Electric Power Systems Research 172 (2019) 22–31

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

A strategy for transmission network expansion planning considering T

multiple generation scenarios
⁎
Patrícia F.S. Freitas, Leonardo H. Macedo, Rubén Romero
Department of Electrical Engineering, São Paulo State University, Avenida Brasil 56, Centro, 15385-000, Ilha Solteira, SP, Brazil

A R T I C LE I N FO A B S T R A C T

Keywords: Transmission network expansion planning (TNEP) is subject to several constraints that are driven by new market
Generation scenarios rules, such as generation uncertainty, demand growth, seasonal climate changes and technological advances.
Mixed-integer linear programming When considering multiple generation scenarios, due to different operational conditions, it is necessary to find
Power systems optimisation an expansion plan that allows the expanded system to operate properly for each scenario. This paper presents a
Transmission network expansion planning
strategy that provides several expansion plans that operate efficiently in all the previously defined generation
scenarios and that keep reasonable values for the expansion costs. Thus, the developed mathematical model
allows us to find adequate expansion plans with expansion costs that can be controlled, with small infeasibilities
in the operation of the system that may not be significant in the long-term planning. The proposed approach is
applied to the 24-bus system and the Colombian 93-bus system. The results indicate that the plans obtained using
the proposed strategy present much lower expansion costs compared to the conventional TNEP model with
multiple generation scenarios, with very low infeasibilities in the operation of the system, which can be cor-
rected in the short-term expansion planning.

1. Introduction renewable generation sources (dispatchable or not), competitive market

operation requirements or any other type of uncertainty, the expansion
Long-term transmission network expansion planning (TNEP) is a plan should allow the expanded system to operate in a suitable way for
classic problem of power systems. The objective is to find the optimal different generation scenarios. Especially when considering generation
expansion plan that identifies the transmission lines that must be in- and demand uncertainties, the most used way to represent these un-
stalled in the electrical system to allow a proper operation within a certainties is to build a set of representative generation and demand
predefined planning horizon with the lowest investment cost. The data scenarios and to expand the electric system taking them into account
of this problem are the current topology of the network, the candidate [3,4]. The process of constructing future scenarios for generation and
lines for addition, the generation and demand for a planning horizon demand is based on historical and forecasted data, while strategies must
and the investment restrictions. The optimal expansion plan should be used to define a reduced set of representative scenarios [5]. This
define where, how many and when the new lines should be installed. work addresses the TNEP problem considering multiple generation
Several mathematical models and optimisation techniques have scenarios and proposes a strategy to generate expansion plans with
been used to solve the TNEP problem. The direct current (DC) model is nonprohibitive expansion costs.
the formulation most often used to represent the electric network and In the TNEP problem, for each mathematical model and depending
the electric quantities. However, there are more relaxed models, such as on the complexity of the system to be solved, there is a most suitable
the transportation model, and more accurate models, like the alter- optimisation technique to be used. These techniques can be the exact
nating current (AC) model. In this work, we use the DC model [1,2] to methods [6], heuristics [7], or metaheuristics [8]. In Reference [6], the
formulate the problem. authors propose a mixed-integer linear disjunctive formulation for the
The traditional modelling for this problem considers only one gen- TNEP problem to be solved by a branch-and-bound algorithm together
eration scenario. The generation scenario choice is arbitrary; however, with a greedy randomized adaptive search procedure. Reference [7]
it is common to use the most probable or the worst-case scenario. Thus, presents a constructive heuristic algorithm for the TNEP problem, that
the generation in each generation bus is fixed at a previously specified adds one line to the network at each iteration, based on the result of a
value. Nevertheless, when it is intended to consider climatic problems, linear programming problem. The authors of Reference [8] present a

⁎
Corresponding author.
E-mail addresses: patricia.fernanda@ifsp.edu.br (P.F.S. Freitas), leohfmp@ieee.org (L.H. Macedo), ruben@dee.feis.unesp.br (R. Romero).

https://doi.org/10.1016/j.epsr.2019.02.018
Received 21 August 2018; Received in revised form 25 December 2018; Accepted 15 February 2019
Available online 07 March 2019
0378-7796/ © 2019 Elsevier B.V. All rights reserved.
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

Nomenclature θsref Reference angle in scenario s

σij Overload factor for corridor ij
Indices cij Cost of a new line in corridor ij
di Load demand at bus i
¯
i Index for buses fij Maximum power flow allowed for a line in corridor ij
ij Index for corridors ¯ new
s Index for scenarios fij Maximum power flow allowed in the lines of corridor ij
y Index for new lines with overload
_g i, s Minimum generation at bus i , scenario s
¯
Sets gi, s Maximum generation at bus i , scenario s
gi*, s Ideal generation at bus i , scenario s
B Set of buses ¯
C Set of corridors nij Maximum number of lines that can be built in corridor ij
S Set of scenarios nijo Number of existing lines in corridor ij
¯
Y Set of candidate lines ri Maximum load shedding at bus i

Functions Variables

v Investment to build new lines θi, s Voltage angle at bus i , scenario s

fijo, s Power flow in the existing lines of corridor ij, scenario s
Constants fij, y, s Power flow in line y of corridor ij, scenario s
gi,s Generation at bus i , scenario s
α Penalty factor for the load shedding ri,s Fictitious generation corresponding to the load shedding
β Penalty factor for generation displacements at bus i , scenario s
x ij Reactance of a line in corridor ij wij,y Binary variable that decides if line y will be built in cor-
δ Limit for the percentage of total load shedding in the ridor ij
system
¯
θ Maximum voltage angle

strategy based on a constructive heuristic algorithm to reduce the problem considering three objective functions: the total social cost,
search space of the TNEP and a particle swarm algorithm to solve the maximum regret and the maximum adjustment cost. The uncertainties
problem. In this work, we use the linear disjunctive model that is an of the problem are considered using scenarios. Reference [13] presents
exact optimisation formulation, which considers the DC model modified a mixed-integer nonlinear programming model for the generation and
to have an equivalent optimal solution, with the advantage that it can transmission expansion planning using a stochastic formulation with
be solved by a mixed-integer linear programming optimisation solver. probabilistic constraints to represent the uncertainties related to the
Regarding the planning horizon, the TNEP problem can be classified demands, availability of the power plants and the transmission capacity
as static or multistage. With static planning, there is only a planning factor of the lines. In Reference [14], the authors present a Benders
horizon, and with multistage planning, the planning horizon is divided decomposition algorithm to solve the TNEP problem considering a
into several stages and new transmission lines are installed at each stochastic formulation to represent the high penetration of renewable
stage. In this work, only static planning is considered. Reliability cri- generation in transmission systems considering network contingencies.
teria and uncertainties in the generation and demand can also be con- The authors of Reference [15] present a stochastic adaptive robust
sidered in the TNEP problem. optimization approach for the generation and transmission expansion
One of the pioneering works in analysing the TNEP problem con- planning problem considering uncertainties in the demand and gen-
sidering multiple generation scenarios was presented by Fang and Hill eration costs. However, these studies address the TNEP problem with
[9]. The described expansion planning considers competitive market multiple generation scenarios integrated into much more complex op-
operation requirements. However, in the proposal, independent plans timisation strategies. Also, the tests show systems that need to add few
are processed for each generation scenario, and only one of them is transmission lines; therefore, it is not possible to verify the high ex-
determined to be the most appropriate, considering an approach based pansion costs due the need to meet several generation scenarios.
on the minimisation of the maximum regret. Thus, the authors do not Reference [16] presents a robust optimisation approach based on a
present a single expansion plan that simultaneously meets all the pre- Benders decomposition algorithm to solve the TNEP problem con-
sented generation scenario requirements, and because of that, when the sidering uncertain generation and load. The tests include systems with
solution obtained by the method is considered to operate in a different renewable generation, and it is possible to verify that the investment
scenario, for which it was not originally planned, the load shedding in costs are high when compared to the conventional TNEP problem.
the system is very high. Reference [17] also presents a robust approach to the TNEP problem,
Other optimisation proposals considering a set of generation sce- that considers both uncertain future demand growth and generation
narios can be found in recent works such as References [10–15], that availability. The solution method is based on a constraint-and-column
use stochastic programming to represent uncertainties. Reference [10] generation method. In Reference [18], an approach for the robust TNEP
presents a stochastic programming formulation for the TNEP problem problem with improved performance is presented. The results of the
that considers the uncertainties related to wind generation and the load, proposed algorithm are compared with Reference [17], demonstrating
and uses a Benders decomposition algorithm to solve the problem. The the performance improvement in terms of computational times. In
authors of Reference [11] present a method based on risk/investment to Reference [19], a method for robust multistage TNEP is presented. The
solve the TNEP problem considering multiple generation and load fu- results for the robust multistage planning present lower costs than the
ture scenarios. In Reference [12], the authors use a non-dominated robust static model.
sorting genetic algorithm (NSGA-II) to solve the multi-stage TNEP Reference [20] presents a hybrid NSGA-II/Chu-Beasley algorithm

23
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

for solving the multi-objective TNEP problem considering multiple 2. Mathematical formulation
generation scenarios. The objective function minimises both the in-
vestment in new lines and the total load shedding in the system. Two The most used mathematical modelling for the TNEP problem is the
test systems are considered, the Garver 6-bus and the 24-bus systems, DC model. This model considers only the active power balance in the
and it is possible to verify that the investment cost is drastically in- system and neglects the losses. Even so, it is a mixed-integer nonlinear
creased when the multiple generation scenarios are considered in the programming problem that can be highly complex. Thus, when an
problem—for the 24-bus system, the investment cost is more than six optimisation solver is to be used, the DC model is replaced by the linear
times higher than the investment cost for a single generation scenario. disjunctive model, which is a mixed-integer linear programming pro-
Reference [21] presents an imperialist competitive algorithm to blem (MILP) that has the same optimal solution of the DC model [6,31].
solve the TNEP problem considering the uncertainty in wind generation
and the costs of investments, repair, maintenance and losses, together 2.1. Linear disjunctive model with multiple generation scenarios
with the AC operation of the network. In Reference [22], the authors
use the NSGA-II algorithm to solve the TNEP problem considering load The linear disjunctive model [6] can be adapted for the TNEP
correlation as a multi-objective problem with the following objective problem with multiple generation scenarios. In this case, the network
functions: investment cost, congestion cost and risk cost. Reference must be expanded to properly operate in the planning horizon for each
[23], proposes an approach for solving the TNEP problem in large-scale one of the different generation scenarios previously defined. This model
systems based on a network reduction strategy. assumes the form (1)–(11).
Heuristics and meta-heuristics for solving the TNEP problem con-
sidering deregulated power systems can be found in References
min v = ∑ ∑ cij wij, y
ij ∈ C y∈Y (1)
[24–28]. In Reference [24], a combined genetic algorithm and linear
programming method is used to solve the TNEP problem in a deregu- subject to
lated market environment. In Reference [25], an improved harmony
search algorithm is used, and N–1 security constraints are considered. ⎛ o ⎞ ⎛ o ⎞
∑ f +
⎜ ji, s ∑ f ji, y, s −
⎟ ∑ f +
⎜ ij, s ∑ fij, y, s + gi*, s ∀ i ∈ B, s ∈ S
⎟
Reference [26] presents a particle swarm optimisation algorithm for the ji ∈ C ⎝ y∈Y ij ∈ C y∈Y
⎠ ⎝ ⎠
multistage TNEP problem in electricity markets considering N–1 se-
= di
curity constraints. Reference [27], presents a heuristic approach for the
TNEP problem in electricity markets. Finally, Reference [28] presents a (2)
method based on the NSGA-II algorithm and a fuzzy decision-making (θi, s − θj, s )
strategy for TNEP in a deregulated environment, considering the AC fijo, s = nijo ∀ ij ∈ C , s ∈ S
xij (3)
operation of the system. In all these works, it can be verified that the
investment costs are much higher than the solution for the traditional |fijo, s | ≤ nijo f¯ij ∀ ij ∈ C , s ∈ S (4)
TNEP problem that considers a monopoly electricity market.
In this paper, we discuss the TNEP problem with multiple genera- |x ij fij, y, s − (θi, s − θj, s )| ≤ 2θ¯ (1 − wij, y ) ∀ ij ∈ C , y ∈ Y , s ∈ S (5)
tion scenarios, proposing to find a single expansion plan that operates
properly in each one of the previously specified generation scenarios. It |fij, y, s | ≤ f¯ij wij, y ∀ ij ∈ C , y ∈ Y , s ∈ S (6)
can be seen that an expansion plan considering multiple generation
scenarios presents a high investment cost. Therefore, the proposal of |θi, s| ≤ θ¯ ∀ i ∈ B, s ∈ S (7)
this paper also presents strategies to find good-quality expansion plans
with lower investment costs. ∑ wij, y ≤ n¯ ij ∀ ij ∈ C
In summary, this paper presents a mathematical model for the TNEP y∈Y (8)
problem that considers multiple generation scenarios and expands the
wij, y ≤ wij, y − 1 ∀ ij ∈ C , y ∈ Y |y>1 (9)
mathematical model so that it is possible to generate expansion plans
with a significant reduction in the expansion costs and with small in-
θsref = 0 ∀ s ∈ S (10)
feasibilities in the operation of some generation scenarios, which can be
corrected in the short-term expansion planning. wij, y ∈ {0, 1} ∀ ij ∈ C , y ∈ Y (11)
It must be emphasised that the solutions obtained for the long-term
TNEP must be analysed and reinforced with the use of other tools, such The objective function (1) represents the investment for the con-
as power flow analysis, reactive power planning, short circuit, transient struction of new transmission lines. Eq. (2) represents Kirchhoff’s cur-
and stability analysis. Therefore, even solutions that are feasible to the rent law for each bus of the system in each generation scenario. Con-
DC model, with very high investment costs, may be infeasible for the straint (3) represents Kirchhoff’s voltage law for each fundamental loop
complete AC model. The advantage of applying the proposed approach formed by the existing lines in each generation scenario. Constraints (4)
is that it can produce a set of high-quality solutions with low investment and (6) determine the transmission capacity limits for the existing lines
costs, and the network planner can choose a solution that is adequate and each new line, respectively, in each generation scenario. In these
for other requirements. constraints, the use of the absolute value is necessary because the flows
The proposed model was implemented in the mathematical mod- are bidirectional. Constraint (5) represents Kirchhoff’s voltage law for
elling language AMPL [29] and solved with the commercial solver each fundamental loop generated by a new line in each generation
CPLEX [30]. The IEEE 24-bus system is used for the tests, for which a scenario. Note that the addition of the y th line means that wij, y = 1, and
set of results with much lower expansion costs (compared to the results in this context, the power flow Eq. (5) for a new line is equivalent to the
for the traditional TNEP model with multiple generation scenarios) and power flow Eq. (3) of an existing line. If wij, y = 0 , then the y th line is not
with very small infeasibilities in the operation of the system were ob- built in corridor ij and the parameter 2θ¯ must be large enough in order
tained. to not impose a limit on the angular difference (θi, s − θj, s ) between
The rest of the paper is organised as follows: Section 2 presents the buses i and j in Eq. (5). Constraint (7) represents the angular limit at
mathematical formulation proposed to solve the TNEP problem with each bus in each generation scenario. Constraint (8) represents the limit
multiple generation scenarios and other alternative formulations; in for the number of new lines that can be installed in each corridor.
Section 3, we present the tests and results, as well as an analysis of these Constraint (9) represents a surrogate constraint that forms the se-
results; the main conclusions are presented in Section 4. quential installation of new lines in corridor ij and avoids the

24
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

generation of equal solutions with different values of wij,y. Eq. (10) transmission capacity of a line is established. The only change occurs in
new
imposes the angular reference to the system in each generation sce- f¯ij = σij f¯ij for each transmission line in the database. In the tests, va-
nario. Finally, (11) indicates that the variable wij,y must be binary, re- lues of σij, ranging from 1.00 to 1.05 (maximum permissible overload of
presenting the biggest source of complexity in the problem. 5%), were used.
The objective function and the constraints in this model are prac-
tically the same as in the model with a single generation scenario, with 2.4. Linear disjunctive model with multiple generation scenarios and small
the addition of the index s in gi*, s , to represent the generation scenarios, load shedding
and in the operation variables fijo, s , fij, y, s and θi, s , to represent the op-
eration of the system in each scenario. Also, it should be noted that for The linear disjunctive model for the TNEP problem with multiple
|S| generation scenarios, the number of operation variables increases generation scenarios can also be modified so that the expansion plans
|S| times, as does the number of constraints (2)–(7) and (10). However, can be obtained at lower costs by allowing small load shedding in some
the number of binary variables that represent the addition of trans- generation scenarios. In modern systems, considering the smart grid
mission lines, which is the main source of complexity of the problem, paradigm, load shedding can represent the load flexibility (demand
remains unchanged. response). With the long-term TNEP, it may be acceptable to allow a
Considering multiple generation scenarios reduces the feasible re- small load shedding to achieve expansion plans with a significant re-
gion of the problem, therefore, the mathematical model (1)–(11) pro- duction in expansion costs. In this context, the mathematical formula-
vides expansion plans with very high costs when compared to the ex- tion is shown in (15)–(19).
pansion plans considering a single generation scenario. This fact is
verified in the results of this paper. Thus, the following sections present min v = ∑ ∑ cij wij, y + α ∑ ∑ ri, s
several alternative mathematical models that allow us to generate ex- ij ∈ C y∈Y i∈B s∈S (15)
pansion plans for the TNEP problem with multiple generation scenarios subject to (3)–(11)
with a significant reduction of the expansion costs (and small in-
feasibilities for some generation scenarios). ⎛ o ⎞ ⎛ o ⎞
∑ f + ∑ f ji, y, s −
⎜ ji, s ⎟ ∑ f +
⎜ ij, s ∑ fij, y, s + gi, s ∀ i ∈ B, s ∈ S
⎟
ji ∈ C ⎝ y∈Y ⎠ ij ∈ C ⎝ y∈Y ⎠
2.2. Linear disjunctive model with multiple generation scenarios and small
generation displacements + ri, s = di
(16)
The mathematical modelling presented in the previous section can
be modified so that it is possible to displace the generations in a small ∑ ∑ ri, s ≤ (1 − δ ) ∑ di
range. Thus, with a small variation of the ideal generation points, an i∈B s∈S i∈B (17)
expansion plan can be generated with a reduction in the expansion
0 ≤ ri, s ≤ r¯i ∀ i ∈ B, s ∈ S (18)
costs. The new model only changes the objective function to (12) and
the power balance constraint (2) to (13), and it adds constraint (14) to
0 ≤ gi, s ≤ gi*, s ∀ i ∈ B, s ∈ S (19)
the model (1)–(11). The resulting model is shown in (12)–(14).

min v = ∑ ∑ cij wij, y + β ∑ ∑ |gi, s − gi*, s| Eq. (15) represents the objective function that minimises the costs of
ij ∈ C y∈Y i∈B s∈S (12) investment in transmission lines and the costs of load shedding in all
the generation scenarios. Eq. (16) replaces Eq. (2) for the active power
subject to (3)–(11) balance in each bus. In fact, the load shedding is represented as an
artificial generation ri,s in each demand bus in each generation scenario.
⎛ o ⎞ ⎛ o ⎞
∑ f +
⎜ ji, s ∑ f ji, y, s −
⎟ ∑ f +
⎜ ij, s ∑ fij, y, s + gi, s ∀ i ∈ B, s ∈ S
⎟
Constraint (17) imposes a limit for the total load shedding in the
ji ∈ C
⎝ y∈Y ⎠ ij ∈ C ⎝ y∈Y ⎠ system. Note that when δ = 0, the total load shedding in the system is
= di unconstrained, while when δ = 1, the total load shedding in the system
must be zero. Constraint (18) restricts the artificial generation (the load
(13)
shedding) to a maximum value, r¯i ≤ di , at each bus of the system.
_g i, s ≤ gi, s ≤ g¯i, s ∀ i ∈ B, s ∈ S Constraint (19) limits the generation at each generation bus to the ideal
(14)
value. Note that, since in (15)–(18) the demand will decrease with the
It should be noted that, in the original model, the generation gi*, s is a load shedding, (19) is different from (14), allowing only a decrease in
known parameter, and in the modified model, the generation gi,s is a the ideal generation values, so that the power balance (16) is ensured.
variable that can assume any value in a small range, from _g i, s to ḡi, s , The parameter α (MUS$/MW) represents the cost related to load
previously specified, but that is encouraged to assume the predefined shedding. Its value should be adjusted so that a small load shedding,
ideal value. Thus, the objective function is still linear, because the ab- under certain conditions, is preferable than building an additional
solute value of its second term can be easily linearised. Parameterβ , transmission line. Thus, good-quality expansion plans can be found.
given in MUS$/MW, only maintains the generations at the ideal values
if there is no possibility of reducing the expansion costs. 2.5. Linear disjunctive model with multiple generation scenarios, small line
overloads, small load shedding and small generation displacements
2.3. Linear disjunctive model with multiple generation scenarios and small
line overloads The three modelling strategies previously shown independently can
be integrated into a single, more generic mathematical model. This
In this modelling strategy, the linear disjunctive model of Section mathematical model can be used for each type of strategy, to in-
2.1 is not changed. However, the data used will be modified to allow a corporate either two strategies simultaneously or the three strategies at
small overload on the transmission lines, allowing expansion plans with the same time. Therefore, in a more general way, a single expansion
lower investment costs. plan can be found with small displacements of the ideal generations,
This strategy consists of allowing the overload of the transmission small load shedding and small overloads in the transmission lines. The
lines in acceptable percentages, i.e., the values for the overloads should formulation of this mathematical model changes the objective function
be small. Therefore, an increase of up to 5% in the maximum and adds some constraints, as shown in (20)–(24).

25
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

min v = ∑ ∑ cij wij, y + α ∑ ∑ ri, s + β ∑ ∑ |gi, s − gi*, s| only shown to give an overview of the costs of independent expansion
ij ∈ C y∈Y i∈B s∈S i∈B s∈S (20) plans, such as those made in [9]; they do not represent the expansion
strategy proposed in this work. Also, these expansion costs can be used
subject to (3)–(11)
for comparative analysis with those found with the strategies proposed
⎛ o ⎞ ⎛ o ⎞ in this work. Obviously, the expansion plans found are feasible only for
∑ f + ∑ f ji, y, s −
⎜ ji, s ⎟ ∑ f +
⎜ ij, s ∑ fij, y, s + gi, s ∀ i ∈ B, s ∈ S
⎟ each corresponding generation scenario, and they do not operate
ji ∈ C ⎝ y∈Y ⎠ ij ∈ C ⎝ y∈Y ⎠ properly for the other scenarios. Since [9] did not use an exact method
+ ri, s = di to solve the TNEP problem, optimality could not be ensured. In fact, the
(21) expansion plans obtained in [9] are not optimal. Table 2 shows the
optimal solution for each generation scenario G1–G4.
∑ ∑ ri, s ≤ (1 − δ ) ∑ di Table 3 present the load shedding values for the expansion plans of
i∈B s∈S i∈B (22) Tables 1 and 2, in which it can be verified that the load shedding is zero
0 ≤ ri, s ≤ r¯i ∀ i ∈ B, s ∈ S only for the generation scenarios for which the corresponding plan was
(23)
obtained.
_g i, s ≤ gi, s ≤ g¯i, s ∀ i ∈ B, s ∈ S The proposal presented in [9] chooses the plans for G1 and G3 as
(24)
solutions for the problem. It can be verified, in Table 3, that both plans
The objective function represented by Eq. (20) replaces objective present high values of load shedding.
function (1), constraint (21) replaces constraint (2) and constraints
(22), (23) and (24) must be added to the basic model represented in
3.1.2. Expansion plan using the linear disjunctive model with multiple
(3)–(11). In addition, the database must be modified to allow small
generation scenarios
increments of the transmission capacity of the lines, using the re-
Table 4 shows the optimal expansion plan considering the four
lationship shown in Section 2.3.
generation scenarios simultaneously and using the mathematical
The linear disjunctive model is particularly important because it is
modelling shown in Section 2. Thus, this expansion plan allows the
used as the base for the development of modified mathematical models,
expanded system to operate properly for each one of the four different
such as those presented in this section. The expansion plans found by
generation scenarios. The expansion cost is 532 MUS$, with 15 new
the mathematical models proposed in this paper are compared and
transmission lines. Obviously, this optimal expansion plan allows the
analysed in the following section.
system to operate adequately for the four generation scenarios, but the
expansion cost increases significantly when compared to the isolated
3. Tests and results
plans shown in Tables 1 and 2.
In order to evaluate the performance of the presented models, the
IEEE 24-bus system (with 41 corridors) and the Colombian 93-bus 3.1.3. Expansion plans found with the modified models for the IEEE 24-Bus
system (with 155 corridors) were used, considering the TNEP problem system
with four different generation scenarios (G1, G2, G3, G4), that re- Tests were performed with the modified models, and many alter-
present each season of the year, with equal durations. The mathema- native expansion plans were found. In the tests, the following para-
tical models were programmed in AMPL and solved using the CPLEX meters were used: β= 0.01 MUS$/MW, a small value only to avoid
solver version 12.8 with default parameters. In the tests, a computer inadequate displacements of the generations in relation to the ideal
with a 3.2 GHz Intel® Core™ i7-8700 processor with 16 GB of RAM was value, overloads in the transmission lines varying between 1% and 5%
used. and parameter α ranging from α= 0.3 MUS$/MW to 0.6 MUS$/MW.
The data for this system is presented in the appendix.
3.1. 24-Bus system The best results are shown in Table 5. The first column of Table 5
indicates the number of the plan; the second column shows the three
The data of the four-generation scenarios of the IEEE 24-bus system control parameters used to obtain the corresponding plan; the third
presented in the appendix were designed by Fang and Hill [9], in which column shows the expansion cost; the fourth column shows the max-
the IEEE 24-bus system data is also found. The total demand in the imum line overload obtained in the solution, considering all the sce-
system is 8550 MW. The maximum addition of three transmission lines narios; the fifth column shows the total load shedding for the solution,
in each corridor will be allowed in all the test cases and it is assumed in i.e., the sum of ri,s in all buses and in all scenarios; and the sixth column
this case that the maximum load shedding at a bus is equal to the de- shows the maximum generation displacement percentage in a bus in the
mand of the bus, r¯i = di , and the maximum total load sheading in the obtained solution.
system unconstrained. The dashes in Table 5 indicate that the corresponding strategy was
not used. Thus, plan number 5 was obtained when only a 4% overload
3.1.1. Expansion plan using the linear disjunctive model with a single was allowed on the transmission lines and, in this context, a plan of 472
generation scenario MUS$ was obtained. In this plan, the largest power flow capacity vio-
Table 1 presents the results when the system is expanded, con- lation appears in scenario G1 in line 12–13, which carries 519.55 MW,
sidering each generation scenario individually, as proposed in [9]. violating the limit of 500 MW by 3.91%, but it represents an expansion
Table 1 shows four expansion plans, one plan for each individual plan with an investment reduction of 60 MUS$ compared to the 532
generation scenario and the respective investment cost. These plans are MUS$ plan without violations.

Table 1
Expansion plan for each individual scenario as presented by [9].
Scenario Number of new lines Cost (MUS$)

G1 n1−5 = 1, n3−24 = 1, n6−10 = 1, n7−8 = 3, n14−16 = 1, n15−21 = 1, n15−24 = 1, n16−17 = 2, n16−19 = 1, n17−18 = 1 454
G2 n3−24 = 1, n6−10 = 1, n7−8 = 1, n10−12 = 1, n14−16 = 1, n15−21 = 1, n15−24 = 1, n16−17 = 2, n17−18 = 1, n2−8 = 1 451
G3 n1−5 = 1, n3−24 = 1, n6−10 = 1, n7−8 = 2, n10−12 = 1, n16−17 = 1, n14−23 = 1 292
G4 n3−24 = 1, n6−10 = 1, n7−8 = 2, n10−12 = 1, n12−13 = 1, n14−16 = 1, n15−24 = 1, n16−17 = 1 376

26
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

Table 2
Optimal expansion plan for each individual scenario.
Scenario Number of new lines Cost (MUS$)

G1 n1−5 = 1, n3−24 = 1, n6−10 = 1, n7−8 = 2, n14−16 = 1, n15−24 = 1, n16−17 = 2, n16−19 = 1, n17−18 = 2 390
G2 n1−5 = 1, n3−24 = 1, n6−10 = 1, n7−8 = 1, n10−12 = 1, n14−16 = 1, n15−24 = 1, n16−17 = 2, n17−18 = 2 392
G3 n6−10 = 1, n7−8 = 2, n10−12 = 1, n14−16 = 1, n16−17 = 1, n20−23 = 1 218
G4 n3−24 = 1, n6−10 = 1, n7−8 = 2, n9−11 = 1, n10−12 = 1, n14−16 = 2, n16−17 = 1 342

Table 3 Table 5
Load shedding for the expansion plans presented in [9] and for the optimal Expansion plans for the IEEE 24-bus system.
solutions for each scenario.
Plan Control parameters Cost Maximum Total load Maximum
Plan Cost G1 (MW) G2 (MW) G3 (MW) G4 (MW) Total load (MUS$/MW) (MUS line shedding generation
(MUS shedding $) overload (MW) displacement
$) (MW) (%) (%)
σ α β
G1 [9] 454 0.00 118.58 395.94 164.35 678.87
G2 [9] 451 380.65 0.00 352.74 94.04 827.43 1 – – – 532 – – –
G3 [9] 292 660.00 660.00 0.00 120.10 1440.10 2 1.02 – – 516 1.91 – –
G4 [9] 376 359.39 357.76 240.58 0.00 957.73 3 1.03 – – 512 2.08 – –
G1 optimal 390 0.00 124.98 387.26 167.46 679.70 4 – – 0.01 500 – – 7.69
G2 optimal 392 372.05 0.00 352.32 95.98 820.35 5 1.04 – – 472 3.91 – –
G3 optimal 218 479.97 386.07 0.00 132.73 998.77 6 1.02 – 0.01 472 2.00 – 10.59
G4 optimal 342 361.65 357.76 276.18 0.00 995.59 7 – 0.60 – 470 – 58.63 –
8 1.05 – – 450 4.11 – –
9 – 0.40 – 450 – 92.29 –
10 – 0.45 0.01 450 – 30.58 7.56
Table 4
11 1.03 – 0.01 450 3.00 – 6.17
Optimal expansion plan for the four generation scenarios considered simulta- 12 1.01 0.35 – 450 1.00 54.95 –
neously. 13 1.02 0.35 – 450 2.00 26.25 –
14 1.03 0.40 – 450 3.00 12.01 –
Corridor Number of new lines Cost (MUS$)
15 1.04 0.40 – 450 4.00 1.21 –
16 1.01 0.40 0.01 450 1.00 11.98 11.11
1–5 1 22
17 1.02 0.40 0.01 450 2.00 1.18 11.11
3–24 1 50
18 1.03 0.45 0.01 450 3.00 0.00 6.17
6–10 1 16
7–8 2 32 19 – 0.30 0.01 428 – 77.48 7.99
10–12 1 50 20 1.01 0.30 0.01 428 1.00 62.42 11.11
14–16 1 54 21 1.02 0.35 0.01 428 2.00 54.90 11.11
15–24 1 72 22 1.03 0.40 0.01 428 3.00 47.37 11.11
23 1.04 0.45 0.01 428 4.00 39.85 11.11
16–17 2 72
24 1.05 0.50 0.01 428 5.00 32.32 11.11
16–19 1 32
25 1.02 0.30 0.01 356 2.00 264.31 11.11
17–18 2 40
26 1.03 0.35 0.01 356 3.00 235.66 11.11
20–23 1 30
13–14 1 62 27 1.04 0.40 0.01 356 4.00 206.74 11.11
Total 15 532 28 1.05 0.40 0.01 356 5.00 179.27 11.11
29 1.05 0.35 0.01 306 5.00 306.44 11.11
30 1.05 0.30 0.01 276 5.00 390.93 11.11

Plan number 7 was found when only load shedding was allowed,
and in this context, a plan of 470 MUS$ was found. In this plan, there is It should be noted that even expansion plans with very low levels of
a load shedding of 45.26 MW at bus 10 in scenario G1 and 13.37 MW at investment, such as the plan number 27, with an investment of 356
bus 10 in scenario G4, totalling a load shedding of 58.63 MW (note that, MUS$ (a reduction of 176 MUS$ in relation to the plan without vio-
although the plan found in [9] for G1, shown in Table 1, has an ex- lations), operate in a better condition than any of the plans found for
pansion cost of 454.00 MUS$, its total load shedding is 678.87 MW). each generation scenario individually, like those presented previously
This proposal represents a reduction of 62 MUS$ when compared to the in Tables 1 and 2.
plan without violations. Plan number 4 was found when only small As an example, we show the lines added in expansion plans 4 and 7:
displacements of the generations around the ideal values were allowed;
in this context, a plan of 500 MUS$ was found. In this plan, there is a
maximum displacement of 7.69% for the generation of bus 15 in sce-
• Plan 4, with investment of 500 MUS$: n 1–5 = 1, n3–24 = 1, n6−10 =
1, n7−8 = 2, n9–11 = 1, n10–12 = 1, n14–16 = 1, n15−24 = 1, n16−17
nario G3 (the ideal generation of 325 MW is increased to 350 MW). = 2, n16−19 = 1, n17−18 = 1, n20−23 = 1.
However, this proposal represents a cost reduction of 32 MUS$ when
compared to the plan without violations.
• Plan 7, with investment of 470 MUS$: n1−5 = 1, n3−24 = 1, n6−10
= 1, n7−8 = 2, n10−12 = 1, n14–16 = 1, n15–24 = 1, n16−17 = 2;
Among the plans found using more than one strategy, plan number n16−19 = 1, n17−18 = 2, n20−23 = 1.
17 is interesting, with an investment of 450 MUS$ (a reduction of 82
MUS$ compared to the plan without violations). This plan produces an Finally, regarding the computational effort, the computational times
overload of 2% in line 12–13 in scenario G1 and in line 15–21 in sce- were low in all cases, being less than five seconds in each one.
nario G2 (510 MW, for a capacity of 500 MW), 0.13% in line 15–21 in
scenario G1 (1001.32 MW, for a capacity of 1000 MW) and 1.20% in
3.1.4. Sensitivity analysis
line 12–13 in scenario G4 (506.00 MW, for a capacity of 500 MW), a
In order to evaluate the behaviour of the proposed models when
maximum displacement of the generation of 11.11% (from 315 MW to
some control parameters change, a sensitivity analysis was carried out.
350 MW) in bus 23 and scenario G1 and a load shedding of 1.18 MW in
Four tests were performed, with the objective of showing how the in-
bus 10 and scenario G1.
vestment costs and load shedding values are dependent on the control

27
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

parameters. shedding of 0 MW is obtained.

For the test shown in Fig. 1(a), δ = 0 and σ = 1.00, i.e., the total
load shedding in the system is unconstrained and no overload is al- 3.2. Colombian 93-Bus system
lowed in the lines. Besides that, model (15)–(19) is considered, and
therefore, the generations can vary from zero until the ideal values, as Complete data of the Colombian 93-Bus system is available in [32].
shown in (19). In this case, for α varying from 0 MUS$/MW until 1.2 The total demand in the system is 14,559 MW. The maximum addition
MUS$/MW, the objective function goes from 0 MUS$ until 532 MUS$, of four transmission lines in each corridor will be allowed in all the test
while the load shedding varies from 3871.89 MW until 0 MW. The va- cases and it is assumed in this case that the maximum load shedding at
lues of objective function and load shedding for plans 7 and 9 from a bus is equal to the demand of the bus, r¯i = di , and the maximum total
Table 5 can be identified in Fig. 1(a). load sheading in the system unconstrained.
Fig. 1(b) shows that the investment cost varies from 306 MUS$ until
500 MUS$ and the load shedding varies from 642.71 MW until 0 MW 3.2.1. Expansion plans found with the proposed models for the 93-Bus
when the model (20)–(24) is solved with δ = 0 , σ = 1. 00 , β = 0.01 MUS system
$/MW and α ∈ [0, 1.8] MUS$/MW. Note that in this case, even for Table 6 presents the results of the proposed method for ten tests
α = 0 , i.e., no cost for load shedding, there is an investment of 306 MUS using different control parameters.
$. This happens because constraint (24) limits the minimum generation Plan number 1, shown in Table 6, operates without any constraint
in the system, while the power balance Eq. (21) must be satisfied, in- violation in all the generation scenarios of the model (1)–(11). How-
directly imposing a limit for the load shedding in the system. Plans 4, ever, its investment cost is 779.31 MUS$. By applying the proposed
10 and 19 from Table 5 can be identified in Fig. 1(b). strategies, it was possible to obtain good-quality expansion plans with
Fig. 1(c) shows the results for the same case shown in Fig. 1(b), but objective functions down to 511.26 MUS$ (plan 7), what represents a
with σ = 1. 05, for α ∈ [0, 0.7] MUS$/MW. Note that, in this case, the reduction of 34.39% in the investment cost when compared to plan
investment cost varies from 260 MUS$ until 450 MUS$ (lower values number 1.
than the previous case, since an overload is allowed in the lines) and the Plans number 5 and number 10 are interesting. In plan number 5, by
load shedding varies from 928.17 MW until 0 MW. Plans 24, 28, 29 and only allowing a generation displacement of up to 10%, it was possible
30 from Table 5 can be identified in Fig. 1(c). to obtain a solution with an investment cost of 587.29 MUS$, what
Finally, Fig. 1(d) shows the results for the model (15)–(19) with represents a reduction of 24.64% in the investment cost when com-
α = 0.01 MUS$/MW (a small value for the load shedding), σ = 1. 00 (no pared to plan number 1. Plan number 10 presents an investment cost of
overload in the lines) and δ = [0, 1]. In this case, when δ = 0 , the total 513.14 MUS$, what represents a reduction of 34.15% in the investment
load shedding in the system is unconstrained in (17) and Fig. 1(d) cost when compared to plan number 1, and it was obtained allowing a
shows a solution with an investment cost of 0 MUS$ and a total load maximum overload of 5% in the system’s lines, a load shedding of
shedding of 3871.89 MW. When no load shedding is allowed in the 25.35 MW (0.17% of the total load) and generation displacements of up
system, a solution with an investment cost of 532 MUS$ and a total load to 10%.

Fig. 1. Sensitivities of the objective function and total load shedding in the system considering: (a) δ = 0 , α ∈ [0, 1.2] MUS$/MW, σ = 1. 00 , in model (15)–(19), (b)
δ = 0 , α ∈ [0, 1.8] MUS$/MW, σ = 1. 00 and β = 0.01 MUS$/MW, (c) δ = 0 , α ∈ [0, 0.7] MUS$/MW, σ = 1.05 and β = 0.01 MUS$/MW, (d) δ ∈ [0, 1], α = 0.01 MUS
$/MW, σ = 1. 00 , in the model (15)–(19).

28
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

Table 6
Expansion plans for the Colombian 93-bus system.
Plan Control parameters (MUS$/MW) Cost (MUS$) Maximum line overload Total load shedding Maximum generation displacement Computational time (min)
(%) (MW) (%)
σ α β

1 – – – 779.31 – – – 26.7
2 1.05 – – 721.38 4.82 – – 119.5
3 – 0.40 – 775.29 – 4.67 – 74.13
4 – 0.30 – 712.64 – 201.11 – 170.33
5 – – 0.01 587.29 – – 10.00 47.88
6 1.05 0.40 – 699.21 5.00 27.84 – 121.67
7 1.05 0.30 – 511.26 5.00 639.11 – 305.70
8 1.05 – 0.01 530.38 5.00 – 10.00 23.87
9 1.05 0.40 0.01 522.62 5.00 5.05 10.00 15.94
10 1.05 0.30 0.01 513.14 5.00 25.35 10.00 18.52

The numbers of lines constructed in these plans are as follows: to reduce the high investment cost of the expansion plan that is ob-
tained when the traditional problem is solved.
• Plan 1, with investment of 779.31 MUS$: n 43–88 = 2, n57–81 = 2, Tests were conducted using the IEEE 24-bus system and the
n15–18 = 1, n30–64 = 1, n30–65 = 1, n57–84 = 2, n55–84 = 2, n59–67 = Colombian 93-bus system. The results indicate that the modifications
1, n55–62 = 2, n27–29 = 1, n29–64 = 1, n48–63 = 1, n62–73 = 1, n45–81 incorporated in the proposed mathematical model lead to solutions
= 2, n72–73 = 2, n19–82 = 2, n82–85 = 1, n68–86 = 2. with substantial reductions in the investment costs, with very small
• Plan 5, with investment of 587.29 MUS$: n43–88 = 2, n45–54 = 2, infeasibilities in the operation compared to other approach presented in
n30–65 = 1, n55–57 = 1, n55–84 = 1, n56–57 = 1, n55–62 = 1, n16–21 = the literature. Thus, many alternative expansion plans can be found by
1, n27–29 = 1, n29–64 = 1, n48–63 = 1, n62–73 = 1, n54–56 = 1, n72–73 simply changing the penalty and overload factors. A sensitivity analysis
= 2, n19–82 = 2, n82–85 = 1, n68–86 = 1. was conducted to evaluate the influence of the values of the penalty and
• Plan 10, with investment of 513.14 MUS$: n43–88 = 2, n45–54 = 1, overload factors in the investment costs. By allowing lower penalty
n57–84 = 1, n55–84 = 1, n56–57 = 1, n55–62 = 1, n27–64 = 1, n19–66 = factors it was possible to obtain lower investment costs. The expansion
1, n48–63 = 1, n62–73 = 1, n54–56 = 1, n72–73 = 1, n19–82 = 1, n82–85 plans found by the presented mathematical models can be considered of
= 1, n68–86 = 1. great practical interest, and the best alternatives can be identified when
the requirements of the planning entity are taken into account.
Finally, regarding the computational effort, since the TNEP problem
is solved offline, the computational times shown in Table 6, between
15.94 min and 305.70 min are acceptable. Acknowledgment

4. Conclusion This work was supported by the Coordination for the Improvement
of Higher Education Personnel (CAPES) – Finance Code 001, the
This paper proposes a method to solve the transmission network Brazilian National Council for Scientific and Technological
expansion planning problem with multiple generation scenarios. The Development (CNPq), under grant 305852/2017-5, and the São Paulo
objective is to find a set of expansion plans that operate adequately for Research Foundation (FAPESP), under grants 2014/23741-9 and 2015/
the different generation scenarios considered. Strategies were presented 21972-6.

Appendix

Table 7 presents the data of minimum and maximum generation (a small range variation) for each generation bus in each scenario of the 24-bus
system. The ideal generation data for the four generation scenarios of the IEEE 24-bus system is presented in Table 8, as are the demand data [9]. The
data for the corridors and lines of the IEEE 24-bus system are shown in Table 9. Finally, Fig. 2 shows the IEEE 24-bus system with 41 corridors used
in the tests.

Table 7
Generation data for the IEEE 24-bus system–multiple scenarios within a generation range.
Bus G1 (MW) G2 (MW) G3 (MW) G4 (MW)

Min Max Min Max Min Max Min Max

1 540 576 430 470 530 576 480 560

2 540 576 540 576 530 576 480 560
7 860 900 690 750 860 900 770 830
13 1720 1773 1380 1450 1400 1500 1540 1640
15 620 645 620 645 290 350 550 620
16 420 465 420 465 250 310 400 450
18 1160 1200 1160 1200 550 640 680 750
21 1150 1200 1150 1200 900 990 1000 1110
22 860 900 860 900 860 900 860 900
23 290 350 900 990 1900 1980 1350 1450

29
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

Table 8
Generation and demand data for the IEEE 24-bus system.
Bus Load (MW) G1 G2 G3 G4
(MW) (MW) (MW) (MW)

1 324 576 465 576 520

2 291 576 576 576 520
3 540 0 0 0 0
4 222 0 0 0 0
5 213 0 0 0 0
6 408 0 0 0 0
7 375 900 722 900 812
8 513 0 0 0 0
9 525 0 0 0 0
10 585 0 0 0 0
11 0 0 0 0 0
12 0 0 0 0 0
13 795 1773 1424 1457 1599
14 582 0 0 0 0
15 951 645 645 325 581
16 300 465 465 282 419
17 0 0 0 0 0
18 999 1200 1200 603 718
19 543 0 0 0 0
20 384 0 0 0 0
21 0 1200 1200 951 1077
22 0 900 900 900 900
23 0 315 953 1980 1404
24 0 0 0 0 0

Table 9
Data for the IEEE 24-bus system’s corridors.
Corridor Power flow Reactance Line cost Number of
capacity (MW) (p.u.) (MUS$) existing lines

1–2 175 0.0139 3 1

1–3 175 0.2112 55 1
1–5 175 0.0845 22 1
2–4 175 0.1267 33 1
2–6 175 0.1920 50 1
3–9 175 0.1190 31 1
3–24 400 0.0839 50 1
4–9 175 0.1037 27 1
5–10 175 0.0883 23 1
6–10 175 0.0605 16 1
7–8 175 0.0614 16 1
8–9 175 0.1651 43 1
8–10 175 0.1651 43 1
9–11 400 0.0839 50 1
9–12 400 0.0839 50 1
10–11 400 0.0839 50 1
10–12 400 0.0839 50 1
11–13 500 0.0476 66 1
11–14 500 0.0418 58 1
12–13 500 0.0476 66 1
12–23 500 0.0966 134 1
13–23 500 0.0865 120 1
14–16 500 0.0389 54 1
15–16 500 0.0173 24 1
15–21 500 0.0490 68 2
15–24 500 0.0519 72 1
16–17 500 0.0259 36 1
16–19 500 0.0231 32 1
17–18 500 0.0144 20 1
17–22 500 0.1053 146 1
18–21 500 0.0259 36 2
19–20 500 0.0396 55 2
20–23 500 0.0216 30 2
21–22 500 0.0678 94 1
1–8 500 0.1344 35 0
2–8 500 0.1267 33 0
6–7 500 0.1920 50 0
13–14 500 0.0447 62 0
14–23 500 0.0620 86 0
16–23 500 0.0822 114 0
19–23 500 0.0606 84 0

30
P.F.S. Freitas, et al. Electric Power Systems Research 172 (2019) 22–31

Fig. 2. IEEE 24-bus system.

References renewable generation and loads, IEEE Trans. Power Syst. 28 (2013) 4558–4567,
https://doi.org/10.1109/TPWRS.2013.2267058.
[17] C. Ruiz, A.J. Conejo, Robust transmission expansion planning, Eur. J. Oper. Res.
[1] R. Romero, A. Monticelli, A. Garcia, S. Haffner, Test systems and mathematical 242 (2015) 390–401, https://doi.org/10.1016/j.ejor.2014.10.030.
models for transmission network expansion planning, IEE Proc – Gener. Transm. [18] R. Mínguez, R. García-Bertrand, Robust transmission network expansion planning
Distrib. 149 (2002) 27–36, https://doi.org/10.1049/ip-gtd:20020026. in energy systems: improving computational performance, Eur. J. Oper. Res. 248
[2] G. Latorre, R.D. Cruz, J.M. Areiza, A. Villegas, Classification of publications and (2016) 21–32, https://doi.org/10.1016/j.ejor.2015.06.068.
models on transmission expansion planning, IEEE Trans. Power Syst. 18 (2003) [19] R. García-Bertrand, R. Mínguez, Dynamic robust transmission expansion planning,
938–946, https://doi.org/10.1109/TPWRS.2003.811168. IEEE Trans. Power. Syst. 32 (2017) 2618–2628, https://doi.org/10.1109/TPWRS.
[3] S. Lumbreras, A. Ramos, The new challenges to transmission expansion planning. 2016.2629266.
Survey of recent practice and literature review, Electr. Power Syst. Res. 134 (2016) [20] C.A.C. Florez, R.A.B. Ocampo, A.H.E. Zuluaga, Multi-objective transmission ex-
19–29, https://doi.org/10.1016/j.epsr.2015.10.013. pansion planning considering multiple generation scenarios, Int. J. Electr. Power
[4] R. Hemmati, R.-A. Hooshmand, A. Khodabakhshian, Comprehensive review of Energy Syst. 62 (2014) 398–409, https://doi.org/10.1016/j.ijepes.2014.04.063.
generation and transmission expansion planning, IET Gener. Transm. Distrib. 7 [21] M. Moradi, H. Abdi, S. Lumbreras, A. Ramos, S. Karimi, Transmission expansion
(2013) 955–964, https://doi.org/10.1049/iet-gtd.2013.0031. planning in the presence of wind farms with a mixed AC and DC power flow model
[5] L. Baringo, A.J. Conejo, Correlated wind-power production and electric load sce- using an imperialist competitive algorithm, Electr. Power Syst. Res. 140 (2016)
narios for investment decisions, Appl. Energy 101 (2013) 475–482, https://doi.org/ 493–506, https://doi.org/10.1016/j.epsr.2016.05.025.
10.1016/j.apenergy.2012.06.002. [22] S. Abbasi, H. Abdi, S. Bruno, M. La Scala, Transmission network expansion planning
[6] L. Bahiense, G.C. Oliveira, M. Pereira, S. Granville, A mixed integer disjunctive considering load correlation using unscented transformation, Int. J. Electr. Power
model for transmission network expansion, IEEE Trans. Power Syst. 16 (2001) Energy Syst. 103 (2018) 12–20, https://doi.org/10.1016/j.ijepes.2018.05.024.
560–565, https://doi.org/10.1109/59.932295. [23] S. Lumbreras, A. Ramos, F. Banez-Chicharro, L. Olmos, P. Panciatici, C. Pache,
[7] L.L. Garver, Transmission network estimation using linear programming, IEEE et al., Large-scale transmission expansion planning: from zonal results to a nodal
Trans. Power. Appar. Syst. PAS-89 (1970) 1688–1697, https://doi.org/10.1109/ expansion plan, IET Gener. Transm. Distrib. 11 (2017) 2778–2786, https://doi.org/
TPAS.1970.292825. 10.1049/iet-gtd.2016.1441.
[8] I.M. de Mendonça, I.C. Silva Junior, B.H. Dias, A.L.M. Marcato, Identification of [24] R.-C. Leou, A multi-year transmission planning under a deregulated market, Int. J.
relevant routes for static expansion planning of electric power transmission systems, Electr. Power Energy Syst. 33 (2011) 708–714, https://doi.org/10.1016/j.ijepes.
Electr. Power Syst. Res. 140 (2016) 769–775, https://doi.org/10.1016/j.epsr.2016. 2010.11.020.
05.011. [25] A. Rastgou, J. Moshtagh, Improved harmony search algorithm for transmission
[9] R. Fang, D.J. Hill, A new strategy for transmission expansion in competitive elec- expansion planning with adequacy-security considerations in the deregulated
tricity markets, IEEE Trans. Power Syst. 18 (2003) 374–380, https://doi.org/10. power system, Int. J. Electr. Power Energy Syst. 60 (2014) 153–164, https://doi.
1109/TPWRS.2002.807083. org/10.1016/j.ijepes.2014.02.036.
[10] J. Zhan, C.Y. Chung, A. Zare, A fast solution method for stochastic transmission [26] G.-R. Kamyab, M. Fotuhi-Firuzabad, M. Rashidinejad, A PSO based approach for
expansion planning, IEEE Trans. Power Syst. 32 (2017) 4684–4695, https://doi. multi-stage transmission expansion planning in electricity markets, Int. J. Electr.
org/10.1109/TPWRS.2017.2665695. Power Energy Syst. 54 (2014) 91–100, https://doi.org/10.1016/j.ijepes.2013.06.
[11] M. Rahmani, R.A. Romero, M.J. Rider, Risk/investment-driven transmission ex- 027.
pansion planning with multiple scenarios, IET Gener. Transm. Distrib. 7 (2013) [27] A. Capasso, A. Cervone, R. Lamedica, L. Palagi, A new deterministic approach for
154–165, https://doi.org/10.1049/iet-gtd.2011.0742. transmission system planning in deregulated electricity markets, Int. J. Electr.
[12] P. Maghouli, S.H. Hosseini, M.O. Buygi, M. Shahidehpour, A scenario-based multi- Power Energy Syst. 73 (2015) 1070–1078, https://doi.org/10.1016/j.ijepes.2015.
objective model for multi-stage transmission expansion planning, IEEE Trans. 06.021.
Power Syst. 26 (2011) 470–478, https://doi.org/10.1109/TPWRS.2010.2048930. [28] S. Abbasi, H. Abdi, Multiobjective transmission expansion planning problem based
[13] J.Á López, K. Ponnambalam, V.H. Quintana, Generation and transmission expan- on ACOPF considering load and wind power generation uncertainties, Int. Trans.
sion under risk using stochastic programming, IEEE Trans. Power Syst. 22 (2007) Electr. Energy Syst. 27 (2017) 1–15, https://doi.org/10.1002/etep.2312.
1369–1378, https://doi.org/10.1109/TPWRS.2007.901741. [29] R. Fourer, D.M. Gay, B.W. Kernighan, AMPL: A Modeling Language for
[14] S. Lumbreras, A. Ramos, F. Banez-Chicharro, Optimal transmission network ex- Mathematical Programming, 2nd ed., Thomson, Duxbury, 2003.
pansion planning in real-sized power systems with high renewable penetration, [30] IBM, CPLEX Optimization Studio v12.8, (2018), p. 594 (accessed June 23, 2018),
Electr. Power Syst. Res. 149 (2017) 76–88, https://doi.org/10.1016/j.epsr.2017.04. http://www-01.ibm.com/support/docview.wss?uid=swg27050618.
020. [31] S. Binato, M.V.F. Pereira, S. Granville, A new Benders decomposition approach to
[15] L. Baringo, A. Baringo, A stochastic adaptive robust optimization approach for the solve power transmission network design problems, IEEE Trans. Power Syst. 16
generation and transmission expansion planning, IEEE Trans. Power Syst. 33 (2018) (2001) 235–240, https://doi.org/10.1109/59.918292.
792–802, https://doi.org/10.1109/TPWRS.2017.2713486. [32] LaPSEE Power System Test Cases Repository, (2018) (accessed December 10, 2018),
[16] R.A. Jabr, Robust transmission network expansion planning with uncertain http://www.feis.unesp.br/#!/lapsee.

31