t log-log plot is usually expressed for the
minimum melting and total clearing times for fuses [6]. From
the fuse characteristic on the log-log curve, it is better to approximate by the second order polynomial function. The general equation describing the fuse characteristics curve can be
expressed as the following equation:
log(t ) a u log( I ) b
(1)
where:
t and I are the associated time and current,
a and b can be know from the curve fitting.
Circuit breakers (CB) and reclosers usually located at beginning and middle of the main feeders. The general characteristics of those devices can be shown as the following equation:
A
t(I )
B
(2)
p
M 1
where
t :operating time of inverse-time overcurrent device,
I : fault current seen by the device,
M : ratio of I / I pickup ( I pickup is relay current set point),
B. Fault Position Method

Fault position method is used to calculate voltage sags in
this paper. Bus m is selected as the bus where the sensitive
equipment is connected. The voltage sag at bus m caused by
three-phase fault along the line k j will be shown as following [8-9].
A fault at a fictitious position p on the line k j , defined
as the ratio of length between bus k and fault location to the
length of the line k j or p Lkp / Lkj . The voltage at bus
of recloser operation (see Fig. 2).
m , when a fault occurs at the position p between k and j ,

can be calculated from
012
Vm012 Vm012 [ Z mp
]I 012
(3)
p
The voltage equation in (3) is in terms of the transfer bus
012
impedance Z mp
between bus m and the fictitious bus p .
Fig. 4. Protective coordination.
012
[ Z mp
]
012
012
(1 p)[ Z mk
] p[ Z mj
]
(4)
I 012
is calculated based on the type of faults. The value of
p
matrix Z 012
pp can be calculated from
Z 012
pp
(1 p) 2 [ Z kk012 ] p 2 [ Z 012
jj ]
2 p (1 p )[ Z kj012 ] p(1 p)[ z kj012 ]
012
kj
(5)
is a diagonal 3x3 matrix of impedance of transmission
line k j .
The pre-fault voltage at the fault position p is
A. Sensitive Equipment
Normally, sensitive equipment is analyzed by using voltage threshold [7-8]. The voltage threshold is determined by
fixed duration of fault current. Therefore, if the duration of
fault current is greater than fixed duration, it will impacts on
sensitive equipment. This paper will present an approach to
determine vulnerability time t fault SE based on remaining voltage at sensitive equipment bus.
Sensitive equipment will be impacted by voltage sags
V fault when a fault occurs in system. Then t fault SE will be de-
(6)
termined by using sensitive equipment characteristics. If a

fault duration is less than t fault SE , sensitive equipment can
where Vkpref and V jpref are the prefault voltages at buses k
withstand. Therefore, the fault time t fault SE should be input to
and j , respectively.
When a three-phase fault occurs, the voltage sag at bus m
can be expressed as
1
Z mp
(7)
Vmfault Vmpref 1 V ppref
Z pp
protection coordination for considering impacts of voltage

sags on sensitive equipment.
V ppref
Vkpref (V jpref Vkpref ) p
IV. PROBLEM DEFINITION AND SOLUTION

The problem addressed in this paper can be stated as follows: Assuming a three-phase fault and calculating the fault
current and voltage in test system, it is determined which protective device interrupts the fault. Moreover, this paper concentrates on fuse and recloser operations based on sensitive
equipment characteristics.
We have assumed that a three-phase fault occurs; the fault
current and voltage are input to the algorithm. Some qualities
are estimated as follows: 1) the magnitude of the fault current
I fault , 2) the duration in which the fault current I fault flows in
the protective device, t fault SE . These two parameters will be
compared to fuse and recloser characteristic curves.
I fault , t fault SE is defined as the protective device operation
point which must be between the coordination range and
above fuses minimum melting time in the case of a fuse operation, or on the reclosers fast or delayed curve in the case
B. Recloser-Fuse Operations
When a permanent or temporary fault occur, a recloser or
fuse should blow to isolate the short-circuit condition. Times
corresponding to I fault on the fuses time-current curves
(TCC) and the reclosers TCC curves are determined as
t fault fuse MM and t fault fuse TC , t fault rec fast and t fault rec slow respectively as shown in Fig. 4.
For considering sensitive equipment, the time t fault SE when
the fault occurs must be greater than t fault fuse MM or
t fault fuse TC ,
t fault fuse MM d t fault SE or t fault fuse TC d t fault SE
(8)
Any fuse in which their corresponding t fault fuse MM and

t fault fuse TC satisfy the above requirement indicate an operation
of the fuse.
reclosers TCC curves. The fuses TCC curves must be located between the reclosers fast and delayed curves.
(9)
t fault rec fast d t fault fuse MM d t fault rec slow
t fault rec fast d t fault fuse TC d t fault rec slow
(10)
168
eBook for You
012
The value of matrix Z mp
is
V. APPLICATION AND RESULTS

The proposed method is applied to the RBTS bus 2 (see
Fig. 5) [10]. Three-phase fault occurs at bus 18 in the test system. The fault position method is used to calculate the fault
current and voltage. Bus 2 is assumed to be a sensitive bus. In
the simulation, the operation range of the recloser and fuses
are set to be 200 to 6000A. The recloser has fast and slow
pickup currents of 184A and 452A respectively. The lateral is
protected by a 100 K-type fuse link.
Let us make the following assumptions: 1) time-current
and 2) sensitive equipment characteristics are available (see in
Fig. 6). Simulation steps can be summarized as follows:
A. Case 1: Simulation of Recloser-Fuse Coordination with

Consideration of Sensitive Equipment.
Assume the utilities are protected by a 100 K-type fuse link
and recloser on the main feeder upstream from the fuse. Voltage sag measured at the sensitive equipment was analyzed by
using the fault position method. The duration and magnitude
of the fault current are estimated to be t fault SE 0 .5 s and
2 .45 kA based on the approach described before. The
I fault
results indicate that the fuse coordinations well with the recloser considering sensitive equipment are 100 K-type and 80
K-type, respectively. Figures 7 and 8 show the operation point
on the recloser and fuse coordination curve.
3
10
Step 2)
Step 3)
Step 4)
Step 5)
Input system configuration, load information, and

sensitive equipment data.
Set protection coordination for circuit breaker,
reclosers and fuses.
Select the bus that SE will be installed on.
Calculate faults in the test system and then determine the vulnerability time which will impact to
SE.
Evaluate and determine the operation point for
recloser and fuse.
10
10
Time(s)
Step 1)
10
-1
10
10
10
Current(A)
10
Fig. 7. Fuse 100 K-type coordination with the recloser to clear three-phase
fault at bus 18.
3
10
10
Time(s)
10
10
Fig. 5. RBTS bus 2 test system.
-1
10
100
-2
10
90
80
10
10
Current(A)
10
Fig. 8. Fuse 80 K-type coordination with the recloser to clear three-phase fault
at bus 18.
70
60
50
Fig. 9 shows the coordination of protective devices (recloser and fuses: 80 K-type and 100 K-type) in four main
feeders with the sensitive equipment located at bus 2 in feeder
1. In Figure 9b, 80 K-type fuse may not protect the sensitive
equipment when fault occurs at bus 23 in the feeder 2.
40
30
20
10
0.2
0.5
1.0
10
100
Duration of voltage sag in seconds
Fig. 6. Voltage sag ride-through capacity curve from 0 to 100s.

169
eBook for You
-2
10
eBook for You
(a) Fault at bus 18 in the main feeder 1.
(b) Fault at bus 23 in the main feeder 2.
(c) Fault at bus 38 in the main feeder 3.
(d) Fault at bus 56 in the main feeder 4.

Fig. 9. Protection coordination with 100 K-type and 80 K-type fuse when fault occurs at bus 18, 23, 38 and 56 in the main feeder 1, 2, 3 and 4, respectively.
170
Fig. 10. Protection coordination with 100 K-type and 80 K-type fuse when fault occurs along line 11 (from bus 5 to bus 18).
[6] S. Chaitusaney and A. Yokoyama, Prevention of Reliability Degradation from Recloser-Fuse Miscoodination Due To Distribution GeneraB. Case 2: Simulation with Fault Along Lines
tion, IEEE Transactions on Power Delivery, Vol. 23, No. 4, October
Based on the fault location method, a three-phase fault will
2008.
be occur along line 11 (from bus 5 to bus 18) on the feeder 1 [7] C.H. Park, G. Jang, Stochastic Estimation of Voltage Sags in a Large
as shown in Fig. 10. It is shown that 100 K-type fuse can coJuly 2007.
ordinate with recloser to protect sensitive equipment if the [8] Le Viet Tien and Thavatchai Tayjasanant, An Analytical Approach for
fault occurs. The estimated fault current flowing in the fuse
Stochastic Estimation of Voltage Sags in Power Systems, ECTI-CON
2008, Vol. 2, May 14-17, 2008.
were 2.33 kA, 2.62KA and 2.45kA corresponding with fault
location 0.1, 0.5 and 1, respectively. Vulnerable duration is [9] Y.S. Lim and G. Strbac, Analytical Approach to Probabilistic
Prediction of Voltage Sags on Transmission Networks, IEEE Proc.0.5s based on the sensitive equipment characteristic. Figure 10
Gener. Transmi. Distrib, Vol. 149, No. 1, January 2002.
shows that if a fault occur at midline, 80 K-type fuse cannot [10] R.N. Allan, R. Billiton, I. Sjarief, L. Goel, and K.S. So, A Reliability
protect the sensitive equipment, but 100 K-type fuse can proTest System for Education Purposes Basic Distribution System Data
and Results, IEEE Transactions on Power System, Vol. 6, No. 2, May
tect with the protection coordination.
1991.
VI. CONCLUSION
VIII. BIOGRAPHIES
The proposed method is to detect fuse and recloser opera- Le Viet Tien was born in Vietnam, in 1980. He received the B. Eng. in elections with consideration of impacts on sensitive equipment in trical engineering from Hanoi University of Technology, Hanoi, Vietnam in
2003. He is currently pursuing the Ph.D. degree at the Power System Research
distribution system. The method requires sensitive equipment Lab., Department of Electrical Engineering, Faculty of Engineering, Chudata, and time-current characteristic curves of the protective lalongkorn University.
devices. The fault position method is used as a fast method for

fault calculation. The effectiveness of the techniques is tested
for the impacts of voltage sags and protection coordination on
sensitive equipment. Results for the test system show good
results.
VII. REFERENCES
[1]
[2]
[3]
[4]
[5]
IEEE Std 1159-1995: IEEE Recommended Practice for Monitoring

Electric Power Quality.
M.H.J. Bollen, Understanding Power Quality Problems: Voltage Sags
and Interruption, New York: IEEE Press, 2000.
R.C. Dungan, M.F. McGranaghan, and H.W. Beaty, Electrical Power
Systems Quality, McGraw Hill, 1996.
J.M. Gers and E.J. Holmes, Protection of Electricity Distribution Networks. London, U.K.: Inst. Elect. Eng., 2004.
L. Comassetto, D.P. Bernardon, L.N. Canha and A.R. Abaide, Automated Coordination and Optimization Tool of Protection Devices for
Distribution System, PowerEng., April 12-14, 2007, Portugal.
Akihiko Yokoyama (M78) was born in Osaka, Japan, on October 9, 1956.

He received the B.S., M.S. and D.Eng. degrees from The University of Tokyo,
Tokyo, Japan, in 1979, 1981, and 1984, respectively. He has been with the
Department of Electrical Engineering, The University of Tokyo, since 1984
and is currently a Professor in charge of Power System Engineering. He is a
member of IEEJ and CIGRE.
Thavatchai Tayjasanant (M06) received the B.Eng. degree from Chulalongkorn University, Bangkok, Thailand, in 1994, the M.Sc. degree from the
University of Manchester, Manchester, U.K. (formerly UMIST) in 1996, and
the Ph.D. degree from the University of Alberta, Edmonton, AB, Canada, in
2005, all are in electrical engineering. Currently, he is a Lecturer at Chulalongkorn University where he has been since 1998. His research interest is
in the area of power quality.
Bundhit Eua-Arporn (M96) received both Bachelor and Master degrees
from Elec. Eng. Dept., Faculty of Engineering, Chulalongkorn University. In
1992, he received Ph.D. degree from Imperial College of Science Technology
and Medicine, London, UK. He is currently a professor at the Elec. Eng.
Dept., and deputy director for Research affair at the Energy Research Institute,
Chulalongkorn University.
171
eBook for You
Journal of International Council on Electrical Engineering Vol. 2, No. 1, pp. 45~52, 2012
45
Subsequence Action to Eliminate Blackout after Detecting Islanding using

Solid State Transfer Switch Implemented in PSCAD/EMTDC
Nguyen Duc Tuyen, Goro Fujita*, Toshihisa Funabashi** and Masakatsu Nomura**
Abstract One operation circumstance in Dispersed Generation (DG) concepts, an island, is formed
when one or more DGs and an aggregate of local loads are disconnected from the main grid and then,
remain operational as an islanded entity. This condition of islanding is conventionally undesirable.
Consequently, the customers will suffer the blackout. Recently, this blackout phenomenon has been
eliminating by facilitating islanded operation. This paper presents a passive method in detecting islanding
for DGs by measuring the negative-sequence impedance differences between islanded mode and gridconnected mode at the line frequency. Subsequently, the islanded entity is connected to a backup system
by Solid State Transfer Switch (SSTS) in order to smoothly and quickly keep electricity available to
customers.
1. Introduction
The currently generalized industrial standard requires
disconnecting all the DG sources from the islanded entity as
fast as possible. For example, under the IEEE 1547
standard, that period is within two seconds [1]. For a
smooth transition from a grid-connected mode to an
islanded mode of operation, fast islanding detection (ID) is
essential to allow DGs to adapt their control strategy to the
new operational one.
The ID methods can be separated into two basic
categories: communication and local one. Local detection
methods can further subdivided into active and passive
detection schemes [2]. In 2008, C.Wrinch developed a
simple control scheme which is easy to accomplish in
software as well as hardware [3]. It presents passive ID by
using
negative-sequence
impedance
measurement.
Information about the step change of actual grid negativesequence impedance could be used for detecting an
islanding situation. That method helps detect an islanding in
as fast as a few cycles for the specific settings described
afterwards in the simulation in Section 4. Thus, it facilitates
an islanding's subsequence actions which fall into two
options. The first one is to connect immediately to a backup
system. The second option is to regulate control topology of
DG which is still in operation inside the islanded system.
Corresponding Author: Power System Lab., Shibaura Institute of
*
**
Technology, Japan (m609504@shibaura-it.ac.jp)

Power System Lab., Shibaura Institute of Technology, Japan
Meidensha Corporation, Japan
Received: December 21, 2011; Accepted: December 28, 2011
If subsequent to islanding, the voltage and frequency

remain under the limits (which means the DGs can stiffly
track islanded power demand with enough active and
reactive power), then a proper control topology can
maintain the autonomous operation. Nowadays, research
approaches are focusing on making the control topology of
DG in an islanded entity become perfect. This so-called
autonomous operation needs further refinement due to its
complexity.
Conversely, in this research, a simpler concept is
presented which coincides with the first mentioned option
after islanding. Distribution systems usually work in radial
operation that contains a high number of backup grid
terminals. Solid State Transfer Switch (SSTS) conventionally transfers the preferred utility to a backup grid in the
case of fault occurring at the preferred utility [4]. Instead of
mechanical circuit breakers, SSTS can promote the power
quality by its superior advantages. Therefore, the proposed
solution ignores any blackout to consumers.
2. Islanding Detection Based on Negative-Sequence

Impedance Measurement
2.1 Islanding detection methodology
The ID method used in this paper uses the theoretically
accurate concept of impedance measurement when the
negative-sequence components of voltage and current are
available. This idea is shown in Fig. 1. The premise is that
the large impedance difference between an island and a
172
eBook for You
Keywords: Islanding detection, DG, SSTS, Blackout, Autonomous operation.
Subsequence Action to Eliminate Blackout after Detecting Islanding using Solid State Transfer Switch Implemented in ~
grid-connected condition supposedly has a very low Non

Detection Zone for radial systems with stiff network
connections, that is . Such a change
in the impedance would allow easier threshold settings for
ID.
Z DG
Z utility
Z load
E utility
V DG
Fig. 1. Principle of using impedance ID method.

The unbalanced system might be transformed into the
symmetrical system described in Fig. 2. The system
equivalent voltage source, remains as the only source
and the current from the and sequences are
indeed supplied by the . The is used because it

is considered that the normal resources generate only
positive-sequence components. Thus, subscript "+"
represents positive-sequence components and subscript "-"
stands for negative-sequence components. From the Fig. 2,
a current source that is produced at the load for negativesequence and zero-sequence components is a result of
being divided up at node . is the point where three
symmetrical currents join together and represents
mutual components. From the current and voltage
derivation in symmetrical components, with this
configuration, the current flows out of the unbalanced load
and goes into the system. That allows the use of Ohm's law
for negative-sequence components measurement to get the
negative-sequence impedance of system, as stated in
Eq.(1).
I0
Z 0sys
Z sys
Z load
V
sys
+
Z +sys
A typical radial system is depicted in Fig. 3. In Fig. 3, the

utility source is connected to a local system on the left side,
and the DG source is connected to the distribution system
on the right side. The ID relay is used by the DG to detect
the opening of circuit breaker C, causing an islanding
phenomenon. The current transformer and voltage
transformer at the breaker B can be used as the point of
measurement's equipment for the ID relay. Also, if the other
current transformer and voltage transformer at feeders 1, 2
and 3 are close to the DG source, which means they have
connections to the ID relay of DG, they could be the point
of measurement for ID.
ZUtility
I V
m
I+
Z mload
I0 + I + I+
3. Transfer Switch
Z +load
3.1 Circuit breaker as transfer switch
V+
Fig. 2. Symmetrical-component current flow in an unbalancedsystem expanded circuit.

2.2 Implementation strategy for impedance islanding

detection
Fig. 3. Set up of ID method.
Z 0load
V0
islanding sensor relay by measuring the impedance away

from an unbalanced node (so-called the point of
measurement) and then comparing it with the already
known threshold before and after the islanding state occurs.
The measured negative-sequence current, and negativesequence voltage, waveforms consist of the real and
imaginary components. The step magnitude change of
negative-sequence impedance is adequate for islanding
alarm signals.
(1)
This method of impedance measurement lets the

negative-sequence components be used as a real-time anti-
The conventional way of using circuit breakers (CB)

creates an undesirable problem. CB reliability depends on
mechanical structure in taking action that is sometimes at
low levels. In reality, while using CB, there is a closing
time between the time when the contacts start to close and
when they finally make contact. Somewhere in between, an
arc may strike across the contacts as they are closing.
This is known as "pre-strike" as seen in Fig. 4.
173
eBook for You
46
Nguyen Duc Tuyen, Goro Fujita, Toshihisa Funabashi and Masakatsu Nomura
47
time delay, TD. That is a time delay (in ms) before the poles
actually commence to close after the Timed Breaker Logic
component (or other signal) has initiated closing. By
randomly changing TD with the Multiple Run component
over a period of one cycle (0 - 16.667 ms) during successive runs, line energizing studies can be accomplished.
3.2 Solid state transfer switch
In the open position, the withstand voltage of the CB

will be a per unit value of rated voltage. The time varying
value of voltage across the open contacts is depicted as an
absolute function of the AC voltage across the contacts. As
the contacts close, the withstand voltage reduces as the
separation distance between the contacts reduces. When the
voltage acrossing the contacts exceeds the reducing
withstand voltage of the insulating medium between them,
pre-strike occurs.
A user-defined component has been developed to model
the pre-strike effect of CB closing as shown in Fig. 5. The
trip signal from the ID relay is fed in as seen at the upper
part. A part of a program code for this component written
by Fortran language is shown at the lower part. SWST59
called in the third row is the Fortran file written by
Manitoba HVDC Research Centre. [5].
Fig. 6. Schematic representation of the SSTS as a link-device.
User built component

0:Close
1:Open
Bka
BKA_backup
Bk
Trip
TripBK_backup
SWIT...
BREAKER D...
16.6667
eBook for You
Fig. 4. Pre-strike effect in closing CBs [5].
It is well-known that the SSTS can be used very

effectively to protect sensitive loads against voltage sags,
swells and other electrical disturbances, within a time scale
of milliseconds which is sufficiently fast to preserve the
operation of even the most sensitive customer equipment
loads. The basic configuration of this device consists of two
three-phase AC solid-state switches, one for the main feeder
and one for the backup feeder. These switches have an
arrangement of back-to-back connected thyristors, as
illustrated in the schematic diagram of Fig. 6.
Bkb
Statistical
Breaker
Close Bkc
TD
BKB_backup
BKC_backup
msec
BREAKER DELAY
10
While an islanding scene is detected by the ID relay, the

control system swaps the firing signals to the thyristors in
the link switches, i.e., the backup feeder is activated. Since
the static transfer switch will transfer upon an under-voltage
condition within one quarter of a 60 Hz cycle (4.167 ms),
the customer's load will not be interrupted and the sensitive
load will not suffer a low power quality.
4. Islanding Detection and SSTS Combination

Control System
The ID system is illustrated in Fig. 7. Voltages (

)
Fig. 5. Pre-strike model in PSCAD.

To ensure there is a random initiation of the CB closing,
the breaker pre-strike component has an input-designated
) at the point of measurement (PoM) are

and currents (
extracted to make the fundamental components by applying
Fast Fourier Transform (FFT). Then, these components
(

) are transformed into positive-, negative-
174
and zero-sequence quantities by using symmetrical

transformation ( to ). However, only the

negative-sequence components (
)) are utilized
for the ID purpose. In order to remove high-frequency parts,
the obtained is being processed by the low pass
filter. Since then, the instant negative-sequence impedance
( ) is compared to the tolerance level ( ) to
determine whether the islanding has been occurring or not.
If the measured is higher than the tolerance level,
this first islanding signal (1st IS) is established. In order to
make the ID scheme more reliable, after a few-cycle delay
(60Hz system) of this 1st IS (4-cycle delay in this study),
the negative-sequence impedance of the system, (
)) is checked again to build the second islanding
signal (2nd IS). Then, if it's value is once more higher than
the tolerance level, the islanding takes place and the trip
signal is generated.
After receiving the ID trip signal, the relay starts
monitoring the voltage at the connection point (CP) where
the backup system is going to connect with. This CP is
located inside the islanded entity. At that same instance,
voltages at the backup system's terminal is observed.
Whenever the difference voltages of each phase reach zero
at the first cycle, the thyristors control gate of respective
phases is then fired to turn ON. This synchronization
technique makes the transient current through thyristors
reasonably small. Owing to only the first zero cross
instance of difference voltages should make the firing
signals of the thyristors gate, the control SSTS system has
been made to distinguish the first zero cross with the others.
That means, after each thyristor has been turned ON to
transfer the load to a healthy feeder, the firing signals of
gate are then locked.
va-Backup Terminal
va-CP
u (PoM
a ,b ,c )
i (PoM
a ,b , c )
vc-Backup Terminal
u1(PoM
a ,b , c )
+0
vb-Backup Terminal
vb-CP
vc-CP
i1(PoM
a ,b , c )
5. Wind Power and Typical IEEE 13-bus

The entire validated model is the wind generators connecting to IEEE 13-bus. The authors used this perspective
to test the considered subsequence action after islanding.
5.1 Wind power generator
For modeling the induction generator, models of the
wind speed, and the wind turbine, model blocks provided
by the EMTDC program are hence used. The wind turbine
model consists of three components: wind model, wind
turbine and induction machine. A wind model in PSCAD
library is a four-component model, which can be described
in [5] and [6]. The wind turbine model is described in [5]
and [7]. This test system simulates two similar wind
generators which generate about 1MVA power demand for
each one. Two wind generators of a wind farm are
connected together through the transmission lines before
coupling to IEEE 13-bus grid.
eBook for You
48
Z line
Z line
Fig. 8. Wind generator connects to IEEE 13-bus scheme.

5.2 IEEE 13-bus includes DG
SSTSa
Utility Bus
SSTSb
SSTSc
650
646
645
632
633
CP
u (PoM
)
i (PoM
)
634
PoM
tolerence
611
684
Z sys (t )
671
692
675
Z sys (t + delay )
652
Fig. 7. Islanding detection and SSTS control system.
680
Wind Generators
Fig. 9. IEEE 13-bus.
175
6. Evaluation of Results from Simulation

6.1 Islanding detection
This section shows the ID ability of the proposed system
and how fast the control system makes a backup feeder
connection to the islanded IEEE 13-bus by using the SSTS
against a mechanical CB. The islanding intentionally occurs
at 3[s]-instance by opening utility CB at node 650.
At first, in order to evaluate the ID, the function of SSTS
is locked. Fig. 10 shows the step shift system negativesequence impedance subsequent to an islanding. From
Fig.10, under grid-connected mode, the system impedance
is about . However, when an islanding occurs, this
parameter increases to . Thus, the tolerance level of
IEEE 13-bus system is pre-determined at a . Another
result derived from the lower part of this figure illustrates
the role of the impedance filter. As shown in the Fig. 10
which is subsequent to the islanding, the transient makes
the impedance value oscillate and then violate the
given tolerance level during the short beginning period
which is right after islanding. Nevertheless, while using the
filter, the impedance value, keeps stably
increasing during the transient. The point is that the
detection time (i.e. the duration from islanding occurrence
until impedance values stably get higher the tolerance level)
takes longer. Despite this minor disadvantage, the detection
time keeps within a few cycles under using filters that still
facilitates the subsequence action after islanding is detected.
Im p e d a n ce M a g n itu d e (O h m )
b) System negative-sequence impedance

1.20
Z2
Z2filter
0.0
2.925
2.950
2.975
3.000
3.025
3.050
3.075
3.100
3.125
Fig. 10. System negative-sequence impedance.
6.2 SSTS control

Fig. 11 shows all control signals of ID and SSTS connection. At the top part, islanding signal generates at 3[s]instance. The first round ID signal appears at 3.0145[s]instance. The delay time to activate the second round
measurement is about 4 cycles and this time is controllable
in PSCAD simulation. Thus, after about 0.0667[s] (at
3.08125 [s]-instance), this second ID signal appears. Both
signals have shown that islanding phenomenon had
definitely happened. Then almost immediately right after
the second ID signal, the trip signal appears to initiate the
SSTS system (at 3.08125[s]-instance).
Detection and SSTS Control Signals
Islanding
firstround
Detected
Trip
SSTSa
SSTSb
SSTSc
0.0
1.0
2.0
3.0
4.0
5.0
Fig. 11. All control signals.

Fig. 12 shows the impact of the ID trip signal on SSTS
closing signals and then to the voltages at connection points
and backup terminals. After the trip signals appear, the
SSTSs in each phase then waits until the instance when the
voltage at backup terminal is exactly equal to the voltage at
connection point of islanded entity in order to send the
activating signals to firing gates (at 3.08275[s]-instance for
phase A, 3.0885[s]-instance for phase B, and 3.08525[s]instance for phase C). The thyristor control gate of
respective phases is then fired to turn ON. Consequently,
voltages at connection points coincide with those at backup
terminals. It is also pointed out from this figure that after a
few cycle, the trip signal reset to beginning status (all three
phases A, B and C at 3.1195[s]-instance) because of
negative-sequence impedance of system reduces. After that,
this reset ID relay waits for other detections. Monitoring
voltage action therefore, at the same time, is deactivated. In
short, the proposal system spends 0.08125[s] (4.875 cycles)
to detect islanding, 0.0885[s] (5.31 cycles) to close islanded
entity to backup system, and 0.1195[s] (7.17 cycles) to reset
the ID relay. If the delay time of the second round ID
checking is set at two or three cycles instead of four cycles,
those periods are even further shortened. Even though
results are just from simulation, the time periods are fast
enough to facilitate the islanding's subsequence action.
Besides eliminating the blackout phenomenon, they
hopefully mitigate low power quality made by islanding to
sensitive loads of the islanded entity.
176
eBook for You
The proposed ID technique was tested on IEEE 13-bus

system as shown in Fig. 9 [10]. This system involves multiphase distribution line segments including three-phase-,
two-phase-, and single-phase-line. Modification of this grid
was conducted in order to be suitable for this test. One DG
is inserted at node 680. The point of measurement is settled
at node 634. In this simulation, the connection point, CP,
locates at node 632.
49
a)Voltage at Connection Point at Backup Feeder Side

Vbackup-b
5.0
Voltage(kV)
The voltages of connection point at backup feeder side

and those at load side are shown by two diagrams
respectively in Fig. 13. In the upper part of Fig. 13, at the
instance when SSTSs link to the islanded entity with the
backup source, the rms voltages at the backup system side
drop to smaller values. This is due to the increasing of
entire loads which are connected to the backup system. As
shown in the lower portion of Fig. 13, the rms voltages at
load side recover to the previous values. Then again, in
islanded entity, the wind generators still keep running.
However, its power quality expression is quite poor in not
only voltage magnitude but also in the voltage balance
problem.
Vbackup-c
Vbackup-a
-5.0
2.90
3.00
3.10
b) Voltages ofConnection Point at Load Side

Vload-a
4.0
Voltage(kV)
50
Vload-b
Vload-c
-4.0
2.90
3.00
3.10
Fig. 13. Voltages at connection point.

b) System negative-sequence impedance
LoadVa
BackupVa
-4 0
Firing Gate Signal and Voltages at Connection Points of Phase B

Magnitude
4.0
Trip
-5 0
3.070
SSTSb
3.080
3.090
LoadVb
3.100
2.950
BackupVb
3.110
3.120
3.130
Z2
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
Z2filter
2.975
3.000
SSTSc
LoadVc
3.090
3.100
3.110
3.120
3.100
3.125
3.150
3.175
Currents through SSTSs
BackupVc
1.25
3.080
3.075
3.130
Ia
Ib
Ic
i [kA]
Magnitude
-4 0
3.070
Trip
3.050
Fig. 14. Negative-sequence impedance with SSTS operation.
Firing Gate Signal and Voltages at Connection Points of Phase C

4.0
3.025
-1.00
Fig. 12. SSTS closing based on zero difference between voltages

at the backup terminal and connection point.
1.0
2.0
3.0
4.0
5.0
Fig. 15. Currents through SSTSs.

Voltages of the Connection Point at Load Side
Voltage(kV)
5.0
Vload-a
time
Vload-b
Vload-c
-5.0
0.0
Impedance Magnitude (Ohm)
The negative-sequence impedance of simulation network

during the utility-disconnected state remains as high as
, and after the SSTS takes action, this impedance
value returns to the previous small one with stability. This
phenomenon can be seen in Fig. 14. The transient currents
flowing through SSTSs of three phases based on the closing
instance have been kept within acceptable level. Those
currents are illustrated in Fig. 15.
In contrast with the successful reclosing of the islanded
entity to the backup system, while using the mechanical CB
without the proposed control system as in Fig. 7, the
reclosing effort failed to take effect as seen in Fig. 16.
Those reclosers unsuccessfully switch the backup system to
the islanded entity over and over again. Both the voltages at
the connection point and system negative-impedance under
the impact of reclosers oscillates time after time.
The powers exchange among the wind farm, the backup
source, and the utility during islanding and SSTS
connection period is shown in Fig. 17. Adead of an
islanding, both the wind farm in the role as a DG and utility
0.0
1.0
2.0
3.0
4.0
5.0
4.0
5.0
System negative-sequence Impedance

Z2
Z2filter
4.0
2.0
0.0
0.0
1.0
2.0
3.0
Fig. 16. Reclosing fails while using a mechanical CB.

grid supply IEEE 13-bus loads with adequate power
demands. It is clear that the wind generators do not
themselves supply the whole power consuming of IEEE 13bus and the shortage power are compensated by the utility
grid through node 650. Via the deficiency of active and
reactive power control tools of simulated wind generators,
the islanded entity cannot operate the IEEE 13-bus by itself
and the SSTS device has been turned on to avoid a blackout.
177
eBook for You
Magnitude
4.0
SSTSa
Impedance Magnitude (Ohm)
Firing Gate Signal and Voltages at Connection Points of Phase A

Trip
During islanding, the generated wind power reduces to

approximately zero until the SSTS connects islanded entity
to the backup system. After connecting to backup feeder,
instead of the previous utility source, the shortage power is
supplied by the backup source appropriately.
a) Power from Wind Farm
Qwind
0.0
1.0
2.0
3.0
4.0
5.0
3.0
4.0
5.0
b) Power from Utility

PUti
References
-0.50
0.0
1.0
2.0
c) Power from Backup Feeder

Pbackup
Qbackup
Power
4.00
-0.50
0.0
1.0
2.0
3.0
4.0
5.0
Fig. 17. Power exchange among three systems.

a) Power from the DGs
Power (MW, MVAr)
1.20
Qwind
Pwind
-0.60
0.0
1.0
2.0
3.0
4.0
5.0
3.0
4.0
5.0
Power (MW, MVAr)
b) Power from Utility

3.50
PUti
QUti
-0.50
0.0
This paper introduces a new method to remove blackout

phenomenon in the case of islanding occurrence. The SSTS,
a fast switch, instead of a mechanical CB is used for the
first time for subsequence action after islanding. Before
appearing new superior approaches on DG control topology
to deal with islanded entity, the backup system in the
distribution network can be utilized to keep power available
to customers. Although the results are from a simulation,
the ideas manifest themselves as a potential application in
practice. More experiments should be conducted to
determine how well the proposed system can perform.
QUti
Pow er
3.50
7. Conclusion
1.0
2.0
Fig. 18. Wind generators stop supplying power.

Also from the simulation results, it should be understood
that if there is neither backup power grid nor changing
control topology of DG, i.e., the wind turbines in this case
study, the islanded entity will going to be involved in a
blackout rapidly. This phenomenon is shown in Fig. 18. The
wind generator's output powers gradually reduce to zero.
The consumers in this case cannot avoid a blackout.
[1] IEEE-SA-Standards-Board, IEEE STANDARD for

interconnecting distributed resources with electric
power systems, IEEEStd 1547, p.1-27, 2003.
[2] Xu Wilsun, K.Mauch, S.Martel, ``An assessment of
distributed generation islanding detection methods and
issues for Canada", Natural Resources Canada, 2004.
[3] Wrinch, J.Marti, M.Nagpal, Negative sequence
impedance based islanding detection for distributed
generation (NSIID), Electric Power Conference
EPEC 2008. IEEE Canada, p.1-6, 2008.
[4] P. M. Anderson and Anjan Bose, Modeling and
Analysis of Custom Power Systems by PSCAD/
EMTDC, IEEE Trans. Power Delivery, Vol.17, No.1,
p.266-272, 2002.
[5] Manitoba HVDC Research Center, PSCAD
Application Guide 2008, Alstom T\&D Protection &
Control Ltd, retrieved from https://pscad.com/library,
2008.
[6] P. M. Anderson and Anjan Bose, Stability Simulation
of Wind Turbine Systems, IEEE Trans. Power
Apparatus and systems, Vol. PAS-102, No.12, p.37913795, 1983.
[7] Reynolds, Michael G., Stability of Wind Turbine
Generators to Wind Gusts, Technical Report, Purdue
University Report TR-EE 79-20.
[8] A. Murdoch, R. S. Barton, J. R. Winkelman, S.H. Javid,
Control Design and Performance Analysis of a 6 MW
wind Turbine Generator, IEEE Trans. on Power
Apparatus and Systems, Vol. PAS-102, No. 5, p.13401347, 1983.
[9] Manitoba HVDC Research Center, PSCAD/EMTDC
Power System Simulation Software User's Manual,
Alstom T&D Protection & Control Ltd, Version 4, 2005.
178
eBook for You
Pow er
Pwind
2.00
1.50
1.00
0.50
0.00
-0.50
51
52
[10] http://ewh.ieee.org/soc/pes/dsacom/testfeeders.html
Nguyen Duc Tuyen received a B.S.

degree in electrical engineering from
Hanoi University of Technology,
Vietnam in 2006 and an M.S degree
from Shibaura Institute of Technology,
in 2009. He is now a Ph.D Candidate at
Shibaura Institute of Technology, Japan.
From September 2006 to September 2007, he was an
Engineer of Electricity of Vietnam. Since October 2007, he
has been a Faculty Member in the Department of Electrical
Power System, Hanoi University of Technology. His
current research interests include modeling and control of
distribution power systems and dispersed power generation
systems.
Masakatsu Nomura received a B.E.

degree in electronics engineering and
the M.S. degree in electrical engineering from Nagoya University, Nagoya,
Japan, and the Ph.D. degree from Meiji
University, Tokyo, Japan in 1975, 1977
and 2009 respectively. Since 1977, he
has been with Meidensha Corporation, where he is
currently working in the New Product Development Group,
Tokyo, Japan. His interests are power electronics, control
and factory automation. Dr. Nomura is a member of IEEE,
IEEJ and Japan Society for Simulation Technology.
eBook for You
Goro Fujita was born in January 1970.

He received B.E., M.E. and Ph.D
degrees in electrical engineering from
Hosei University, Tokyo, Japan in 1992,
1994 and 1997, respectively. In 1997,
he was a research student at Tokyo
Metropolitan University. Since 1998,
he has been serving at Shibaura Institute of Technology in
Tokyo, Japan as an associate professor. He is interested in
power system control. He is a member of the Society of
Instrument and Control Engineers (SICE) of Japan, the IEE
of Japan, and IEEE.
Toshihisa Funabashi graduated in

March 1975 from the Department of
Electrical Engineering, Nagoya University, Japan. He received, in March 2000,
a Doctor degree from Doshisha
University, Kyoto, Japan. He joined
Meidensha Corporation in April 1975
and has engaged in research and development on power
system simulations and integration of distributed energy
resources in the power network. Currently, he is a Senior
Fellow of the Power Systems Solutions Engineering
Division. Dr. Funabashi is a chartered engineer in UK, a
senior member of IEEE, a member of IET.
179
IEEJ TRANSACTIONS ON ELECTRICAL AND ELECTRONIC ENGINEERING

IEEJ
2012;
240250
TngTrans
hp cc
bi7:bo
khoa hc giai on 2007-2012
Published online in Wiley Online Library (wileyonlinelibrary.com). DOI:10.1002/tee.21724
Paper
Adaptive Notch Filter for Synchronization and Islanding Detection using
Negative-sequence Impedance Measurement
Nguyen Duc Tuyen , Student Member
Goro Fujita , Member
Toshihisa Funabashi , Member
Masakatsu Nomura , Member
Keywords: islanding detection, impedance measurement, adaptive notch lter
Received 9 July 2010; Revised 23 December 2010
1. Introduction
The growing interest in environmental issues, the progress in
technologies to couple renewable energy sources to the grid, and
the liberalization of the energy market have led to a growing share
of grid-connected dispersed generation (DG). In the context of
DG units, an island is formed when one or more DG units are
disconnected from the utility grid due to accidental or preplanned
events while they are still in operation. The current generalized
industry standard requires disconnecting all the DG units from the
island as fast as possible, for example, the IEEE 1547 standard is
in 2 s [1]. For a smooth transition from a grid-connected mode
to an islanded mode of operation, fast islanding detection (ID) is
essential in order to allow DG units to adapt their control strategies
to new operational strategies. In fact, the detection time interval,
i.e. 2 s, is not fast enough for controller reaction in autonomous
operation.
There exist many ID methods that can be fundamentally
classied into two basic categories: the communication method
and the local one, and the latter can then be subdivided into active
and passive detection schemes [2]. In [3], a simple control scheme
has been developed that is easy to accomplish in software and
hardware as well. With the low cost used for ID, this can support
the negative impedance measurement method. It initially presents
a passive ID by using negative-sequence impedance measurement
a
Correspondence to: Nguyen Duc Tuyen.

E-mail: m609504@shibaura-it.ac.jp
* Power System Laboratory, Shibaura Institute of Technology, 3-7-5,

Toyosu, Koto-ku, Tokyo 135-8548, Japan
** Meidensha Corporation, Ohsaki 2-1-1, Shinagawa-ku, Tokyo 141-6029,
Japan
based on the normal existence of a negative-sequence voltage

in the distribution network. Information about the actual grid
impedance change can be used for detecting islanding problems.
However, this method is not actively controlled in purely balanced
distribution networks or in a small imbalance network because of
the present limitation of the measurement devices. In the case
of distribution networks without unbalanced voltage, it cannot be
implemented. This paper validates the performance of an active ID
method which actively injects a negative-sequence current through
the power electronic interface of a DG unit as a disturbance signal
for ID. In 2008, Karimi et al. [4] proposed a method based on
injecting a disturbance signal into the system through either direct
axis (d -axis) or quadrature axis (q-axis) current controllers of
the interface voltage-sourced converter. However, it just tries to
estimate the magnitude of the corresponding negative-sequence
voltage at the point of common coupling (PCC) to detect islanding.
This method might malfunction in the case of a highly disturbed
utility grid. Negative-sequence voltages from a utility can either
add to or cancel out those produced by the negative-sequence
current injection. When the negative-sequence voltage components
cancel out, the method is within its nondetection zone (NDZ). On
the other hand, the method is a little complicated and the control
methodology depends strongly on external parameters, such as
the values of resistance and inductance of the interface circuit.
Moreover, 5% p.u. negative voltage injection is quite a high value
which impacts on the power quality.
According to Ref. 5, an adaptive notch lter (ANF) is a novel
synchronous device for the DG unitgrid interface because of
its fast reaction time in comparison to other types of signal
processing units. The application of ANF in a DG unit has not
been presented in previous approaches because the phase-locked
loop (PLL) is currently more popular. Nevertheless, in order to
2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.
180
eBook for You
Dispersed generation (DG) has been found promising for satisfying the requirements of high power quality in distributed
systems. One operation situation, namely, an island, is formed when one or more DG systems and an aggregate of local loads are
disconnected from the main grid and remain operational as an islanded entity. Islanding is either due to intentional events, e.g.
maintenance outage, or due to unintentional events, such as faults, and their subsequent switching actions. Islanding is usually
undesirable because of the potential harm it causes to the existing equipment, human safety, power reliability and quality, etc.
Accordingly, anti-islanding schemes are used to immediately detach a DG system from the feeder after islanding. This paper
presents an active method of detecting islanding for DG systems by injecting a small negative voltage into the point of common
coupling, and then measuring the negative system impedance. In this study, an adaptive notch lter is introduced as a synchronous
part instead of a phase-locked loop and as a signal processing unit as well. The proposed control strategy allows DG systems
to detect properly the occurrence of islanding in a balanced distribution system. Simulation results show the overall system
performance including synchronization, power control, and islanding detection capability of the simulated DG system. 2012
Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.
Tng hp cc bi bo khoa hc
giai on NOTCH
2007-2012
ADAPTIVE
FILTER FOR SYNCHRONIZATION AND ISLANDING DETECTION
I0
Z0sys
Z0load
V0
Zsys
Zload
I V
m
E+sys
Zmload
I0 + I + I+
I+
Z+sys
Z+load
V+
Fig. 3. Symmetrical-component current ow in an unbalancedsystem expanded circuit

Fig. 1. Proposed DG unitGrid interface system
make a more reliable synchronous tool for DG units, the PLL
needs be developed into a much more complex device. Besides, it
cannot be reliably used as a signal processing unit. In the proposed
method, ANF not only plays an important role in synchronizing a
DG unit with a utility grid but also processes the input signal for
ID schemes by extracting negative-sequence components as seen
in Fig. 1. The proposed controller methodology is very simple
and can be applied as an additional component of the existing
controller system without changing the entire controller devices.
Because of the simple algorithm of the controller and ANF parts,
this ID method is applicable in case of an almost purely balanced
distribution network.
The proposed method helps to detect the islanding phenomenon
in as fast as a few cycles (of 60 Hz system) for the specic settings
described in the simulation. Thus, it provides the option used not
only for anti-islanding applications but also for the context of
microgrids [6], where a fast ID is required to activate specic
control actions afterwards.
The rest of the paper is organized as follows. Section 2 briey
introduces the methodology of ID by impedance measurement.
Section 3 describes the power electronic interface to control power
and injection of the negative voltage. Section 4 proves why the
ANF can be used as a signal processing unit. Section 5 evaluates
the performance of the method based on simulation test cases in
the Matlab/SimpowerSystem environment. Section 6 discusses the
results of this study. Finally, the paper concludes with Section 7.
sys
V
I
(1)
where the subscript represents the negative-sequence

components.
The measured negative current I and negative voltage V
waveforms are composed of real and imaginary components.
sys
With these components, the negative system impedance Z , the
sys
sys
associated magnitude Z , and the phase Z are calculated as
sys
sys
2. Islanding Detection Based on Negative-Sequence

Impedance Measurement
E utility
CB
sys
sys
sys
and Z = V I . The step magnitude
3.1. DG-utility grid interface A DG unit is connected

to the distribution system through an inverter, as shown in Fig. 1.
The inverter performs two main functions: The rst is controlling
the power output of DG unit. The DG unit generates no reactive
power in order to keep unity power factor and to facilitate the
distribution voltage controlled by utilities. Second, according to the
IEEE Standard 1547, a DG unit should be equipped with an antiID algorithm. From Fig. 1, the power electronic interface includes
determination of the direct current bus voltage of the DG unit VDC ,
a link capacitor Cdc , an inverter, a lter, and a controller.
The power rating of almost all DG units is not higher than a
few megawatts. So, most of the inverters are based on insulated
gate bipolar transistor (IGBT) switches. The action of the power
electronic devices in such an inverter is much faster than that
of mechanical valves. It is recommended to incorporate an LC
low-pass lter between the inverter output and the AC mains, as
E DG
Fig. 2. Principle of using negative-sequence impedance measurement for islanding detection
sys
I
3. Power Electronic Interface
Z DG
Z load
change of negative system impedance is adequate for islanding

alarm signals.
The ID method developed in this paper takes the theoretically

accurate concept of impedance measurement when the negative
components of voltage and current are available. This idea can be
seen in Fig. 2. The premise is that the large impedance difference
between an island and grid-connected condition supposedly has a
very low NDZ for radial systems with strong network connections,
that is Z utility << (Z DG + Z load ). Such a change in impedance
would allow easier threshold settings (i.e., tolerance levels) for
ID.
The unbalanced system is transformed into the symmetrical system described in Fig. 3 [3]. In this system, the system equivalent
Z utility
241
IEEJ Trans 7: 240250 (2012)

181
eBook for You
sys
voltage source E+ remains as the only source and the current

sys
sys
sys
sys
from the Z0 and Z sequences are indeed supplied by E+ . E+
is used because the normal resources are considered to generate
only positive-sequence components. Thus, the subscript + represents positive-sequence components. From Fig. 3, a current source
that is produced at the load for negative and zero sequences is a
result of I+ being divided at node Vm . Vm is the point where three
symmetrical currents join together and m represents the mutual
components. From the current and voltage derivation in symmetrical components with this conguration, the current ows out of
the unbalanced load and goes into the system that allows the use
of Ohms law for negative-sequence component measurement as
stated in (1).
sys
The negative-sequence impedance of system Z can be calculated as in (1) when the negative-sequence current is used [3].
This method of impedance measurement lets the negative-sequence
components to be used as a real-time anti-islanding sensor relay by
measuring the impedance away from an unbalanced node and then
comparing it with an already known threshold before and after the
island state occurs.
intermediary stage called the frame, as follows:
v
2 1 1/2 1/2 aPCC
v
vbPCC
=
v
3/2
3 0 3/2
vcPCC
Vdcref
Vdc(t)
PID
iDGa(t)
iDGb(t)
IDGd
IDGdref
PID
IDGq
T(q)
V DGqref
V DGdref
PID
iDGc(t)
2
V 2DGqref + V DGdref
tan1
IDGqref
qq
vaPCC(t)
vbPCC(t)
IDGqref = 0
(Qref = 0)
and
V +DGqref
cos
vd
=
vq
sin
V +DGdref
ref
qq
vcPCC(t)
Inverter
V DGaref
VDGaref
Switches
control
PWM
i
iq
vq 0
c
id
q
a
C d
v aPC
C
PC
va
(3)
3.3. Three-phase adaptive notch lter in synchronization role One important part of the grid-connected DG units
vd
is the grid synchronization unit. Synchronization techniques play

an important role in providing a reference phase signal synchronized with the grid voltage to meet the power quality standards in
all DG systems. It requires fast and accurate detection. One of the
two main objectives of this paper is to propose a three-phase ANFbased technique. The PLL is a fundamental concept used in various
disciplines of electrical technology [7]. The principal idea of the
PLL is to generate a signal whose phase angle adaptively tracks the
variations of the phase angle of a given signal. The complexity of
phase locking is due to noise, distortions, and frequency variations.
The concept of the PLL is very much similar to that of an ANF.
However, the PLL is fundamentally different from the ANF in the
sense that it actively generates its output signal whereas the ANF
(which is frequency based) extracts it passively from the input signal. ANF represents as a synchronization scheme that offers a high
degree of insensitivity to power system disturbances, harmonics,
and other types of distortion existing in the grid signal as well
as structural simplicity because it requires no voltage-controlled
oscillator as in the case of a PLL. ANF is a basic adaptive structure that can be used to extract the desired sinusoidal component
of a given periodic signal by tracking its frequency variations. The
dynamic behavior of the newly modied ANF is characterized by
the following set of differential equations [5]:
..
2
x. + x = 2 e(t)
(4)
= x e(t)
e(t) = u(t) x
a,b,c refer to actual phase variables
Fig. 5. Vector system in abc dq transformation

shown in Fig. 1, to reduce the high-frequency components of the
inverter currents. A small resistor R is also inserted in series with
the capacitor to suppress possible oscillations. Besides the power
electronic interface, the utility and load parameters considered in
this paper and the design values are given in Table I of Section 5.
The control system is demonstrated in Fig. 4. For this interface
control, both the IDGd and IDGq components of the DG output
current are controlled to be equal to a preset value, i.e. IDGd ref
and IDGqref . Current control consists of PID controllers. Figure 4
shows the abc to dq transformation block (denoted by T ())
and its related ANF that provides the required transformation
angle, i.e. . Superscript represents the ripple components due
to the transformation. The control block diagram for vDGd

ref ,
vDGqref is obtained subsequently. The modulation index m and

angle transformation ref will be implemented in order to get
the abc coordinate voltage reference (which is represented by
= a, b, c) fed into pulse-width modulation (PWM) generation,
which are used as control signals for the inverter. The AC
voltage generated by the inverter is forced to follow the reference
signals produced from the synchronous reference generator of the
controller.
where u(t) is the input signal, e(t) is the error signal, is the
estimated frequency, and and are adjustable real positive
parameters that respectively determine the estimation accuracy and
the capability of the algorithm in tracking the signal characteristics
variations. A trade-off between the (steady-state) accuracy and
(transient) convergence speed can be obtained by adjusting those
design parameters. By increasing , one can achieve faster
convergence speed; however, at the same time, should be
increased to avoid oscillatory behavior. The ANF structure has
three integrators. The initial condition for the one that outputs the
frequency is set to the nominal power system frequency of 2.60
rad/s. The initial conditions for all other integrators are set to zero.
For a single sinusoidal input signal u(t) = A sin (t + ) where A
3.2. abc dq Transformation to control active and

reactive power In the grid-connected mode, the grid is
assumed to be stiff enough, e.g. short-circuit ratio SCR > 5, to
maintain the PCCs voltages and frequency at the rated values
without signicant distortion. As a result, the DG units should
be capable of regulating only the power ows. Therefore, DG
units can utilize the conventional dq-current control strategy. This
section proves the principle of this control methodology as shown
in Fig. 5.
The synchronization methods use a transformation from the abc
to the dq frame, for which an estimation of the phase angle is
required. This transformation can be formulated by means of an
v
v
Fig. 4. Proposed DG control methodology
sin
cos
where vaPCC vbPCC vcPCC are the PCCs voltages and

is an estimation of the phase angle of vaPCC . Note that v
and v are in phase with and orthogonal to vaPCC , respectively, and they have the same magnitude. However, under this
frame transformation, the voltage at the PCC and the current from DGs will have themselves two components at the
and axes. If an ideal estimation of the phase is available,
i.e. = , then dq-frame voltages will be constant values of
vd vaPCC and vq 0. Therefore, after synchronization with
voltages at the PCC, which are compensated by stiff utility grids,
the generated active and reactive power can be handled easily by controlling only the d -current and q-current components
separately.
V +DGaref
ANF
(2)
242
IEEJ Trans 7: 240250 (2012)

182
eBook for You
Pref
N. D. TUYEN ET AL.
giai on NOTCH
2007-2012
ADAPTIVE
x..
1
s
theta_xa
x.
ea
1
Fundamental (x.)
1
s
2*zeta
ea
xa.
xa_
x
theta_xa
theta theta*xa
u2
Input
signal
u(t) 1
theta_xb
theta
2
Frequency (w)
theta
ANF-a
eb
u_a(t)
eb
xb.
e(t)
e(t)
1
s
-gamma
theta
x.
u_b(t)
abc
u(t)
5
Fundamental_90 (-theta*x)
sqrt
ec
ec
xc.
xc_
xc.
theta_xc
theta theta*xc
xb_
3
Amplitude (A)
theta_xb
ANF-b
x*theta
u2
xb_
xb.
theta theta*xb
Frequency
estimator
u_c(t)
4
Harmonic(h)
x.
xa_
xa.
ANF-c
u2
theta_xc
x.
theta
Fig. 6. Implementation of the single-phase ANF
xc_
Fig. 7. Implementation of the three-phase ANF
ua
4
x e (t)
-gamma
eb
1
s
4
theta
3
ec
7
xb.
8
theta*xc
uc
6
ec
9
xc.
Fig. 8. Frequency estimator in three-phase ANF

The dynamical system given by (6) and (7) has a unique periodic
orbit located at
A
x
cos (t + )
0
p (t) = x = A sin (t + )
(9)
(6)
and = . For the three-phase ANF in the steady state, the dened
outputs x and x are
x = A sin (t + ) u (t)
x = A cos (t + ) S90 u (t)
(10)
where x is the fundamental component and x is its 90 phase

shift which is represented by S90 .
A structural block diagram of the rst stage of the algorithm
is shown in Fig. 7. Detailed implementation block diagrams of
the frequency estimator and the sublter (ANF ) are shown in
Figs 8 and 9. The basic structure of the proposed system has two
independent design parameters, i.e. and (a,b,c) .
(7)
=a,b,c
For a three-phase sinusoidal signal, u(t) is given by

Aa sin (t + a )
ua (t)
u(t) = ub (t) = Ab sin (t + b )

uc (t)
Ac sin (t + c )
2
eb
3
theta*xb
ub
5
where is, again, an estimate for .

To derive an equation for estimating , we note that (i)
is the common frequency of the three-phase signals, therefore,
information of all three sublters must be incorporated into the
updated law for frequency estimation; and (ii) the input signal
u(t) and the error signal e(t) incorporate into the updated law
in (5). The term is for scaling. Therefore, the update law for
frequency estimation is proposed to be
=
ea
2
xa.
The third entry of o is the estimated frequency, which is identical

to its correct value . Figure 6 shows the schematic structure of
the proposed technique.
In the three-phase system, a synchronization algorithm can be
implemented in the abc frame by means of three single-phase ANF
systems introduced previously. Each single-phase ANF is a thirdorder dynamic system as in (4), hence the three-phase system has
a nine-order dynamics. However, the three-phase signals have a
common frequency , and therefore there is no need to estimate
the frequency of each phase independently. Moreover, separate
operation of ANFs makes the overall system complex for hardware
implementation and demands a high volume of calculations for
software implementation. Additional order might also cause delays
and reduce the speed of convergence. A three-phase ANF-based
synchronization for three-phase systems is introduced. The threephase-based method uses a seven-order dynamical ANF system.
Consider three identical ANFs [8]:
x = 2 x + 2 e (t); = a, b, c
e (t) = u (t) x
1
ea
1
theta*xa
eBook for You
is magnitude and is phase angle, this ANF space has a unique

periodic orbit located at
A
x
cos (t + )
.

(5)
o = x = A sin (t + )
3.4. Negative-sequence injection topology The proposed ID method is based on injecting a negative-sequence and
fundamental frequency current component through the power electronic interface. Afterward, the ANF detects the corresponding
(8)
243
IEEJ Trans 7: 240250 (2012)

183
N. D. TUYEN ET AL.
the tolerance of the d q transformation controller. Otherwise,

separate complicated positive and negative controls must be used
as in Ref 4. As a result, the DG unit generates ripple active
power. This circumstance is not only facilitated by stiff grids that
can supply unmatched powers between local loads and the DG
output power, but also assist under/overvoltage (UVP/OVP) and
under/overfrequency (UFP/OFP)-based ID methods by eliminating
matching power cases, which is the NDZ of conventional passive
ID methods.
e_alpha
x_alpha.
1
1
s
2*zeta
1
s
2
theta*x_alpha
Fig. 9. ANF of the three-phase ANF system
4. Three-phase adaptive notch lter in signal

processing role
negative-sequence voltages at the PCC and DG units current owing into the PCC, which results in a negative system impedance.
To inject the negative-sequence currents, the conventional positivesequence current controller of the inverter is slightly augmented
with a negative-sequence voltage reference. In this study, the controller uses the ANF to measure a three-phase signal and derives
a single-phase reference signal. The distinct phase references for
individual phases are obtained by adding or subtracting 2/3 radians to the measured phase angle. Such a three-phase design has the
advantage of triple-harmonic cancelation. As shown in Fig. 4, for
the positive-sequence component, three reference phases a, b, c are
composed from the modulation index m and angle ref , formed by
Fast and accurate detection of the negative-sequence components of the utility voltages and currents is a prerequisite for the
proposed ID method. The ANF might be extended to meet this
function. This section proposes an ANF-based signal processor
that has the three-phase synchronization scheme proposed in the
previous section as the main cell. It is developed to extract key
power system information that is required for ID. The input signals
u(t) can be decomposed to positive-, negative- and zero-sequence
components u(t) = u+ (t) + u (t) + u0 (t) and are related to the
input signal u(t) by the following equations [8]:
+
= m sin m +
uaref
+
= m sin m + 2/3
ubref
+
= m sin m + + 2/3
ucref
u+ (t) = T2 u (t) + T1 S90 u (t)

u (t) = T2 u (t) T1 S90 u (t)
u0 (t) = (I 2T2 ) u (t)
(11)
where I is a [3 3] identity matrix and T1 and T2 are 3 3

matrices given by
0
1
1
0
1
T1 = 21 3 1
1
1
0
(14)
1
0.5 0.5
1
0.5
T2 = 13 0.5
0.5 0.5
1
and the negative-sequence components are in the reverse rotation

dened by
uaref
= m sin m +
= m sin m + + 2/3
ubref
= m sin m + 2/3
ucref
(13)
(12)
where, especially, the commonly used subscripts + and are

changed into superscripts for easy expression.
The implementation of negative voltage injection is illustrated
in Fig. 10. Under this construction, the reference voltage fed into
PWM is unbalanced.
However, it is widely known that d q transformation cannot
deal with the negative-sequence component in an unbalanced
system except for separately controlling the negative component.
The reason is that the one-dimensional components d and q will
become periodic. It makes the modulation index m to uctuate.
In fact, the m value cannot be kept constant because of the noise.
Therefore, the reference voltage will hardly be in pure balance.
However, the ripple is not signicant enough for the measurement
devices to keep tracking them. In this proposal, a small negative
voltage (0.5% positive sequence) is introduced to make this signal
more detectable. The results are a ripple modulation index, then
negative currents from the DG unit, and voltages at the PCC.
It should be understood that the ripple sometimes is higher than
Figure11 shows the structure of the linear transformation

to estimate the symmetrical components. The three-phase ANF
introduced in the previous section receives the three-phase signals
ua (t), ub (t), and uc (t). Each sublter provides the fundamental
component of each phase, its 90 phase shift, and its frequency.
Then, all the information can be used to calculate the symmetrical
components.
1/3
1
xa.
1/2
4
ua
va
Va+
1/(2*sqrt(3))
1
ua+
6
theta_xa
vb+
3
xb.
m
1
[0 2*pi/3 2*pi/3]
vb
4
theta_xb
2
xc.
2
phi_m
sin
1/(2*sqrt(3))
3
theta
Uref P1
0.005
[0 2*pi/3 2*pi/3]
1
Pulse
vc
1/2
Discrete 3-phase
PWM generator
% Negative
sequence
injection
vc+
5
theta_xc
2
ub+
5
vb
3
uc+
6
uc
1/3
sin
Fig. 11. Linear transformation of the symmetrical-component calculator
Fig. 10. Negative-sequence voltage injection
244
IEEJ Trans 7: 240250 (2012)

184
eBook for You
1
2
theta
giai on NOTCH
2007-2012
ADAPTIVE
Table I. System parameters

Inverter
Switching frequency
Input DC voltage
Filter inductance
Resistance in series with the inductor
Filter capacitance
Resistance in series with the capacitor
Voltage (line to line)
Load
P
QL
QC
Grid
Frequency
Voltage (line to line)
Short-circuit level
X /R ratio
Controller
Vdc
Id
Iq
ANF
PQ_DG
100
0
Active power
Reactive power
100
15 000 Hz
900 V
0.8 mH
4m
30 F
1
480 V
200
0.1
0.2
0.3
0.4
0.5
0.6
0.4
0.5
0.6
0.4
0.5
0.6
PQ_load
100
50
0
50
0.1
0.2
0.3
PQ_utility
48830 W
45 891.68 Var
45 891.68 Var
200
0
200
400
60 Hz
600 V
40 MV
7
0.1
0.2
0.3
Time
Fig. 12. Power ows within DG system
Ki = 4, Kp = 0.04
Ki = 2, Kp = 0.05
Ki = 2, Kp = 0.05
= 18 000, = 0.6
(a)
1.5
1
0.5
5. Evaluation of Results from Simulation

0
In order to show the performance of the proposed control strategy based on the ANF under grid-connected and islanding operating conditions, the simulation model of the proposed DG system as
in Fig. 1 has been built and evaluated on Matlab/SimpowerSystem
environment. The system parameters of simulation are given in
Table I.
In this case study, the output power of the DG unit is limited to
50 kW. Moreover, the inductance and the capacitance of the load
should be selected so that their resonant frequency equals the fun1
= 2. (60 0.1Hz).
damental frequency of the system, i.e. LC
Satisfying this condition, the overall load circuit has unity power
factor and appears to be purely resistive. Therefore, the load
draws a reactive power which is equal to the DG capacity minus
the loss in the lter and almost no reactive power, i.e. Q =
1
V 2 L
C 0. This scenario represents the worst case for
ID, since no noticeable change in either the voltage or the frequency at the PCC occurs when utility circuit breaker (CB) is
opened. Voltages and currents in this simulation are measured in
p.u.
1
1.5
0.1
0.2
0.3
Time
0.4
0.5
0.6
0.4
0.5
0.6
(b)
2
1.5
1
0.5
0
0.5
1
1.5
5.1. ANF performance in synchronization role
The
test system initially operates in a matched power grid-connected
mode. On opening the switch CB , the system is islanded during t =
0.4 and 0.5 s. The interval from 0.0 to 0.1 s is the system transient
initiated time of the simulated system. Figure 12 shows the reactive
and active power of the DG unit, load, and utility. The active power
generated by the DG unit is matched with the local loads power
demand, while the reactive power of load draws almost zero power
because of the condition of the fundamental frequency resonance.
The DG unit controls its output reactive power at zero to meet
unity power factor.
Figure 13 shows the instantaneous point of the common
couplings voltages and the inverter output currents. Mainly due to
the injection of the negative-sequence currents, the PCCs voltages
and the DG units output currents (before and after the islanding
event) are not balanced and cause almost no change. However,
after reclosing the CB, these voltages and currents still deviate
from the normally operated range even in case of nearly matching
power because of the ripple power generated from the DG unit. In
Fig. 13 also, since a tiny amount of unbalanced voltages is injected,
0.1
0.2
0.3
Time
Fig. 13. Point of common coupling (PCC) voltages and generated

currents by the DG unit. (a) Voltage at PCC (VPCC ) and (b)
generated currents by DG (IDG )
the voltages at the PCC observed are seemingly in balance. In
Fig. 14, the reference voltages from the controller are presented.
With 0.5% injection, the nal reference of the inverter output
voltage looks slightly unbalanced. On the other hand, the currents
through three phases are unbalanced.
In the grid-connected mode, power electronic interface operates
as a current source and injects both negative- and positivesequence currents into the innite bus. When the switch CB is
closed, the voltages and frequency at the PCC are dominantly
dictated by the grid. The estimated frequency in Fig. 15 shows
that, even in the condition of nearly matching power, the frequency
245
IEEJ Trans 7: 240250 (2012)

185
eBook for You
0.5
N. D. TUYEN ET AL.
ANF-phase
U+ref
6
4
0.1
10
0.2
0.3
0.4
0.5
0.6
0.1
0.2
0.3
Uref
0.4
0.5
0.6
0.4
0.5
0.6
0.5
0.6
PLL-phase
6
0
5
4
2
0.1
0.2
0.3
0.4
0.5
0.6
0.1
0.2
0.3
Uref
1
Positive voltages at PCC

1
0
1
0.1
0.2
0.3
Time
0.4
0.5
0.6
1
0
0.1
0.2
0.3
Time
0.4
Fig. 14. Reference voltages

Fig. 16. Extraction ability of ANF
System frequency
390
(a) Negative-sequence injection currents

0.2
380
0.15
370
0.05
360
0.05
0.1
0.1
350
0.2
340
330
0.1
0.2
0.3
Time
0.4
0.5
0.1
0.2
0.3
Time
0.4
0.5
eBook for You
0.15
0.6
(b) Negative-sequence voltage at PCC
0.6
0.02
0.015
Fig. 15. Estimated frequency by ANF
0.01
0.005
still deviates considerably from the nominal values, which can be

used as a back-up ID method.
The ability of ANF in extracting the phase angle of the PCCs
voltage is shown in Fig. 16. The phase angle extracted by the
conventional PLL is also presented for comparison. It should
be noted that, in a matched power condition, if subsequent
to an islanding event the islanded system does not change its
control strategy, its electrical signals may remain within their
acceptable limits even for several seconds. However, any change
in the demanded power may result in large deviations of the
electrical signals and even in the instability of the islanded system.
Therefore, the islanding event must be rapidly detected, and the
control strategy must be immediately modied after the detected
island instance. This will be dealt with in another paper.
0
0.005
0.01
0.015
0.02
0.1
0.2
0.3
Time
0.4
0.5
Fig. 17. Negative-sequence components of VPCC

generated currents
0.6
and DG-
detection is determined by the two aforementioned parameters, i.e.

and (a,b,c) .
5.2.2. Islanding detection time and tolerant level Figure 18 indicates that before islanding (t < 0.4s), the system
negative-sequence impedance is dictated by the low grid impedance
and remains small, i.e. about 0.06 p.u.
Based on the mechanism of impedance ID, a simple method
is proposed to set tolerance levels so that the protection system
functions properly. As mentioned in Refs 9 and 10, the impedance
ID has a small NDZ, which means that choosing the tolerance
levels can be implemented simply in practical systems. The
sensitivity concept is then referred to. The relationship between
the impedance of grid-connected system and that of islanded entity
should be considered. On one hand, the grid-connected impedances
5.2. ANF performance in islanding detection role

5.2.1. Generated negative-sequence components Figure 17 shows the instantaneous values of the negative-sequence
component of DG-generated currents and the PCCs voltages,
respectively, which are detected by the ANF part successfully.
In grid-connected mode (t < 0.4s), the unbalanced voltages at the
PCC are controlled by the stiff grid. And due to effects of the
lter, their values are less than the reference values. Thus, it does
not impact on the power quality but, by contrast, facilitates ID
properly. During the transient periods, especially after reclosing the
CB, the ANF takes time to track those components. This transient
246
IEEJ Trans 7: 240250 (2012)

186
over the tolerance level many times before the islanding truly
occurs. Thus, in those cases, the ID scheme should consider
that no islanding appears. The one-time checking design could
make the ID relay work incorrectly. Therefore, multichecking
is necessary for a proper operation. By contrast, the more the
number of checking times, the longer the detection time is
pushed to increase.
Regarding the tolerance levels, determining the xed value
that is suitable for almost all operation scenarios is never an
easy task. The tolerance level is then determined on the basis
of negative-sequence impedances of the grid-connected mode.
As in Fig. 18, the grid-connected impedance is roughly 0.05
p.u. Therefore, the higher value of the tolerance level should
be chosen. However, how much higher this value should be
above 0.05 p.u is left to the operating engineers. This chosen
tolerance value is a compromise between two things. The rst
one is the disturbance during the grid-connected mode such
as opening induction motors (which also makes the negativesequence impedance to increase [3]) or other perturbation, as
seen in Fig. 18. For instance, if the tolerance level is set at a
small value as much as 0.08 p.u, then at the 0.13 s instance the
increasing impedance phenomenon is detected as an islanding.
And the second one is of high values that even the occurred
island could not bring negative-sequence impedance to reach.
The tolerance level is set at 0.9 p.u is such an example.
In Fig. 19, an example of theoretical analysis shows how to
set up the tolerance levels. This grid is an IEEE 13 bus
system [10]. This system involves multiphase distribution line
segments including three-phase, two-phase, and single-phase
lines. Instead of the innite bus in the previous simulation,
the multiload grid with existing unbalance components is
considered. Modication of this grid was taken into account
in order to be suitable for this analysis. One unit DG is
inserted at node 675. It should be understood that there are
power-consuming loads at every node. Therefore, under this
conguration, once line interruption happens at one of the four
sections (I, II, III, IV as shown in Fig. 19), the islanding takes
place. If islanding happens at section I, the DG unit will supply
for local loads at node 675. The change symptom of negativesequence impedance manifests, which has been disconnected as
shown in Fig. 20. The grid-connected impedance is retained at
the smallest value during the normal operation of the system.
The order of tolerant levels, i.e. TLi (i = 1 to 4), depends on
the respective interruption parts. The highest impedance of the
islanded part occurs when opening section I, while the smallest
islanded impedance is when opening section IV. The tolerance
level of each section is chosen as clearly shown Fig. 20.
System negative impedance mesured from DG view
2.5
p.u
2
System
impedance
reaches
tolerant
level
1.5
Tolerant level 2
Islanding detected
Tolerant level 1
Checking-time 2
Checking-time 1
0.5
0.1
0.2
0.3
Time
0.4
0.5
0.6
Detection-time2
Detection-time1
Fig. 18. System negative-sequence impedance used for islanding

detection and detection time depending on the chosen tolerance
level
can be easily measured live. On the other, in large distribution
systems it is hard to determine the impedances of all the cases
when the respective islanded entities are established. Therefore, a
higher value of the grid-connected impedance should be chosen to
set the tolerance level. Consequently, the ratio between the gridconnected impedance and the tolerance level is referred as the
tolerance factor Ktf of ID relay as follows:
Ktf =
Ztolerant level
gridconnected
(15)
Clearly, the tolerance factor which is greater than 1 can be

regulated under practical systems to ensure the reliable operation
of ID relays in any established islanded entities. Under this
conguration, in this study system, tolerance level 1 coincides
with Ktf = 3.6 and tolerance level 2 coincides with Ktf = 8.3 for
the following analysis.
After islanding (t = 0.4s), the impedance rises sharply and
reaches the threshold in 0.0035 s with the tolerance level 1 and in
0.044 s with the tolerance level 2. For an explanation, tolerance
levels 1 and 2 are chosen for a typical analysis. In Fig. 18, the
initial period of the simulation system shows that the impedance
instantaneously exceeds the setting values during the rst period
and rapidly diminishes afterward. If we run this system in the
snapshot mode which eliminates the initiate period of the DG unit,
the rst high impedance will not be noticed. In fact, the ID relay
could be designed to be locked during the initiating time of the
DG unit in order to get rid of this malfunctioning. The subsequent
reaction of this simulated system is that, almost immediately
after islanding, the system negative-sequence impedance increases
dramatically to much higher values in 0.033 s, and then decreases
again to the previous small values. Once more, the impedance
is increased to higher values than the previous one and this
increased impedance value keeps itself high until the CB is closed
at 0.5 s. In fact, this unstable reaction depends on the arbitrary
characteristics of the system components such as transformers,
lters, loads, etc. This means that in some cases this negativesequence impedance permanently stabilizes at high values once it
reaches the thresholds due to islanding; in other cases, it repeatedly
increases and decreases. Practically, the detection time basically
depends closely on the detection algorithms and the tolerance
levels chosen.
Regarding the detection algorithms, owing to the live negative-
sequence impedance measurement of this method, the ID signal

must cooperatively be checked several times in order to make
it more reliable. As shown in Fig. 18, if there is only a onetime checking, the negative-sequence impedance value will get
Fig. 19. Modied IEEE 13 node using for tolerance level analysis
247
IEEJ Trans 7: 240250 (2012)

187
eBook for You
giai on NOTCH
2007-2012
ADAPTIVE
various practical systems. Moreover, because the ID device is

very important, a new device based on a new theory is important.
Therefore, there are currently still new ID methods that are being
pursued. There are also many uncompleted potential approaches
in the ID eld such as wavelet methods, inverted-based DG, and
correlation-based methods. New methods still keep appearing. It
is hard to make a comparison among all ID methods one by one
because each method works properly in particular networks and
specic DGs. Sometimes one method works in one system but does
not in the others, indeed. In addition, the standard detection time
is different in each country. For example, some countries do not
consider interference with automatic reclosers to be an issue and
allow up to 5 s to detect an islanding and to disconnect DG unit,
while the U.S. standard requires 2 s to implement those steps. The
idea of this paper is to introduce an ID method with advantages
explained in the previous parts to detect islanding events that is
probably suitable for some kinds of real power systems.
A method is feasible in a multi-DG grid, which is the challenge for almost all conventional methods. Since the concept of
using negative-sequence impedance measurement for ID has been
explored by Wrinch et al. [3], it can be concluded that the proposed
method is not only suitable for one DG itself but also for large
networks with any DG penetration levels. However, the scheme
did not fully discover the effect of high penetration of DG units
and motors to the electrical network, and this drawback is being
currently solved by further research which will be reported in the
near future. Necessarily, the assumption that there is not enough
unbalanced voltage and current inside networks is the main component that makes our method work. As in the case of the shortage
of negative-sequence components, DG should generate a small
amount of those into approximately pure balanced networks.
In comparison to the other phenomena that the conventional
passive methods (e.g., frequency drift with positive feedback, voltage amplitude, frequency of terminal voltage) take difculties to
distinguish from, the negative-sequence impedance measurement
method overcomes those challenges by its unique topology. For
example, in case of closing or opening large loads, though the
voltage and current variations are large, the impedance changes
insignicantly because of Z utility << Z huge loads . On the other hand,
in case of short-circuit faults, the impedance of the overall system
becomes even smaller in contrast to an islanding event in which
the system impedance increases.
In Table II, a rough comparison among popular active methods
is given in order to explain the advantages of the proposed method
over the others. The passive method is not considered because of
their wide NDZ. The real detection period then depends on several
other aspects such as logical detection design, signal transfer speed,
and so on. More or less detection time depending on realistic
scenarios is still around this theoretical period of the simulated
system which is really fast (the detection time is 1.53.5 electrical
cycles is quite fast among other ID methods as stated in Ref. 4
because almost all methods need at least one cycle pending the
electromagnetic response. Not only fast but also a small NDZ
make this method novel. It should also be noted that the UL1741
anti-islanding test requires run-on times of 2 s or less. Therefore,
the proposed ID is promising for autonomous operation of the DG
system.
Fig. 20. Measured negative-sequence impedance of the system

from manifests which has been disconnected
As for the two chosen tolerance level values shown in Fig. 18,
the tolerance level 1 accelerates the detection time. Checking-time
1 happens in 0.025 s. The ID time 1, which is counted from the
actual islanding to the detected instance, is equal to 0.0285 s.
However, if checking-time 1 is longer than the period shown in
Fig. 18, the ID scheme probably realizes other small impedance
values after acknowledging the high values during the very rst
time. Consequently, the ID relay is reset again for restarting
another ID. Therefore, the ID scheme that chose tolerance level
1 somehow cannot detect islanding as fast as it desires, although
the islanding nally is detected by the next detection iteration.
Meanwhile, if tolerance level 2 is chosen, the ID relay does not
suffer the relays reset action as in the previous case. To trade off,
the ID time 2 of 0.069 s is slower than the detection time 1.
5.2.3. Sensitivity to utility impedance Figure 21 shows

the response of the negative-sequence impedance detection method
to different values of the utility by changing the utility capacity
from 4 to 0.4 MW. The system negative impedance difference
between the grid-connected mode and the islanding mode in case
of the 0.4-MW utility becomes insignicant. However, Fig. 21
shows that the source stiffness or how large the utility capacity
is has almost no impact on the detection time, but it impacts the
detection levels. It faces obviously more difculty to choose the
tolerance level for the 0.4-MW utility. In case of smaller capacity
grids, this proposed ID method whose tolerance level is set up in
advance depending on the specic parameters of the utility grids
successfully detects the islanding events.
6. Discussion
As for the ID methods, the standardization is now under
discussion. However, the present methods might not be good for
System negative impedance measured form DG view
5
4 MW utility
0.4 MW utility
4.5
4
3.5
Tolerant level
for 0.4 MW
p.u
3
2.5
Tolerant level
for 4MW
2
1.5
7. Conclusion
This paper introduced a new ANF-based approach which acts

as synchronization part for the controller and the extraction of
negative-sequence components as well. The proposed system can
benecially replace the conventional PLL in those applications that
require a more accurate estimation of the parameters in unbalanced
conditions.
0.5
0
0.1
0.2
0.3
Time
0.4
0.5
0.6
Fig. 21. Negative-sequence impedances in different utility capacities

248
IEEJ Trans 7: 240250 (2012)

188
eBook for You
N. D. TUYEN ET AL.
giai on NOTCH
2007-2012
ADAPTIVE
Table II. Active-method comparison

Tested parameters
Speed
Main characteristics
Enhanced disturbance method:

- Sandia frequency shift (SFS)
- Sandia voltage shift (SVS)
- Slip mode frequency shift (SMS)
- Autonomic phase shift (APS)
- Active frequency drift (AFD)
Remote techniques
Current injection
Phase, magnitude and frequency

of voltage or current
Fast
- Malfunction by other transients

(short circuit, loads attachment or detachment, etc)
- Large NDZ with high power quality factor-Q
- Detain an autonomous operation
Status of circuit breakers

Through q- or d-axis
Fast
Very fast
Positive-sequence impedance
Zpositive =
Harmonic impedance
Zharmonic =
Negative-sequence Impedance
Znegative =
dVpositive
dIpositive
Fast
dVharmonic
dIharmonic
Slow
Vnegative
Inegative
Very fast
The investigations reported in this paper show that injection

of an unbalanced signal through the power electronic interface is
a viable approach for ID requirements. Mathematical derivations
and computer simulations were presented to describe the principles
of operation and the performance of the proposed system. This
ID method imposes no additional design requirements to the
current controllers, but only adds a negative-sequence component
to the reference signal. This study indicates that the injection of
about 0.5% of the inverter rated voltage is adequate to detect
the islanding phenomenon within two cycles at 60 Hz. The
reliability of this detection strategy can be enhanced if the counter
block of the detection algorithm is set to ag islanding after a
certain number of times. This enables the detection algorithm to
distinguish islanding from impedance transients or other sustained
perturbations. The established power ripple facilitates OVP/UVP
and OFP/UFP as a back-up strategy.
Although the existing standards prohibit the autonomous operation of DG units due to potential safety issues and inadequate
grounding, the high depth of penetration of DG units and the economic aspects indicate that the islanded mode of operation will be
a viable operational model. There exist ID techniques that aim to
destabilize the system by means of controller gains when the grid
is disconnected and the DG is forced to shut down after islanding
is detected. Such ID methods can only be used as anti-islanding
strategies. However, if an ID method is implemented for a DG
unit which is intended to keep operating as a part of a micro-grid
after islanding occurs, the DG controls are expected to enhance
voltage/angle stability and maintain operation of the microgrid
immediately after ID. The method presented in this paper can also
accommodate this requirement. As soon as islanding is detected,
the current control can be commanded to stop injecting the unbalance signals into the system and change to another control strategy
to keep voltage and current stable under the continuing connecting
loads.
(4) Houshang Karimi RI, Nikkhajoei H. Negative sequence current

injection for fast islanding detection of a distributed resource unit.
IEEE Transactions on Power Electron 2007; 55(1):8593.
(5) Mohsen Mojiri AB, Karimi-Ghartemani M. Time-domain signal
analysis using adaptive notch lter. IEEE Transactions on Signal
Process 2008; 23(1):298307.
(6) Houshang Karimi RI, Nikkhajoei H. Control of an electronicallycoupled distributed resource unit subsequent to an islanding event.
IEEE Transactions on Power Delivery 2008; 23(1):493501.
(7) Timbus RFAV, Liserre M. Synchronization methods for three phase
distributed power generation systems. An overview and evaluation.
Proceedings of IEEE Power Electronics Specialist Conference, 2005;
24742481.
(8) Yazdani D, Mojiri M, Bakhshai A, Joos G. A fast and accurate
synchronization technique for extraction of symmetrical components.
IEEE Transactions on Power Electronics 2009; 24(3):674684.
(9) Kunte RS, Wenzhong Gao. Comparison and review of islanding
detection techniques for distributed energy resources. Power Symposium, NAPS 08. 40th North American, 2009; 18.
(10) D. T. F. W. Group. Distribution test feeders. http://ewh.ieee.org/soc/
pes/dsacom/testfeeders/index.html [Last accessed January 2010].
Nguyen Duc Tuyen (Student Member) received the B.S. degree

in electrical engineering from Hanoi University
of Technology, Vietnam, in 2006 and the M.S
degree from Shibaura Institute of Technology,
Japan, in 2009. He is now pursuing the Ph.D.
degree at Shibaura Institute of Technology. From
September 2006 to September 2007, he was an
Engineer of the Electricity of Vietnam. Since
October 2007, he has been a Faculty Member in the Department of
Electrical Power System, Hanoi University of Technology. His current research interests include modeling and control of distribution
power systems and dispersed power generation systems.
Goro Fujita (Member) was born in January 1970. He received the

B.E., M.E., and Ph.D. degrees in electrical engineering from Hosei University, Tokyo, Japan, in
1992, 1994, and 1997, respectively. In 1997, he
was a research student at the Tokyo Metropolitan University. Since 1998, he has been with the
Shibaura Institute of Technology, Tokyo, Japan,
as an Associate Professor. His research interests
include power system control. Dr Fujita is a member of the Society of Instrument and Control Engineers (SICE) of Japan and the
IEEE.
References
(1) IEEE-SA-Standards-Board. IEEE STANDARD for Interconnecting
Distributed Resources with Electric Power Systems, IEEEStd 1547
Standard, 2003.
(2) Xu Wilsun S, Mauch K. An assessment of distributed generation
islanding detection methods and issues for Canada. Natural Resources
Canada, Technical Report, 2004.
(3) Wrinch MC. Negative sequence impedance measurement for distributed generator islanding detection. Ph.D. dissertation, University
of British Columbia, 2008.
Complicated implementation
Difcult implementation
Neutralize by interferences from grids
Large NDZ with high Q
Malfunction by other transients
Large NDZ with high Q
Depend on harmonic available levels
Facilitates an autonomous operation
Small NDZ, simple implementation
Sensitive threshold setting
249
IEEJ Trans 7: 240250 (2012)

189
eBook for You
Active method
N. D. TUYEN ET AL.
Masakatsu Nomura (Member) received the B.E. degree in electronics engineering and the M.S. degree in electrical engineering, both from Nagoya University,
Nagoya, Japan, in 1975 and 1977,respectively,
and the Ph.D. degree from Meiji University,
Tokyo, Japan, in 2009. Since 1977, he has been
with Meidensha Corporation, where he is currently working in the New Product Development
Group, in Tokyo. His research interests include power electronics,
control, and factory automation. Dr Nomura is a member of the
IEEE and the Japan Society for Simulation Technology.
eBook for You
Toshihisa Funabashi (Member) graduated in March 1975

from the Department of Electrical Engineering,
Nagoya University, Japan. He received the Ph.D.
degree from Doshisha University, Kyoto, Japan,
in March 2000. He joined Meidensha Corporation in April 1975 and was engaged in research
and development on power system simulations
and integration of distributed energy resources
in power networks. Currently, he is a Senior Fellow of the Power
Systems Solutions Engineering Division. Dr. Funabashi is a chartered engineer in the UK, a senior member of the IEEE, and a
member of the IET.
250
IEEJ Trans 7: 240250 (2012)

190
Extended Summary
pp.501509
Using Simulink Simulation to Evaluate Load Following Characteristics

of SOFC Generator with Heat Exchanger Considering Heat Balance
Nguyen Duc Tuyen Non-member (Shibaura Institute of Technology, m108423@shibaura-it.ac.jp)
Goro Fujita Member (Shibaura Institute of Technology)
Ryuichi Yokoyama Senior Member (Waseda University)
Kaoru Koyanagi Senior Member (Waseda University)
Toshihisa Funabashi Senior Member (Meidensha Corporation)
Masakatsu Nomura Member (Meidensha Corporation)
This paper proposes a transient dynamic model of Solid Oxide

Fuel Cell (SOFC) Power Unit which is supplied directly by natural
gas (CH4 ). Figure 1 shows the block diagram of a dynamic model
for a 70-kW SOFC developed in Simulink, based on the electrochemical and thermodynamic characteristics. The output voltage
depends on conditions including fuel ow, air ow, partial pressures, operating temperature and load current. First, load following ability can be achieved by using fuel and air feedback control.
In order to calculate the output voltage, this model point out the
method to calculate the partial pressure of all SOFC species. The
ideal voltage output is calculated by Nernsts equation as following
to accurately predict the reversible potential for single FC:
2
RT PCH4 PO2 T
ln 2
Ecell = E0 kE (T 298) +
E (1)
ne F
PH2 O PCO2 0
eBook for You
Keywords: SOFC, dynamic simulation
Fig. 2. Heat balance inside SOFC power unit with HX

included
and 1 atm pressure. Ecell calculated from Eq. (1) is the open circuit voltage of the fuel cell. However, when the fuel cell is under load, its output voltage is less than Ecell due to activation loss
(act = act,1 + act,2 ), ohmic loss (ohmic ), and concentration loss
(con ). The layers can store electrical energy and behave as a supercapacitor. While the double-layer charging eect is integrated into
the modeling, the fuel cell output voltage now is:
where E0 is the standard reference potential at standard state 298 [K]
Vcell = N0 Ecell VC ohmic act,1 (2)

Energy balance inside SOFC power unit incluled Heat Exchangers (HXs) is the key point of this paper. Both counter ow and
parallel ow types of heat exchangers are selected for comparison.
The operating temperature is the dominant parameter of SOFC. Afterthat, this model can be used to point out how to control operating
temperature by using the excess air into the fuel cell. Excess air is
sent to take sensible heat of hot air and to stabilize the temperature.
The amount of excess air used is determined by the actual operating
temperature.
Fig. 1. Outline of SOFC model
4
191
Paper
Using Simulink Simulation to Evaluate Load Following Characteristics

of SOFC Generator with Heat Exchanger Considering Heat Balance
Non-member
Member
Member
Member
Member
Member
That ever increasing electricity consumption, progress in power deregulation, and rising public awareness for environment have created more interest in fuel cell distributed generation. Among dierent types of fuel cells, solid oxide
fuel cells (SOFCs) manifest themselves as great potential applications due to many advantages such as low emission,
high eciency, and high power rating. On the other hand, SOFC systems are benecial because they can convert fuel
such as natural gas (almost CH4 ) which is supplied by widespread systems in many countries into electricity eciently
using internal reforming. In facts, the load demand changes exibly and fuel cell life time decreases by rapid thermal
change. Its lifetime may be extended by maintaining in appropriate temperature. Therefore, it is important to acquire
the load following performance as well as control of operation temperature. This paper addresses components of the
simple SOFC power unit model with heat exchanger (HX) included. Typical dynamical submodels are used to follow
the variation of load demand at a local location that considers temperature characteristics using the Matlab-SIMULINK
program.
Keywords: SOFC, dynamic simulation
1.
the fuel and air directly, the system eciency therefore can
be improved (1) . And overall heat balance inside SOFC power
unit eecting on operating temperature afterward on output
voltage will be pointed out.
Introduction
Depending upon the application, dierent models are

available in the open literature and there are large dierences
in the level of details in the models presented. Research work
of SOFCs modeling has been begun since 1980s and there
have been a lot of models developed so far. Today modeling
research is pursued in both detailed single SOFC modeling
and system level stack modeling. Nevertheless, lumped models still continue to attract the attention of researchers due to
their simplicity and short calculation time.
A transient dynamic model of SOFC power unit will be
proposed in this paper. Electrochemical and thermal simulations of a SOFC power unit reported in all reference will
be used to identify the key parameters of this SOFC system
from a single cell to a N0 single fuel cell connected in series.
The cells terminal voltage during a load change will be discussed. Also, the double-layer charging eect is taken into
account which is represented by a capacitor. SOFC systems
in particular involve recuperation of heat due to the high gas
inlet temperatures required and the high amount of heat in
the exhaust. And, waste heat recovery used for preheating
2.
Figure 1 shows the block diagram of a dynamic model for

a 70-kW SOFC developed in Simulink, based on the electrochemical and thermodynamic characteristics of the fuel cell.
The output voltage of the fuel cell depends on conditions including fuel ow, air ow, partial pressures, operating temperature, load current.
2.1 Partial Pressure Model
This model calculates
Power System Lab., Shibaura Institute of Technology

3-7-5, Toyosu, Koto-ku, Tokyo 135-8548
Waseda University
3-4-1, Okubo, Shinjuku-ku, Tokyo 169-8555
Meidensha Corporation
2-1-1, Ohsaki, Shinagawa-ku, Tokyo 141-6029
c 2010 The Institute of Electrical Engineers of Japan.

Dynamic Model of SOFC
Fig. 1. Outline of SOFC model

501
192
eBook for You
Nguyen Duc Tuyen

Goro Fujita
Ryuichi Yokoyama
Kaoru Koyanagi
Toshihisa Funabashi
Masakatsu Nomura
the partial pressure of CH4 , O2 , H2 O and CO2 in Eq. (5). The

partial pressures p j of the 4 components j are related to theirs
molar ow. They maybe determined through the equations (2) :
1/K j
qin qout
j a.Kr .i f c (1)
1 + js j
V = N0 [Ecell act ohmic con ] (7)
where K j = q j /p j ; a = 1 or 2; j = Van(ca) /(K j RT );

Kr = N0 /(ne F) (CH4 = 26.1; H2 O = 78.3; O2 = 6.74;
CO2 = 2.91) and s is complex Laplace variable in complex
Laplace domain.
2.2 Load Following Fuel Submodel
This model
calculates the input fuel ow to meet the optimal fuel
utilization (2) . The fuel utilization (U f = qrCH4 /qin
CH4 =
in
(N0 i f c /ne F)/qin
=
(K
i
)/q
)
is
assumed
between
r
f
c
CH4
CH4
U min
= 80% and U max
= 90%, the demand current of the
f
f
SOFC system can be restricted in the range given by:
qin
CH4
U min
f
2Kr
iref cf qin
CH4
2.4 Voltage Loss Model

2.4.1 Activation loss
Activation voltage loss is

caused by an activation energy barrier that must be overcome
before the chemical reaction occurs.
act,2 = act,an + act,ca
i
i
2RT
sinh1
+ sinh1
=
ne F
2i0,an
2i0,ca
= Ract i (8)
Actually, the operating temperature change eects on i0,an
and i0,ca . However in this simulation, this inuence is so
small that can be neglected (4) . The activation loss of an SOFC
is known to be less than the value given by Eq. (8). Therefore, a constant and a temperature-dependent term can also
be added to Eq. (8) for activation voltage drop computation
of SOFC as follows (5) :
U max
f
2Kr
re f
max
imin
f c i f c i f c (2)
The value if c = Pdemand /V f c is used to control fuel ow in

the fuel processor by calculating the qin
CH4 in order to keep
the fuel utilization in constant range. For a certain input
species ow, the demand current of the SOFC system can
be restricted in the range:
max
(if c > imax

ifc
fc )
min
re f
min
ifc
(i f c < i f c ) (3)
ifc =
i , otherwise
fc
Tact = 1 + 2 T + iRact = act,1 + act,2 (9)

act,1 = 1 + 2 T is the part of activation drop aected only
by the fuel cell internal temperature, while act,2 = iRact is
both current and temperature dependent (1 = 4.e4 [V], 2 =
18.e7 [V/K]).
2.4.2 Ohmic overpotential
Ohmic overpotential
occurs because of the resistance to the ow of ions in the
ionic conductors and the resistance to electrons through the
electronic conductors. The main contribution to the ohmic
polarization is from the transport resistance of ion O2 in the
electrolyte. The resistance of electrolyte strongly depends on
the temperature, and its eect cannot be ignored. The resistances of other parts are assumed to be constant because of
the weak dependence on temperature and their contribution
to the total voltage drop is small (6) . Hence,
The optimal utilization factor U opt

is assumed to be 85%,
f
allowing the control of the input ow by the measure of the
output current, so that:
opt
qin
CH4
Kr /U f
1 + f .s
.iref cf (4)
where f = 0.5 is fuel ow lagging time (Again, s is complex

Laplace variable in complex Laplace domain).
2.3 Electrochemical Model
The nal chemical reaction that results from internal reforming (See the Appendix
for more details) inside SOFC stack is as following.
Rohmic = Re0 e10100((1/T )(1/1273)) + Ra + Rc + Ri

(10)
ohmic = Rohmic i (11)
CH4 + 2O2 = 2H2 O + CO2 (5)
2.4.3 Concentration overpotential

During the reaction process, concentration gradients can be formed due to
mass diusions from the ow channels to the reaction sites.
At high current densities, slow transportation of reactants
(products) to (from) the reaction sites is the main reason for
the concentration voltage drop. The concentration overpotential in the fuel cell is dened entirely empirically as:
The voltage generation of the state of the ideal is calculated

by Nernsts equation as follow to predict quite accurately the
reversible potential for single FC (3) :
2
RT pCH4 pO2
ln
Ecell = E0 kE (T 298) +
ne F p2H O pCO2
2
(6)
con = m.en.i = Rcon i (12)

(m = 3 105 [V], n = 8 103 [mA1 .cm2 ])
where E0 is the standard reference potential at standard state

298 [K] and 1 atm pressure. kE = 0.00012 [V/K] is coecient
representing ideal voltage depends on temperature.
Ecell calculated from Eq. (6) is the open circuit voltage of
the fuel cell. However, when the fuel cell is under load, its
output voltage is less than Ecell due to activation/polarization
This equation is entirely empirical which received much

interest lately after many researches nding out accurate
equations, and yields an equation that ts the results very
well (7) .
502

193
eBook for You
pj =
loss (act ), ohmic loss (ohmic ) and concentration/diusion

loss (con ).
The output voltage of the fuel cell stack including N0 individual single fuel cell connected in series can be obtained
as:
Load Following and Heat Balance of SOFC
methodology (8) is used to model the pre-heaters in CH4 SOFC system, which makes use of two non-dimensional
groups: and NTU.
By comparing two heat capacity of hot and cold streams,
the lower and higher values are assigned as Cmin and Cmax , respectively. The ratio of heat capacity rates is then available.
Thus, Cr = Cmin /Cmax and NT U = UA/Cmin ; where, U is
overall heat transfer coecient, which is dened largely by
the system and in many cases it proves to be insensitive to the
operating conditions of the system. With our simulation, we
take U to be a constant value and U = 0.5 [W/(cm2 .K)] with
high pressure gas and A = 20000 [cm2 ] is total heat transfer
area.
The value is dened as the ratio of actual heat transfer
rate (q), and the maximum possible heat transfer rate between
the 2 streams (qmax ): = q/qmax , where the theoretical maximum heat transfer is:
Fig. 2. The charge double layer at the surface of a fuel

cell cathode (7)
2.4.4 Double-layer charging eect

In an SOFC, the
two electrodes are separated by the electrolyte (Fig. 2), and
two boundary layers are formed, e.g., anode-electrolyte layer
and electrolyte-cathode layer. These layers can be charged
by polarization eect, known as the electrochemical doublelayer charging eect, during normal fuel cell operation. The
layers can store electrical energy and behave like a supercapacitor. The model for SOFC considering this eect can
be described by the equivalent circuit shown in Fig. 3 (7) .
C is the equivalent capacitor due to the double-layer charging eect (C = 2F). The voltage across capacitor, VC is:
VC = i C
dVC
(Ract + Rcon ) (13)
dt
qmax = Cmin (T h,in T c,in ) (15)
The double-layer charging eect is integrated into the

modeling, by using VC instead of act,2 and con calculate
Vcell . The fuel cell output voltage now turns out to be:
where the values (T h,in ; T c,in ) are simply the temperature difference between the 2 inlet ports.
We have the NT U relation by these equations with
counter ow and parallel ow, respectively:
1 e[NT U(1Cr )]
ct =
1 Cr e[NT U(1Cr )]
(16)
[NT U(1+Cr )]
1e
pr =
1 + Cr
Vcell = N0 Ecell VC ohmic act,1 (14)

2.5 Heat Exchanger Model
2.5.1 Fuel and air preheaters model
In this model,
both counter ow and parallel ow types of heat exchangers
are selected for comparison. There are 2 HXs. The rst one,
HX1 is used to increse temperature of air and the second one,
HX2 is used to increse temperature of fuel. To determine the
outlet temperatures of the heat exchangers, which vary with
the inlet conditions, the heat capacity rates of the cold and
hot gas streams are calculated:
( 1 ) HX1-air preheater
Therefore, once eectiveness is calculated, based on the

energy balance, the exit temperature of the hot and cold gas
streams from the heat exchanger are:
q
q
; T c,out = T c,in +
(17)
T h,out = T h,in
Ch
Cc
2.5.2 SOFC exhaust temperature
Relatively high
amount of input gas ows are not used and they pass though
SOFC without reaction. With the reactants, therefore it can
be considered their output temperates are operating temperature. However, non-reactants do not have the operating temperature when they come out from SOFC stack. Assuming
that the SOFC conguration as a Heat Exchanger for these
gases, it is similar to calculate their output temperature as in
case of HX in the above section. Fig. 4 is the concept of this
idea.
Heat capacity rates of the cold and hot gas streams are calculated:
78 in
q CN
21 O2 2
Ch [CH4 +N2 +O2 +CO2 +H2 O]
1 U opt
f
qin + 78 qin CN
=
opt
CH4
21 O2 2
Uf
1 U opt
f
out
out
qin
+
O2 C O2 + qH2 O C H2 O + qCO2 C CO2
opt
Uf
Cc [Air] = qin
O2 C O2 +
( 2 ) HX2-fuel preheater
Cc [CH4 ] = qin
CH4 C CH4
Ch [HX2 ] = Ch [HX1 ]
out
Ch [CO2 + H2 O] = qout
CO2 C CO2 + qH2 O C H2 O
in
out
Cc [CH4 + N2 + O2 ] = qout
CH4 C CH4 + qN2 C N2 + qO2 C O2
The eectiveness () and number of transfer units (NT U)

B130 5 2010
503
194
eBook for You
Fig. 3. Equivalent electrical circuit of the double-layer

charging eect inside an SOFC
Fig. 4. Exhaust temperature calculation concept
Following the previous section, in which subscript 1 means

in and subscript 2 means out for discrimination, these
values can be calculated:
Cmin
;
Cmax
qmax = Cmin (T h,1 T c,1 )
Assuming the SOFC conguration as parallel one, the eciency can be calculated:
=
Fig. 5. Heat balance inside SOFC power unit with HX

included
1 e[NT U(1+Cr )]
(18)
1 + Cr
gas streams, and the conductive heat transfer occurs in the

solid and/or porous structures. The electricity generated are
the basic heat components in a fuel cell as shown in Fig. 5.
The general energy balance states that the enthalpy of the reactants (total energy produced by the reaction) and gas sensitive after HXs entering the fuel cell equals the gas sensitive
heat of the products leaving the fuel cell stack plus the sum
the power output, the rate of heat loss to the surroundings and
the rest to maintain the operating temperature. The heat loss
depends on the dierence between operating temperature of
fuel cell stack and ambient temperature which is represented
by Kh .
Based on the diagram in Fig. 5, the temperature equation
and look at the ow of energy per unit time as a representable
expression which is used as dynamic energy balance equation:
and
q = qmax (19)
Hence,
T h,2 = T h,1
q
;
Ch
T c,2 = T c,1 +
q
(20)
Cc
where
T h,1 = T (operating temperature)
T c,1 =
HX2 out
in
qCH4 + T OHX1
(qout
T CH
O 2 + qN 2 )
4
2
out
in
qout
CH4 + qO2 + qN2
S OFC
T h,2 = T CO
= T HS 2OFC
O
2
GS HHX + H i f c V
dT
+ Kh (T T 0 ) (22)
= GS HS OFC + R f c
dt
The dynamic operating temperature can be derived from
Eq. (22) shown in the following Equation:
(reactants output temperature)

T c,2 = T S OFCHX
(non-reactants output temperature)
The isothermal temperature of SOFC exhaust which will
be fed into HX therefore can be calculated as:
T = T ini
T S OFCexhaust
=
T HS 2OFC
O
out
out
out
in
qout
H2 O +qCO2 + T S OFCHX qCH4 +qO2 +qN2
out
out
out
in
qout
H2 O + qCO2 + qCH4 + qO2 + qN2
(21)
t
0
GS HHX + H i f c V
dt
Kh (T T 0 ) GS HS OFC
(23)
The parameters appear in energy balance model are dened

here.
2.6 Thermal Dynamic Model

Thermodynamics is
the study of energy changing from one state to another. The
predictions that can be made using thermodynamic equations
are essential for understanding and modeling SOFC performance since SOFCs transform chemical energy into electrical energy. The temperature and voltage drop aect the life
of material, the model should reect this.
2.6.1 Energy balance
The rst step in determining
the heat distribution in a stack is to perform energy balances
on the system. The total energy balance around the fuel cell
is based upon the power produced, the reactions, and the heat
loss that occurs in a fuel cell. Heat losses include the convective heat transfer occurs between the solid surface and the
G: Thermal generation (Gibbs energy) [J/mol]

(Electrical energy can retrieve a theoretical
maximum)
H: Enthalpy of reaction [J/mol]
V: Terminal voltage of SOFC power unit [V]
i f c : Output current of SOFC power unit [A]
i f c .V: Output power [W]
H i f c V: Thermal Energy [J/mol]
R f c : Heat capacity of SOFC power unit [J/K]. That
is the required energy for increasing 1 [K] for
SOFC power Unit.
R f c [J/K] = C p [J/kgK]. M [kg]
504
1
Rfc

195
eBook for You
Cr =
C p : Specic heat of SOFC power Units material (14)

M: Mass of SOFC power Unit
change. The conceptual model of Eq. (24) is shown in Fig. 6

which is derived by calculating the dissipation coecient of
heat energy from the stack number of moles of each gas
equivalent reaction.
2.6.3 Operating temperature control
The operating temperature control method by using the excess air
(O2 +N2 ) into the fuel cell stack is shown in Fig. 7. Excess air
is sent to take sensible heat of hot air, stabilize the temperature. The amount of excess air(qin
O2 ex ) used is determined
by the actual temperature which is adjusted by the feedback
control as Eq. (26).
Kh : Heat Radiation Coecient [W/K]

T 0 : Ambient temperature [K]
T : Operating temperature [K]
Kh (T T 0 ) is considered to be the energy dissipation. [W]
T ini : The initial temperature [K]
GS HHX : Gas sensitive heat after HX [J/mol]
GS HS OFC : Gas sensitive heat after SOFC stack
[J/mol]
in
qin
O2 ex = qO2 1 + Kair
3.
2.6.2 Heat radiation coecient model
The calculation of the heat radiation coecient Kh which appear in

Eq. (22) is necessary to consider heat characteristics. The
heat radiation coecient is calculated in this model. The
sum of gas sensistive heat from HXs and generated energy
except for converter into electricity is assumed to contribute
in temperature uctuation of exhaust gas and the machine itself. Gibbs free energy is theoretically convertible energy to
electricity. Therefore, remained energy that takes out electric
energy from reactive energy can be separated to gas sensitive energy and heat radiation energy. In the steady-state, the
Eq. (22) can be rewritten:
Therefore, the heat radiation coecient can be given:

GS HHX + H i f c V GS HS OFC
(25)
T ini T 0
Equation (25) is used to calculate Kh and it is considered

that this value keeps constant upon the operating temperature
Fig. 6. Concept of heat loss coecient calculation

model
Fig. 7. Concept of control temperature by excess air

B130 5 2010
Results and Discussion
This section consists of the simulation results obtained in

Simulink for the system shown in Fig. 1 and afterward are the
comments. There are fundamental time delays in this simulation that are pointed out at Appendix.
3.1 Load Following Ability
In this study, SOFC
70 [kW] power unit operating at 1273 [K] temperature works
at normally stable load, and the output command 100 [kW] is
suddenly required. Observing SOFC load following ability,
we need to consider the change of operating temperature.
Figure 8 is diagram of the output of current, voltage, power
and operating temperature in short period test time, 200 [s].
The load suddenly changes at 100 [s] instance. This case considers robust following load change in small time scale. The
current increasing makes the power response change instantaneously. The output power afterward takes 30 [s] to follow
the increasing of load. That is equivalent to 1 [kW/s] velocity.
The slow response of the fuel cell is due to the slow and gradual change in the fuel ow and the chemical reaction which
is proportional to the stack current.
Figure 9 is diagram of the output of current, voltage, power
and operating temperature in long period test time, 10000 [s].
The framework of load change for several time is assumed at
[0, 500, 1500, 2500, 3500, 4500] instance and the respective power are [70 100 70 100 70 100][kW]. This case considers the operating temperature. The temperature respond
velocity with increase of load is 10 [K]/100 [s]. The nal
stable operating temperature is about 1250 [K] obtained at
the 6000 [s]-instance while the voltage and current output is
about 437 [V], 220 [A].
3.2 Heat Exchangers
3.2.1 Comparison between SOFC with and without
HX
The simulation results are showed in Fig. 10. Once
we used HXs, according to the compared simulation results,
the operating temperature will be reduced about 50 [K]. This
is one of the advantages while using HXs. The energy eciency calculation also show that the SOFC with HXs model
has 85% eciency compared to 55% of SOFC without HXs
model.
3.2.2 Comparison between counter and parallel HX
The clear results are pointed out by Fig. 11. The exhaust
temperature of the Counter-HX into atmosphere is lower than
that of the Parallel HX mean the higher energy eciency.
However, the operating temperature of SOFC in case using
the Counter-HX is higher than using the Parallel-HX. The
GS HHX + H i f c V = GS HS OFC + Kh (T T 0 )
(24)
Kh =
T 1273
(26)
1273
505
196
eBook for You
In this model, because the mass of SOFC power unit system

is assumed as 100 kg of iron, the value, Rfc = 46000 [J/K] is
used.
Fig. 8. Dynamics model in small timescale-200 [s]
Fig. 11. HX operating temperature consideration between Counter and parallel Conguration
Fig. 9. Dynamics model in large timescale-10000 [s]
Fig. 12. Current comparison between model with and

without excess air
method. The output power at the various stages makes

change in output voltage, current, operating temperature with
and without excess air shown in the same gure for comparison. Figure 12 represents that the output current almost does
not change while using excess air compared to the case without excess air. Figures 13 and 14 pay attention to the temperature and voltage ranges. The SOFC stack material degradation will be suered by these ranges which might reduces life
time of SOFC stack.
Results express that the operating temperatures and voltage variation range with excess air is smaller over time. The
temperature variation reducing can expectedly prevent the
short fuel cell stack life time because of the rapid changes
in operating temperature. Because the output voltage depends largely on temperatures, when operating temperature
variation is kept under control, the output voltage is therefore
more stable. The results on Fig. 14 show that the variation is
smaller compared to the case without excess air and the output voltage is also higher while using excess air. From this
result, one idea emerges that if we fed SOFC stack with as
much air ow rate as it needs, the variation range would be
smallest. This is actually impractical because of compressor
congurations as well as the electrode pressures. Thanks to
this results, in addition to the real SOFC power unit conguration, the manufacture can decide the value of appropriate
Kair .
Fig. 10. Heat temperature comparison between SOFC

with and without HX
higher operating temperatures make stack materials work in

severer condition. Based on this conclusion, the manufactures could decide which conguration is suitable for their
real model.
3.3 Operating Temperature Control by Excess Air
The amount of the input air is adjusted by feedback control
of the temperature as a measured value. The amount of air
supply has limitation because of pressure limitation as well
as the power consuming of air compressor.
Following is the evaluation for using above excess air
506

197
eBook for You
(1)
(2)
(3)
(4)
(5)
Fig. 13. Temperature comparison between model with

and without excess air
(6)
(7)
(8)
(9)
( 10 )
( 11 )
( 12 )
( 13 )
Fig. 14. Voltage comparison between model with and

without excess air
4.
( 14 )
( 15 )
Conclusion
Appendix
This study discussed the simulation of SOFC dynamic

model running on natural gas (CH4 ). The load following
ability characteristic, the eect of heat exchangers and the
control of operating temperature of the model and has been
studied. The characteristics of the fuel cell (voltage, current
and power) have a slower gradual change at the instant of step
change. Worthy remarks of this paper include:
( 1 ) Calculate heat balance inside SOFC power unit that
eects on the operating temperature and therefore on the output voltage. The N2 gas is also involved in consideration.
( 2 ) Add the heat exchangers into SOFC power unit and
calculate the gas ows temperature attaching to heat balance
to compact SOFC model. Evaluate the dierent heat characteristic of two popular congurations of HX, and therefore
increase system energy eciency.
( 3 ) Evaluate the load following ability of SOFC power
unit by using feedback control the fuel and air ow.
( 4 ) Control the operating temperature by excess air.
Certainly, further modication is required for improvement
of accuracy which is applicable to practices. Accumulating
experience in real apparatus is also expected for development
of SOFC power generation systems.
(Manuscript received July 30, 2009,
revised Jan. 8, 2010)
B130 5 2010
S.H. Chan, C.F. Low, and O.L. Ding: Energy and exergy analysis of simple
solid oxide fuel cell power systems, J. Power Sources, Vol.103, pp.188200
(2002)
F. Jurado: Modeling SOFC plants on the distribution system using identication algorithms, J. Power Sources, Vol.129, pp.205215 (2004)
D. Sanchez, R. Chacartegui, A. Munoz, and T. Sanchez: On the eect of
methane internal reforming modelling in solid oxide fuel cells, J. Power
Sources, Vol.33, pp.18341844 (2008)
S.H. Chan, K.A. Khor, and Z.T. Xia: A complete polarization model of a
solid oxide fuel cell and its sensitivity to the change of cell component thickness, J. Power Sources, Vol.93, pp.130140 (2001)
P.R. Pathapati, X. Xue, and J. Tang: A new dynamic model for predicting
transient phenomena in a PEM fuel cell system, J. Power Sources, Vol.30,
pp.122 (2005)
S. Nagata, A. Momma, and T. Kato: Numerical analysis of output characteristics of tubular SOFC with internal reformer, J. Power Sources, Vol.101,
pp.6071 (2001)
J. Larminie and A. Dicks: Fuel Cell Systems Explained, 2nd ed., Wiley, England (2003)
T. Kuppan: Heat Exchanger Design Handbook, Illustrated ed., Marcel
Dekker, Inc., New York (2000)
M. Uzunoglu and M.S. Alam: Dynamic Modeling, Design, and Simulation
of a Combined PEM Fuel Cell and Ultracapacitor System for Stand-Alone,
IEEE Trans. Energy Conversion, Vol.21, No.3 (2006)
C. Wang and M.H. Nehrir: A Physically Based Dynamic Model for Solid
Oxide Fuel Cells, IEEE Trans. Energy Conversion, Vol.22, No.4, pp.887
897 (2007)
P. Piroonlerkgul, W. Kiatkittipong, A. Arpornwichanop, A. Soottitantawat,
W. Wiyaratn, N. Laosiripojana, A.A. Adesina, and S. Assabumrungrat:
Integration of solid oxide fuel cell and palladium membranereactor:
Technical and economic analysis, Int. J. Hydrogen Engergy, Vol.34, No.9,
pp.38943907 (2009)
EG and G Technical Services, Inc.: Fuel Cell Handbook, 7th ed., Morgantown, West Virginia, USA: U.S. Department of Energy (2004)
C. Wang, M.H. Nehrir, and S.R. Shaw: Dynamic Models and Model Validation for PEM Fuel Cells Using Electrical Circuits, IEEE Tran. Energy
Conversion, Vol.20, No.2, pp.442451 (2005)
http://www.engineeringtoolbox.com/specic-heat-metals-d 152.html
P. Aguiar, D. Chadwick, and L. Kershenbaum: Modelling of an indirect
internal reforming solid oxide fuel cell, Chemical Engineering Science,
Vol.57, pp.16651677 (2002)
1.
Nomenclature
v or V: Voltage, V
E: Voltage potential, V
P: Power, kW
Pdemand : Required power of load, kW
R: Resistance,
R = 8314: Universal gas constant, J/(kmol.K)
R f c : Heat capacity of SOFC power unit, J/K
C: Capacitor or Heat capacity rate
i: Output current density, A/cm2
T : Operating temperature, K
T 0 = 293: Ambient temperature, K
T ini = 1273: Initial temperature, K
p: Pressure, atm
q: Flow rate, kmol/s or heat transfer rate
F = 96487: Faraday constant, C/mol
N0 = 384: Number of cells connected in series
U f : Fuel Utilization, %
Kr : Constant, mol/C
ne = 8: Number of electrons are transfer between
layers respect to 1 mol CH4
K j = q j /p j : Valve molar constant, kmol/(s.atm)
507
198
eBook for You
References
Kair = 100: Coecient of excess air

i f c : Output current of SOFC power unit, A
iref cf : Reference current regulate input fuel ow rate,
A
qinj : Flow rate of species j fed into stack, kmol/s
qout
j : Flow rate of species j exhaust from stack,
kmol/s
qrj : Mole ow of species j take part in reactions,
kmol/s
i0,an = 0.53: The apparentcurrent exchange density of anode, A/cm2
i0,ca = 0.2: The apparentcurrent exchange density
of cathode, A/cm2
Ract : Equivalent activation resistance, /cm2
Rohmic : Equivalent ohmic resistance, /cm2
Rcon : Equivalent concentration resistance, /cm2
Ra : Resistance of anode, /cm2
Re0 = 0.0248: Resistance of electrolyte at 1273 [K],
/cm2
Rc : Resistance of cathode, /cm2
Ri : Resistance of interconnector, /cm2
HX2
T CH
: Temp. of CH4 after HX2, K
4
T : Temperature
2.
CH4 = 26.1: Response time for methane ow

H2 O = 78.3: Response time for water ow
O2 = 6.74: Response time for oxygen ow
CO2 = 2.91: Response time for carbon dioxide ow
e = 0.8: Electrical response time
f = 0.05: Fuel ow lagging time
f u = 5: Fuel processor response time
comp = 1: Air processor response time
Reformer Model
The model considers internal reformer model for simplifying SOFC power unit and reducing cost, a reaction when
steam and oxygen generated from methane. The formulas
show the internal reforming reaction process:
3.
CH4 +H2 O 3H2 +CO (A1)

1 [mol] CH4 generates 3 [mol] H2 and 1 [mol]CO. These
products then react with O2 as the following equations:
T OHX1
: Temp. of O2 after HX1, K
2
3
3H2 + O2 3H2 O (A2)
2
1
CO+ O2 CO2 (A3)
2
from
SOFC, K
Combining these three Eqs. (A1), (A2) and (A3) match the
combustion reaction of methane:
Greek letters
j = Van(ca) /(K j RT ): Time constant in reaction, s
Van(ca) : Volume of anode or cathode, m3
1 , 2 : Empirical coecients of activation loss
: Eectiveness, %
CH4 +H2 O + 2O2 3H2 O + CO2

Or CH4 +2O2 2H2 O + CO2 (A4)
Finally, the Eq. (A4) is equivalent to Eq. (5). Thus, practically internal reforming model is used to model the reaction
of methane combustion. This is the key advantage of SOFCs
high operating temperature. This detail model is explained in
Ref. (15).
Subscripts
j: CH4 , O2 , H2 O, CO2
f: fuel
fc: fuel cell
e: electron
act: activation
con: concentration
c: cool
h: heat
ct: counter
pr: parallel
SOFC: Solid Oxide Fuel Cell
HX: Heat Exchanger
GSH: Gas Sensitive Heat
ini: initial
Nguyen Duc Tuyen (Non-member) received the B.S. in electrical engineering from Hanoi University of Technology, in
2006 and M.S. from Shibaura Institute of Technology,
in 2009. He is now the Ph.D. Candidate at Shibaura
Institute of Technology, Japan. Currently, he is a Faculty Member in the Department of Electrical Power
System, Hanoi University of Technology, Vietnam.
His current research interests include modeling and
control of distribution power systems and dispersed
power generation systems.
Goro Fujita (Member) was born in January 1970. He received the

B.E., M.E. and Ph.D. degrees in electrical engineering from Hosei University, Tokyo,Japan in 1992,
1994 and 1997, respectively. In 1997, he was a
research student at Tokyo Metropolitan University.
Since 1998, he has been served at Shibaura Institute
of Technology in Tokyo, Japan. At present, he is an
associate professor. He is interested in power system
control. He is a member of the Society of Instrument
and Control Engineers (SICE) of Japan, the IEE of
Japan, and IEEE.
Superscripts
r: react
min: minimum
max: maximum
opt: optimal
re f : reference
in: input
out: output
508

199
eBook for You
S OFC
T CO
: Temp. of CO2 exhaust from SOFC, K
2
T HS 2OFC
O : Temp. of H2 O exhaust from SOFC, K
T S OFCHX : Temp. of non-reactants exhaust
Time delay

Ryuichi Yokoyama (Member) was born in August 1944. He received
the degrees of B.S., M.S., and Ph.D. in electrical engineering from Waseda University, Tokyo, Japan, in
1968, 1970, and 1973 respectively. He entered Mitsubishi Research Institute in 1974. From 1978 he
served at Tokyo Metropolitan University. In 2007, he
transferred to Waseda University. His eld of interests includes control and optimization of large-scale
systems and application of articial intelligence to
power systems. He is a member of the Society of
Instrument and Control Engineers (SICE) of Japan, the IEEJ, CIGRE, and
IEEE.
Toshihisa Funabashi (Member) graduated in March 1975 from the

Department of Electrical Engineering, Nagoya University, Japan. He received, in March 2000, a Doctor
degree from Doshisha University, Kyoto, Japan. He
joined Meidensha Corporation in April 1975 and has
engaged in research and development on power system simulations and integration of distributed energy
resources in the power network. Currently, he is a
Senior Fellow of the Power Systems Solutions Engineering Division. Dr. Funabashi is a chartered engineer in UK, a senior member of IEEE, a member of IET.
Masakatsu Nomura (Member) received the B.E. degree in electronics engineering and the M.S. degree in electrical engineering from Nagoya University, Nagoya, Japan, and
the Ph.D. degree from Meiji University, Tokyo, Japan
in 1975, 1977 and 2009 respectively. Since 1977, he
has been with Meidensha Corporation, where he is
currently working in the New Product Development
Group, Tokyo, Japan. His interests are power electronics, control and factory automation. Dr. Nomura
is a member of IEEE, IEEJ and Japan Society for
Simulation Technology.
eBook for You
Kaoru Koyanagi (Member) received a B.S. degree in Applied

Physics from Tokyo University of Education, Japan in
1971.and a Dr.(Eng) degree in Electrical Engineering
from Tokyo Metropolitan University, Japan in 2000.
From April, 1971 to June, 2008, he was with Toshiba
Corp. working on power system analytical engineering at the Heavy Apparatus and Engineering Lab.,
and with TEPCO SYSTEMS Corp. He is currently
with Waseda University as a visiting researcher, and
is involved with research and development in the eld
of distributed generations, micro-grid, recycle energy and their controls. He
is a Chartered Engineer in UK, and a senior member of IEEJ, IET and IEEE.
B130 5 2010
509
200

Cac Bai Bao Khoa Hoc 2007-2012 (Phan II CUNG CAP DIEN)

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Cac Bai Bao Khoa Hoc 2007-2012 (Phan II CUNG CAP DIEN)

Transféré par

Droits d'auteur :

Formats disponibles

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

eBook for You

B mn H thng in - i hc Bch Khoa H Ni

Tng hp cc bi bo khoa hc giai on 2007-2012

Fault Distribution Modeling Using Stochastic

Manuscript received August 2, 2005; revised December 5, 2006. This work

information about the voltage sag is mainly obtained by

0885-8977/$25.00 2007 IEEE

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

The fault rate mainly depends on fault position, fault type,

Fig. 1. Example of bivariate normal distribution.

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

tion because it is the most common stochastic model which has

weighted factors of the fault rates of

means and standard deviations of two

The line failure rate is normally expressed in the number of

The probability for a fault to occur at the fault position

and is large enough,

For the distribution system, geographically, if network nodes

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 1, JANUARY 2008

following approximation where and

are the coordinates of

Fault rate for the line segment

number of sags down to under

B mn H thng in - i hc Bch Khoa H Ni

Fig. 2. Mapping of the IEEE123-bus radial distribution test feeder.

V. RESULT DEMONSTRATION AND ANALYSIS

eBook for You

C. Development of Voltage Sag Indices

Tng hp cc bi bo khoa hc giai on 2007-2012

B. Evaluation of Inuences of the Fault Distribution Modeling

eBook for You

Table I. The test system with actual dimensions in feet is mapped

Tng hp cc bi bo khoa hc giai on 2007-2012

different deviation values of fault distribution

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 1, JANUARY 2008

[11] M. R. Qader, M. H. J. Bollen, and R. N. Allan, Stochastic prediction

B mn H thng in - i hc Bch Khoa H Ni

Seung-Il Moon (M93) received the B.S. degree

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012

Prediction of Voltage Sags in Distribution

Index Terms-- power quality, voltage sag, distribution

MONG power quality phenomena, the voltage sag

II. INDICES FOR VOLTAGE SAG ASSESSEMENT

Figure 1. ITI curve for susceptibility of computer equipment.

eBook for You

Abstract-- This paper presents a method of assessing a

Tng hp cc bi bo khoa hc giai on 2007-2012

REDICTION OF VOLTAGE SAG IN DISTRIBUTION SYSTEM

Figure 2. Model of voltage sag

This work uses the method of fault position [1] for

eBook for You

Tng hp cc bi bo khoa hc giai on 2007-2012