IEEE TRANSACTIONS ON POWER SYSTEMS , VOL. 21, NO. 2, MAY 2006 997
A New Approach for Substation Expansion Planning
Mohammad Sadegh Sepasian, Hossein Seifi, Senior Member, IEEE, Asghar Akbari Foroud, Seyyed Hadi Hosseini,
and Esmaeil Mohseni Kabir
Abstract Substation expansion planning (SEP), as a major to find the optimal solution while observing the losses to be less
module of power system planning modules, is addressed in this
than a prespecified value. Genetic algorithm (GA) is used in [10]
paper. For long-term transmission SEP, a new approach based on a
to optimize the problem, while load uncertainties are considered
genetic algorithm (GA) optimization tool is proposed, in which by
by a fuzzy-based technique. Reference [11] considers SEP as
defining appropriate objective function and various constraints,
a nonlinear optimization problem, while in [12], simulated an-
new as well as expansion of existing substation requirements are
determined. The proposed algorithm is successfully tested for the nealing is employed to find the solution.
Iranian Power Grid in 2011.
While in [2], [4], and [10], the optimum solutions are found
from a list of available candidates, [6] and [9] are able to auto-
Index Terms Genetic algorithm (GA), optimization, power
system planning, substation expansion planning (SEP). matically find the solutions. However, they could not be readily
applied to large-scale problems [13].
In [13], a heuristic algorithm is proposed in which the op-
I. INTRODUCTION
timum required substation capacities are initially found. There-
ITH electric power consumption growth, desired new
after, they are allocated by minimizing a linearized model of
transmission system elements are needed to overcome
active power losses.
W
the possible lack of adequacy problems so that with least costs,
In this paper, a new SEP procedure is proposed in which the
various operational constraints are met. In the so-called sub- mathematical clustering technique is used to find, initially, a list
station expansion planning (SEP), the problem is to determine
of feasible candidates by observing the limitations on substation
the required expansion capacity of the existing substations as
capacities, feeder capacities, and voltage regulations. Then, GA
well as the allocation and size of new substations together with
is used to solve an optimization problem in which the alloca-
the required availability times. The existing SEP algorithms are
tions and capacities of new substations as well as the expansion
normally based on several candidate substation buses. The user
requirements for the existing ones are determined. The numer-
should select the candidates, and an optimization algorithm will
ical results for the Iranian Power Grid are shown as a large-scale
choose the best solutions. If the candidates change, the solution
practical case.
may change.
Over the years, SEP has received attention in the literature.
II. PROBLEM FORMULATION
Reference [1] uses an integer programming approach to solve
While the optimization problem defined for power system
the problem, while [2] employs a dynamic programming algo-
planning may have different objective functions as well as var-
rithm. In [3], branch-and-bound dynamic programming is used
ious constraints [14] [16], in the SEP problem, the objective is
in which the optimum solution is found while observing all pos-
to determine a mixture of both expansion plans for the existing
sible expansion candidates as well as the cost of active power
substations and the need for new substations so that the loads
losses. In [4], the objective function is the cost of supplying the
are adequately supplied.
loads by multiplying the loads by their respective distances from
The loads to be supplied are widely geographically dis-
the substation. Reference [5] proposes an approach in which
tributed. For substation planning, the normal procedure is to
the problem is decomposed into two series problems in which
initially determine the distribution substation requirements
the first determines the optimal substation capacities, while in
and moving upward, to finally determine the transmission
the second, linear programming is employed to determine the
substation requirements. This approach, although accurate and
supplying feeders. In [6], mixed integer linear programming is
practical for short and midterm planning, may prove impractical
used, while in [7] and [8], linear programming is employed to
for long-term studies (say, five years onward) of transmission
tackle the problem. Reference [9] employs integer programming
substations, as the transmission owner (developer) may wish
to determine the possible allocation and size of the substations
Manuscript received August 4, 2005; revised December 5, 2005. Paper no.
(either new or expansions on existing) without involving in
TPWRS-00487-2005.
many details of the downward grids (subtransmission and
M. S. Sepasian is with Power and Water, Industry University, Tehran, Iran
(e-mail: [e-mail]). distribution). One way to overcome this problem is to propose
H. Seifi is with Tarbiat-Modares University, Tehran, Iran, and is also with Iran
an algorithm in which the geographically distributed loads
Power System Engineering Research Center (e-mail: [e-mail]).
are somehow assigned to transmission substations. Although
A. Akbari Foroud is with Semnan University, Semnan, Iran (e-mail: aak-
[e-mail]). this does not happen in practice, the final transmission alloca-
S. H. Hosseini is with Zanjan University, Zanjan, Iran (e-mail: siedhadi@hot-
tions and capacities can prove appropriate, provided various
mail.com).
constraints are properly observed. The assigning procedure
E. M. Kabir is with Tavanir Co., Tehran, Iran (e-mail: [e-mail]).
Digital Object Identifier 10.1109/TPWRS.2006.873406 is, however, crucial as an unsuitable procedure can result in
0885-8950/$20.00 © 2006 IEEE
998 IEEE TRANSACTIONS ON POWER SYSTEMS , VOL. 21, NO. 2, MAY 2006
improper solutions. In Section II-A, the system under study is a) : A load may be supplied by several nearby
described. The general optimization problem is presented in substations. The cost is dependent on the distance between the
Section II-B. A detailed algorithm is explained in Section III. load center and the substation as follows:
A. System Under Study
(2)
The existing Iranian Power Grid is made up of 400-, 230-,
132- and, 63-kV transmission and subtransmission elements.
The corresponding grid lengths are roughly 12 000, 23 000,
in which
16 000, and 35 000 km, respectively. The sum of MVA capacity
cost of all downward grids;
of the transmission substations (transmission to subtranmis-
cost of the feeder for supplying the th load by
sion) is in the order of 90 000 MVA. The summer peak of the
conductor (see Section III-B for details);
grid in 2004 has roughly been 29 000 MW.
distance between the th load and the th sub-
The transmission and subtransmission grids are highly inter-
station;
connected, managed by 16 regional utilities owned by a holder
S set of all selected substations;
company (Tavanir). The transmission grid (i.e., 400 and 230 kV)
L(j) set of all loads connected to the th substa-
planning is monitored by Tavanir, while for the subtransmission
tion.
grid, the task is performed by regional utilities.
b) : A major cost is the investment cost for all sub-
With rapid annual growth of 7% 9% electric consumption,
stations defined as
the grid is confronted by a challenging planning problem for the
years to come. The problem described is this paper addresses
SEP for large-scale transmission systems. It is specifically ap-
plied to 16 utilities of Iranian Power Grid for the summer elec-
tric power peak in 2011 without involving much in the details of
(3)
downward systems (subtransmission and distribution). A later
where
paper will address the corresponding network planning proce-
fixed cost for the th substation;
dure.
variable cost factor for the th substation;
capacity of a new substation ;
B. Optimization Problem
S set of all new substations and all expanded substa-
tions;
The SEP problem may be defined as an optimization proce-
SE set of all expanded substations;
dure in which all the investment costs as the well as operational
amortizing coefficient for the existing substation ;
costs have to be minimized while numerous constraints are met.
capacity of the existing substation .
The final solution should determine
c) : Obviously, the closer a substation is to an
" expansion capacity of any existing substation (provided
existing transmission grid, the more attractive it is, in terms of
feasible);
general costs. To consider this effect, a term is in-
" allocation and size of any new substation;
cluded in (1), as defined in (4) consisting of two terms, namely,
" investment costs.
a fixed cost for the right of way, tower, etc. (dependent on the
voltage level) and a variable cost (dominantly conductor cost)
In the following subsections, the objective functions as well
dependent on the line capacity. Therefore
as the constraints are described in detail.
1) Objective Functions: The aim is to supply the loads
through all transmission (transmission to subtransmission)
(4)
substations so that
(1)
where
fixed cost of the upward grid for energizing substa-
is minimized. is the overall plan cost. Other terms are
tion ;
described below. Note the following.
variable cost factor of the upward grid for ener-
" Each load is presented with its value (in MW) and geo- gizing substation ;
graphical characteristics (X and Y) in the horizon year. upward grid capacity for energizing substation ;
" Each load is assumed to be radially supplied by a trans- distance between substation to the nearest feeding
mission substation (to ignore the distribution and sub- point of HV transmission network.
transmission effects). Although not accurate in practical It is evident that does not show the exact distance for
terms, this approach facilitates the planning procedure practical situation. It somehow considers the upward grid in
with due attention to some practical considerations (see problem formulation so that substations far from the existing
the following sections). network are not justified.
SEPASIAN et al.: NEW APPROACH FOR SUBSTATION EXPANSION PLANNING 999
d) : The losses of the downward grid (as opera-
tional losses) should also be minimized. That is why is
introduced as
(5)
where
cost of the downward grid losses calculated as in
base year (for 30 years operational period);
conductor resistance of the feeder supplying the
Fig. 1. Normal crossover.
th load (for details, see Section III-B);
change of MVA of the th load with respect to
the least amount of extra load should be applied to
the base value (current year) (for details, see Sec-
this substation.
tion III-A);
c) Maximum and minimum installation capacities:
as before.
e) : Another term to be considered is the cost of
(9)
transformer losses (operational losses), denoted by and
defined as
In (9), refers to required reserve capacity for the th substa-
tion. For the existing substations, refers to the maximum
expansion capacity of the substation. For this type of substation,
(6)
may be set at a value less than its existing capacity (or even
zero). In that case, the substation may be de-rated (the extra ca-
pacity is considered as a benefit) or even totally removed, pro-
where
vided the optimization procedure finds it economical.
fixed cost of the th substation losses;
d) Standard capacities:
variable cost factor due to substation losses for full
(10)
load conditions;
cost of transformer losses calculated as in base year
which shows that the substations should be selected from a set
(for 30 years operational period);
of standard list (available from planning departments).
actual load of the th substation is MVA.
e) Thermal capacity of the upward lines: Similar to (7),
2) Constraints: The following constraints are considered in
(11) applies to upward transmission lines
the optimization problem:
(11)
" for the downward grid: thermal capacity of the feeder
for supplying the load (Section II-B2a) and with accept-
where is a set of candidate substations.
able voltage drop (Section II-B2b);
" for the substations: maximum and minimum installation
C. Solution Algorithm
capacities (Section II-B2c) as well as standard capacities
The following subsections briefly review candidate selection,
(Section II-B2d);
GA algorithm, and a refined algorithm.
" for the upward grid: thermal capacity of the upward
1) Candidate Selection: To narrow the search space, an ap-
transmission line (Section II-B2e).
proach is proposed by which a list of suitable candidates is se-
a) Thermal capacity of the downward feeder:
lected for the new substations. The approach is based on the
mathematical clustering concept. For some details, see the Ap-
(7)
pendix.
2) GA Algorithm: GA is a metaheuristic approach used for
where
optimization problems. Some chromosomes are initially gener-
set of loads;
ated. Two operators, namely, crossover and mutation, are there-
required capacity for supplying the th load.
after applied, and new chromosomes are then generated.
b) Voltage drop:
The decision variables considered in the chromosomes are in
fact the supplying buses as
(8)
(12)
where where
actual voltage drop for load ; th chromosome;
acceptable voltage drop; feeding bus number for supplying the th load.
factor for considering the fact that an already ex- Two crossover operators, namely, normal and mathematical,
isting substation has some voltage problems, and are applied as shown in Figs. 1 and 2, where is a random
1000 IEEE TRANSACTIONS ON POWER SYSTEMS , VOL. 21, NO. 2, MAY 2006
TABLE I
LOAD CENTERS IN SEP FOR 2011
Fig. 2. Mathematical crossover.
information that may still be useful in rejected chromosomes,
this penalty factor is linearly increased (through iterations) from
zero toward a very high value. The fitness function is in fact the
cost as detailed in (1).
3) A Refined Algorithm: As for a large-scale system, the
search space is quite large ( , where is the number of load
centers, and is the number of existing as well as candidate
substations), a two-stage GA is proposed in which, following
the first final run by GA, the evaluated substations are ranked ac-
cording to their number of selection times. Thereafter, a second-
Fig. 3. Normal mutation.
stage GA is run with the candidate buses equal to two times the
number of the first-run results; this time is from the ranked list.
The tests showed that this procedure improves the results.
III. SPECIFIC DETAILS
The problem formulation was fully discussed in Section II.
Some of detailed descriptions are provided in this section.
A. Load Model
Fig. 4. Most suitable mutation.
It is assumed that the existing network supplies the base load
of each  load center, and the new downward grid should be
planned for the load increase. As a result, each load (as denoted
by its magnitude and geographical properties, i.e., X and Y)
is divided into a base value and an increase. For planning the
downward grid, only the increase part is considered, while for
substation loadings and upward grid design, both parts (base and
increase) are considered.
As noted earlier, there are 16 utilities throughout the country.
The number of load centers considered for each region within a
GIS context software is shown in Table I. The load centers are in
the form of small villages or similar, in terms of demand values.
Fig. 5. Dual displacement mutation.
B. Downward Grid
number [0, 1]. Regarding the mutation operator, four options
The downward grid of the system under study is in fact the
are proposed to improve the optimization procedure:
substransmission level of the Iranian Power Grid. It comprises
" Normal Mutation as shown in Fig. 3, where is a
63- and 132-kV elements. For cost analysis of the downward
random number in the th variable range.
grid, four curves are used as shown in Fig. 6, where the hor-
" The Most Suitable Mutation as shown in Fig. 4, where
izontal axis shows the typical standard conductors available,
is the most suitable substation for supplying the th
while the vertical axes are  thermal capacity,  voltage drop,
load.
 investment cost, and  resistance, respectively. A linear ap-
" Substation Elimination Mutation in which a substation
proximation is assumed between the points, as indicated. The
is randomly selected, and all of its connecting loads are
way these curves are used is as follows. Initially, based on ac-
disconnected and connected to its closest substation.
ceptable voltage drop for the specified load (b), a conductor
" Dual Displacement Mutation as shown in Fig. 5.
size is selected . Also, based on line thermal capacity (a),
In each stage, a fitness value is calculated for each popula- a conductor size is selected, too . is selected
tion with assigning a penalty factor to the infeasible solutions
as the final choice (for the case demonstrated, ). For the se-
(i.e., the ones violating the constraints). To speed up the conver- lected conductor, the cost and resistance are then determined
gence properties of algorithm and at the same time, to use the (c,d). Right-of-way sitting difficulties are also observed by a
SEPASIAN et al.: NEW APPROACH FOR SUBSTATION EXPANSION PLANNING 1001
TABLE II
UTILITIES INFORMATION
D. Transmission Substation
The types of transmission substations considered are
400/132-, 400/63-, 230/132-, and 230/63-kV. Various capacities
are available and considered. Alternatives and classifications,
as required for each utility region, are also considered. Within
cities, the reserve is considered as 1/3 of the substation capacity.
For other regions, the reserve is so chosen not to overload the
substation with one transformer, out. As the substation voltage
level is not initially known, both 400- and 230-kV options are
considered. The solution algorithm finds the best solution.
IV. NUMERICAL RESULTS
A. General
As noted earlier, the proposed algorithm is applied to the Ira-
nian Power Grid involving 16 electric utilities, the details of
which are shown in Table II For each utility, the load is shown
for the last summer peak (2004) as well as for the horizon year
(2011). Within each utility, the load is predicted for each con-
sumption center in the capacity of a small village. Provided the
estimated load is within the capacity of available subtransmis-
Fig. 6. Line selection curves.
sion feeders, the predicted load is considered at its geographical
characteristics (respective X and Y). Otherwise, the load is dis-
large number factor in the investment cost. Moreover, advocat-
cretized so that several loads are considered at the same center.
able loads (to a specific substation) are also considered.
For large cities, such as the capital (Tehran), the loads are dis-
tributed among existing subtransmission substation geograph-
ical characteristics.
C. Upward Grid
Fig. 7 depicts the load predictions for the consumption centers
of a typical utility (Tehran). Table III shows the characteristics
The upward grid of the Iranian Power Grid comprises 400- (capacity and cost) of both subtransmission and transmission
and 230-kV elements. The line capacity and the type are de- lines employed. The costs are in Iranian monetary units (R).
termined based on capacity, voltage level, and the type of sup- The costs of transmission substations (the standard capacities
plying substation. employed for the Iranian Power Grid) are shown in Table IV.
1002 IEEE TRANSACTIONS ON POWER SYSTEMS , VOL. 21, NO. 2, MAY 2006
Fig. 7. Load centers in Tehran.
Fig. 8. Candidate substations in Tehran region.
TABLE III
TABLE V
TYPICAL LINES COST AND THERMAL CAPACITY
GA PARAMETERS AND TECHNICAL/COST PARAMETERS
TABLE VI
RESULT OF SUBSTATION EXPANSION PLANNING FOR TEHRAN ELECTRIC
UTILITY. (A) EXISTING SUBSTATIONS. (B) NEW SUBSTATIONS
TABLE IV
CAPACITY AND COST OF TYPICAL SUBSTATIONS: (A) 230/63 (KV) AND 230/132
(KV) SUBSTATIONS. (B) 400/63 (KV) AND 400/132 (KV) SUBSTATIONS
B. Candidate Selection
Within each utility, the algorithm capability is initially as-
sessed by the proposed technique implementation discussed in
Section II-C1. Thereafter, infeasible candidates (infeasible in
terms of either locational, environmental, or similar considera-
various constraints are met (see Section II-B2). The details for
tions) are removed, and extra possible candidates (based on, say,
a typical region are shown in Table VI. To further assess the
experience) are added. Fig. 8 shows the resulting candidates for
capabilities of the proposed GA in comparison with other GAs,
a typical utility (Tehran).
Table VII shows five typical GA cases (case 1 through case 5), in
which the crossover and mutation operators are set as demon-
C. Substation Allocation Results
strated. The results for the proposed GA are shown as case 6.
As discussed in Section II-C2, GA is employed for final al- The computational time for each case is also shown for 100 GA
location of new substations as well as on deciding for the ex- rerun on a Pentium IV 3.2-GHz personal computer. As is evi-
isting substations expansions. With the GA parameters and tech- dent, although the computational time is increased, the overall
nical/cost parameters shown in Table V, GA is rerun 100 times cost is reduced. As the studies conducted are for long-term plan-
for each utility to find the best solution. The final solution is ning, the computation burden is quite acceptable. Note that the
the best in terms of all costs incurred (see Section II-B1), while results are for Tehran Regional Electric Utility (TREC) as one
SEPASIAN et al.: NEW APPROACH FOR SUBSTATION EXPANSION PLANNING 1003
TABLE VII
Iranian Power Grid for the horizon year are shown in Table VIII
RESULTS OF VARIOUS GAs FOR TEHRAN REGIONAL ELECTRIC UTILITY
and Fig. 9, Table IX demonstrate the computational time for
each utility.
V. CONCLUSION
A new approach for the SEP problem was proposed in this
paper. GA was employed as the solution tool of the optimiza-
tion problem. The algorithm does not require the detailed de-
scriptions of the downward as well as upward grids. Therefore,
it can be readily applied to long-term studies. The algorithm
TABLE VIII
was tested for the Iranian Power Grid in 2011 with quite accept-
SEP SUMMARIES FOR 2011
able results. The authors are investigating an alternative solution
technique so that the results can be determined more quickly.
APPENDIX
The basics of mathematical clustering are described in [17].
Basically we are going to  cluster the load centers based on
some specific characteristics. In this paper, the load centers are
clustered based on their feeding substations. The algorithm em-
ployed is as follows.
1) Generate two lists for load centers (LL as load list) and all
existing substations (SL as substation list).
2) With due attention to voltage drop constraint, assign each
load to its nearest substation. Remove the assigned loads
from the LL.
3) Choose the most overloaded substation. If no substation
is overloaded, go to step 6.
4) Rank the loads, assigned to the aforementioned substation
(step 3), based on their least distance to the next substa-
tions. From the ranking list, remove the loads (with the
least distance) from the above substation in such a manner
that overloading is removed.
5) Add the unassigned loads to the LL and remove the above
substation from the SL.
6) Provided no load remains in the list, the algorithm ends.
Otherwise, go to step 7.
7) Choose the largest load from the LL and select its nearest
loads so that their overall sum is in the range of the ca-
pacity of a substation. The center of gravity of these loads
is considered as a candidate substation.
8) Add the candidate substation to the SL and repeat the pro-
cedure from step 1.
The above procedure is in fact an engineering-based approach in
Fig. 9. SEP results for 2011.
which a quite high feasible candidate may be reduced to quite
low and manageable candidate from which the optimum ones
TABLE IX may be chosen by the proposed GA.
COMPUTATIONAL TIMES FOR ALL UTILITIES
REFERENCES
[1] U. G. W. Knight,  The logical design of electrical networks using linear
programming methods, Proc. Inst. Elect. Eng. A, vol. 107, pp. 306 316,
1960.
[2] J. V. Old-eld and M. A. Lang,  Dynamic programming network flow
procedure for distribution system planning, in Proc. Power Industry
Computer Applications Conf., 1965.
of the largest in the country. The computational time depends on
[3] R. N. Adams and M. A. Laughton,  Optimal planning of power net-
the number of load centers (see Table I) and the number of ex-
works using mixed integer programming, Proc. IEEE, vol. 121, no. 2,
isting substations (see Table II). While the overall results for the pp. 139 147, Feb. 1974.
1004 IEEE TRANSACTIONS ON POWER SYSTEMS , VOL. 21, NO. 2, MAY 2006
[4] D. M. Crawford and S. B. Holt,  A mathematical optimization technique Mohammad Sadegh Sepasian was born in Tehran in 1967. He received the
B.Sc., M.Sc. and Ph.D. degrees from Tabriz University, Tabriz, Iran, Tehran
for locating and sizing distribution substations and deriving their optimal
University, and Tarbiat Modares University, Tehran, in 1990, 1993 and 1999,
service areas, IEEE Trans. Power App. Syst., vol. PAS-94, no. 2, pp.
respectively.
230 235, Apr. 1975.
He is currently with Power and Water University, Tehran (Ministry of Energy)
[5] E. Masud,  An iterative procedure for sizing and timing distribution
as an Assistant Professor. His research interests include power system planning
substations using optimization techniques, in Proc. IEEE PES Winter
as well as distribution planning issues.
Meeting, Feb. 1974, pp. 1281 1286.
[6] T. Gönen and B. L. Foote,  Distribution system planning using mixed-
integer programming, Proc. Inst. Elect. Eng., vol. 128, no. 2, pp. 70 79,
Mar. 1981.
Hossein Seifi (SM 01) was born in Shiraz, Iran, in 1957. He received the B.Sc.
[7] M. J. Carson and G. Corn-eld,  Design of low-voltage distribution net- degree from Shiraz University, Shiraz, Iran, in 1980 and the M.Sc. and Ph.D
works, Proc. IEEE, vol. 120, no. 5, pp. 585 593, May 1973. degrees from the University of Manchester Institute of Science and Technology,
Manchester, U.K., in 1987 and 1989, respectively.
[8] ,  Computer aided design of low-voltage distribution networks, in
He then joined Tarbiat-modares University, Tehran, Iran, where he is currently
Proc. IEE Computer Aided Design Conf., vol. 86, 1972, pp. 121 124.
a Full Professor, and at the same time is the Head of the Iran Power System
[9] T. Belgin,  Distribution system planning using mixed integer program-
Engineering Research Center (IPSERC). He was on sabbatical leave in 1995
ming, ELEKTR-IK, vol. 6, no. 1, pp. 37 48, 1998.
with ABB Network Partner, Switzerland. His research interests include power
[10] M. R. Haghifam and M. Shahabi,  Optimal location and sizing of
system planning and operational issues.
HV/MV substation in uncertainty load environment using genetic
Dr. Seifi is a Senior Member of the Iran Association of Electrical Engineers.
algorithm, Elect. Power Syst. Res., vol. 63, pp. 37 50, 2002.
[11] K. Yahav and G. Oron,  Optimal location of electrical substation in re-
gional energy supply systems, in Proc. IEEE EEIS, Jerusalem, Israel,
Nov. 5 6, 1996, pp. 307 310. Asghar Akbari Foroud was born in Hamadan, Iran, in 1972. He received the
B.Sc. and M.Sc. degrees from Tehran University and Tarbiat-modares Univer-
[12] D. E. Bouchard, M. M. A. Salama, and A. Y. Chilchani,  Optimal
sity, Tehran, Iran, in 1993 and 1996, respectively.
feeder routing and optimal substation sizing and placement using
He is now a faculty member with Semnan University, Semnan, Iran. His
guided evolutionary simulated annealing, in Proc. IEEE CCECE,
research interests include network expansion planning and power system dy-
1995, pp. 684 691.
namics.
[13] D. Hongwei, Y. Yixin, and H. Chunhua et al.,  Optimal planning of dis-
tribution substation locations and sizes model and algorithm, Elect.
Power Energy Syst., vol. 18, no. 6, pp. 343 357, 1996.
[14] G. Latorre, R. D. Cruz, J. M. Areiza, and A. Villegas,  Classification
Seyyed Hadi Hosseini was born in Saenghaleh, Zanjan, Iran, in 1974. He re-
of publications and models on transmission expansion planning, IEEE ceived the B.Sc. degree from Tehran University and the M.Sc. and Ph.D. degrees
Trans. Power Syst., vol. 18, no. 2, pp. 938 946, May 2003. from Tarbiat-modares University, Tehran, Iran, in 1996, 1999, and 2005, respec-
tively.
[15] I. J. Ramírez-Rosado and J. Antonio Domínguez-Navarro,  Possibilistic
His research interests include power transformer design, high voltage and
model based on fuzzy sets for the multiobjective optimal planning of
insulation, and network expansion planning.
electric power distribution networks, IEEE Trans. Power Syst., vol. 18,
no. 4, pp. 1801 1810, Nov. 2003.
[16] J. Choi, A. El-Keib, and T. Tran,  A fuzzy branch and bound-based
transmission system expansion planning for the highest satisfaction level
Esmaeil Mohseni Kabir was born in Gilan, Iran, in 1951. He received the B.Sc.
of the decision maker, IEEE Trans. Power Syst., vol. 20, no. 1, pp.
degree from Iran University of Science & Technology, Tehran, Iran, in 1974 and
476 484, Feb. 2005.
M.Sc. degrees from Sharif University, Tehran, in 2005.
[17] W. Li, Adaptive Fuzzy Systems and Control, Design and Stability Anal- He is currently a Deputy in planning affairs with Tavanir Co., Tehran. His
ysis. Englewood Cliffs, NJ: Prentice-Hall, 2000. research interests include all aspects of power system planning.