Received: 23 October 2018 Revised: 18 January 2019 Accepted: 25 February 2019

DOI: 10.1002/2050-7038.12029

RESEARCH ARTICLE

A new AC OPF tool for sub‐transmission networks

considering distribution switching actions and load‐
transferring capability

Mohammad Ali Tavakkoli | Nima Amjady

Department of Electrical Engineering,

Semnan University, Semnan, Iran
Abstract
In practical power systems, in order to study the results of optimal power flow
Correspondence (OPF) for a sub‐transmission network (STN), the downstream primary distri-
Mohammad Ali Tavakkoli and Nima
Amjady, Department of Electrical bution network is usually modeled as constant load. However, in practical dis-
Engineering, Semnan University, Semnan, tribution grids, the operators are constantly transferring load from one primary
Iran.
distribution feeder to another one because of, eg, congestion problems or main-
Email: mo‐[email protected];
[email protected] tenance plans. Considering distribution switching actions can improve the per-
formance of sub‐transmission OPF tool that typically aims at reducing the
operation cost of STN. This paper presents a new sub‐transmission Alternating
Current (AC) OPF (ST‐AC‐OPF) considering load transferring (LT) between
primary distribution feeders implemented through distribution maneuvers.
The proposed OPF tool is tested on an illustrative practical example as well
as on a real‐world Tehran STN. The numerical results show that the proposed
model can obtain a more optimal operating point than conventional ST‐AC‐
OPF in terms of the operation cost and load shedding.

KEYWORDS
load transferring (LT), primary distribution feeder, sub‐transmission AC OPF (ST‐AC‐OPF), sub‐
transmission network (STN), switching action

1 | INTRODUCTION

1.1 | Motivation and background

Optimal power flow (OPF) aims at optimizing the operating state of power system, eg, by minimizing the total produc-
tion cost, subject to the technical and economic constraints of power system. For practical power systems, OPF is usu-
ally formulated as a large‐scale complex mathematical optimization problem. For instance, AC OPF is nonlinear and
nonconvex because of including AC power flow equations. Both mathematical and heuristic methods have been used
to solve OPF. A review of earlier mathematical OPF solution methods, such as sequential quadratic programming, New-
ton's method, linear programming, and interior point methods, can be found in previous studies.1,2 To enhance the per-
formance of these methods, various decomposition and transformation techniques, such as Benders decomposition with
special ordered sets3 and Signomial convex transformation,4 have been presented in the newer mathematical OPF solu-
tion methods. In addition, heuristic optimization methods have been also proposed to solve OPF problems and a review
of these methods can be found in other studies.5,6
Int Trans Electr Energ Syst. 2019;29:e12029. wileyonlinelibrary.com/journal/etep © 2019 John Wiley & Sons, Ltd. 1 of 17
https://doi.org/10.1002/2050-7038.12029
2 of 17 TAVAKKOLI AND AMJADY

Recently, switching actions as an effective and important set of decision variables have been considered in OPF prob-
lem. These switching actions change the network topology. Various switching actions on transmission or distribution
lines, busses, and shunt devices have been reviewed in Glavitsch.7 In the following, transmission switching and distri-
bution switching research works are briefly reviewed.
Previous studies8,9 have introduced new algorithms including bus‐bar switching and line switching to deal with line
overload and voltage correction. Line switching has been used in previous studies10-13 to improve economic dispatch
and unit commitment schedule. In Khanabadi et al,14 the issue of voltage security has been addressed by transmission
line switching, and it has been shown that voltage profile can be improved in this way. Previous studies14,15 have studied
the effect of network switching actions on locational marginal prices (LMPs). Dehghanian and Kezunovic16 have stud-
ied the impact of transmission switching on the system reliability and have concluded that appropriate transmission
switching actions could improve the system reliability. In Reddy and Bijwe,17 maximizing reliability and minimizing
switching cost have been considered as two conflicting objectives of OPF. Previous studies18,19 have evaluated the loss
reduction of power system using transmission switching and have concluded that transmission switching could reduce
the system losses. Based on the studies,7-19 it is concluded that transmission switching can have the advantages of (a)
improving the voltage profile, (b) enhancing the system reliability, and (c) reducing the power losses.
Recently, some distribution operation research works have studied the impacts of distribution reconfiguration.
Gomes et al20 have presented a distribution system reconfiguration approach based on distribution OPF for loss reduc-
tion. In Khodr et al,21 other decision variables of distribution OPF have been considered, in addition to distribution sys-
tem reconfiguration, for distribution loss reduction. In previous studies,22,23 distribution reconfiguration has been used
for distribution energy loss minimization considering daily load profiles. Alonso et al24 have studied the impact of dis-
tribution network reconfiguration on distribution loss and reliability through a multi‐objective framework. In
Bernardon et al,25 distribution network reconfiguration has been used to decrease the power loss of distribution and
sub‐transmission systems. Load transferring (LT) in distribution network has been used in previous studies26-28 to opti-
mize the distribution network operating point, while the impact of this LT on the upstream sub‐transmission grid has
not been considered. In addition, to the best of the authors' knowledge, none of the transmission switching and
distribution switching research works has presented a sub‐transmission AC OPF (ST‐AC‐OPF) that can minimize
sub‐transmission operation cost and load shedding. Moreover, optimizing distribution switching actions along with
optimizing STN control actions has not been addressed yet. Addressing these unsolved issues is a main focus of this
paper. Developing a specific OPF tool for STNs considering their practical characteristics is another focus of this paper
as described below.
STN is the intermediate grid between transmission and distribution networks, which receives energy from transmis-
sion system and delivers it to distribution network.29 For instance, in Iran's power system, the voltage levels of transmis-
sion network are 400 kV and 230 kV, the voltage levels of STN are 132 kV and 63 kV (mostly 63 kV), and the voltage
levels of primary distribution network are 33 kV, 20 kV, and 11 kV (mostly 20 kV). Unlike transmission network, STN is
usually operated radially in Iran to limit short‐circuit fault currents. However, its radial structure is different from the
radial structure of distribution grid as feeders of distribution grid usually have several sections, several connected loads,
and several switches unlike the feeders of STN. The OPF tools used for STNs in practice assume that the load located at
distribution busses are constant, such as the OPF tool produced by the ALSTOM company of France30 and used in Teh-
ran sub‐transmission dispatch center of Iran. However, in real‐world distribution grids, the operators are constantly
transferring load from one primary distribution feeder to another one because of, eg, congestion problems or mainte-
nance plans. Today, dispatch centers can remotely monitor and control distribution switches. If LT in distribution net-
work is not considered in a sub‐transmission OPF tool, its results might not be optimum for the STN. This research
work also focuses on this subject area, ie, developing a new specifically‐designed OPF tool for STNs that can consider
and optimize distribution LT.

1.2 | Contributions
The main contributions of this paper can be summarized as follows:

1. A new specifically designed AC OPF model for sub‐transmission network is presented in this research work. Unlike
currently used OPF tools in practical sub‐transmission networks assuming constant load in distribution busses, the
proposed ST‐AC‐OPF tool can model and optimize LT between primary distribution feeders. Although some trans-
mission and distribution OPF models including switching actions have been recently presented in the literature, to
TAVAKKOLI AND AMJADY 3 of 17

the best of the authors' knowledge, there is no sub‐transmission OPF tool that can model LT of primary distribution
feeders along with the STN control actions to optimize the operating point of the STN and primary distribution
network.
2. Unlike sub‐transmission feeders, primary distribution feeders usually include several sections, several connected
loads, and several switches. Thus, many various switching actions and maneuvers are typically possible in a distri-
bution grid. The proposed ST‐AC‐OPF tool can optimize these switching actions along with optimizing sub‐
transmission network control actions. In other words, the proposed ST‐AC‐OPF can employ the flexibility caused
by distribution maneuvers to obtain a better operating point for both sub‐transmission and distribution networks.
In addition, since the proposed ST‐AC‐OPF model is mathematically formulated as a mixed‐integer nonlinear pro-
gramming (MINLP) optimization problem, an efficient signomial convexification method has been applied to make
it computationally tractable.
3. The effectiveness of the proposed ST‐AC‐OPF tool in terms of reducing sub‐transmission operation cost and load
shedding as well as reducing computation burden, compared with conventional ST‐AC‐OPF tool, is shown on a
real‐world Tehran STN.

1.3 | Assumptions

1. For the sake of simplicity and better illustrating the underlying ideas of the proposed ST‐AC‐OPF model, the uncer-
tainty sources (eg, the uncertainty of renewable Distributed Generation (DG) connected to STN) as well as contin-
gencies are neglected in this research work. Various scenario approaches,31 stochastic programming techniques,32,33
and robust optimization methods34 are available for OPF uncertainty modeling in the literature, which can be
merged with the proposed ST‐AC‐OPF tool to model the uncertainty sources. In addition, contingencies can be
included in the proposed ST‐AC‐OPF by considering contingency constraints.35
2. The proposed ST‐AC‐OPF tool can model STN as well as primary distribution feeders. However, the upstream trans-
mission system is not considered in the proposed ST‐AC‐OPF, which is the common industry practice for sub‐
transmission OPF tools. Optimizing transmission system settings considering the load of sub‐transmission system
is usually performed by transmission OPF.

The rest of this paper is organized as follows. In Section 2, the proposed ST‐AC‐OPF model considering LT and opti-
mizing distribution switching actions is introduced. In Section 3, the numerical results obtained from applying the pro-
posed ST‐AC‐OPF model to an illustrative practical example as well as a real‐world Tehran STN are presented and
compared with the results of a conventional ST‐AC‐OPF. Finally, conclusions are drawn in Section 4.

2 | MODEL DESCRIPTION

In this section, the concept of LT between primary distribution feeders using an illustrative real‐world example is intro-
duced. Afterward, a conventional ST‐AC‐OPF model is presented. Then, the proposed ST‐AC‐OPF model considering
the LT concept is detailed to optimize distribution switching actions along with optimizing STN control actions. The last
part of this section presents application of signomial convexification‐based solution method to solve the proposed ST‐
AC‐OPF model.

2.1 | An illustrative real‐world example to introduce the proposed LT concept

Figure 1 shows a small part of Tehran transmission and sub‐transmission networks. In this figure, Bus 1 is the 230 kV
side of the upper 230 kV/63 kV substation, while Bus 2 and Bus 3 are the 63 kV side with a normally open circuit
breaker (CB) between them. Bus 2 and Bus 3 transfer the energy received from the transmission grid through two
63 kV lines to two 63 kV/20 kV substations A and B. In these two substations, Bus 4 and Bus 5 indicate the 63‐kV side,
while Bus 6 and Bus 7 are the 20‐kV side. Primary distribution feeders A and B are connected to the distribution Busses
6 and 7, respectively. Transmission and sub‐transmission parts with the border between them are indicated in Figure 1.
In conventional sub‐transmission OPF tools, loads of primary distribution feeders A and B (ie, PdA,QdA and PdB,QdB)
4 of 17 TAVAKKOLI AND AMJADY

FIGURE 1 A small part of real‐world Tehran transmission/sub‐transmission grid

are considered constant. However, the portion of downstream grid of this real‐world network is as shown in Figure 2. It
is seen that there are three 20 kV switches, S1, S2, and S3, as well as four loads, L1, L2, L3, and L4, between two primary
distribution feeders, A and B. Switches S1, S2, and S3 represent three maneuver points of the distribution network. The
statuses of these switches are continuously changed by the associated distribution system operator. For instance, in one
operation mode, S1 and S2 are closed while S3 is open (in which loads L1, L2, and L3 should be fed by feeder A, and load
L4 should be fed by feeder B); and in another operation mode, S1 and S3 are closed while S2 is open (in which loads L1
and L2 should be fed by feeder A, and loads L3 and L4 should be fed by feeder B). Thus, by this distribution maneuver,
load L3 is transferred from feeder A to feeder B. This illustrates the concept of LT between primary distribution feeders.
The proposed ST‐AC‐OPF tool models this concept and the associated distribution switching actions (such as switching
of S1, S2, and S3 in Figure 2) to optimize the operating point of the sub‐transmission network and primary distribution
network. In practice, to implement these distribution switching actions determined by the sub‐transmission OPF, the
sub‐transmission dispatch center can send a request for the distribution dispatch center to implement the required
switching actions. It is noted that a distribution dispatch center, although is aware of the loading condition of its pri-
mary distribution feeders, may not be able to perform the LT as it cannot monitor the ST network to observe the effect
of the LT on the STN. For instance, an LT between primary distribution feeders in Figure 1 may lead to congestion in
the upstream ST transformers and lines (ie, congestion of 63 kV/20 kV transformers and 63 kV lines in Figure 1). How-
ever, this congestion cannot be determined (and thus cannot be avoided) in the distribution dispatch center as it is not
aware of the loading condition and capacity of ST transformers and lines. Alternatively, the proposed ST‐AC‐OPF,
which optimizes both ST control actions and distribution switching actions (using the information of both ST network
and primary distribution feeders), can perform the LT between primary distribution feeders without causing congestion
for the above ST network. This illustrates the importance of the proposed ST‐AC‐OPF.
To better illustrate the LT concept, Figure 2 includes only two feeders with three switches between them. However,
as will be shown below, the proposed ST‐AC‐OPF can model any number of feeders with any number of switches
between them.

FIGURE 2 Downstream grid of feeders A and B of Figure 1

TAVAKKOLI AND AMJADY 5 of 17

2.2 | Formulation of a conventional ST‐AC‐OPF

A conventional AC OPF, typically used for STNs, can be formulated as follows:

Min OF Pi ; Qi ; Pgi ; LSi ; Qci ; tcij ; (1)

OF ¼ ∑i∈ΨB EPi × Pi ⏟ þ ∑i∈ΩB CLS × LSi ⏟ þ ∑i∈ΩB Cgi × Pgi ⏟ ; (2)

Cost1 Cost2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Cost3
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}

Pi þ Pgi − Pdi þ LSi ¼ ∑j∈ΩB Pbij ; i ∈ ΨB ; (3)

   
Qi þ Pgi × tg cos−1 ðPFgi Þ þ Qci − ðPdi − LSi Þ × tg cos−1 ðPFdi Þ ¼ ∑j∈ΩB Qbij ; i ∈ ΨB ; (4)

Pgi − Pdi þ LSi ¼ ∑j∈ΩB Pbij ; i ∈ ðΩB − ΨB Þ; (5)

   
Pgi × tg cos−1 ðPFgi Þ þ Qci − ðPdi − LSi Þ × tg cos−1 ðPFdi Þ ¼ ∑j∈ΩB Qbij ; i ∈ ðΩB − ΨB Þ; (6)

 2  
Pbij ¼ tcij × Gij × V 2i − V i × V j × tcij × Gij cosδij þ Bij sinδij if tap setting is at bus i ; (7)

 2  
Qbij ¼ − tcij × Bij × V 2i − V i × V j × tcij × Gij sinδij − Bij cosδij if tap setting is at bus i; (8)

 
Pbij ¼ Gij × V 2i − V i × V j × tcij × Gij cosδij þ Bij sinδij if tap setting is at bus j; (9)

 
Qbij ¼ −Bij × V 2i − V i × V j × tcij × Gij sinδij − Bij cosδij if tap setting is at bus j; (10)

 2  2  2
Pbij þ Qbij ≤ Sbmax
ij ; i; j ∈ ΩB ; (11)

V min
i ≤ V i ≤ V max
i ; i ∈ ΩB ; (12)
 
Pgmin
i ≤ Pgi ≤ Pgmax
i ; Qgmin
i ≤ Pgi × tg cos−1 ðPFgi Þ ≤ Qgmax
i ; i ∈ ΩB ; (13)

 2
ðPi Þ2 þ ðQi Þ2 ≤ Smax
i ; i ∈ ΨB ; (14)

Qcmin
i ≤ Qci ≤ Qcmax
i ; i ∈ ΩB ; (15)

0 ≤ LSi ≤ Pdi ; i ∈ ΩB ; (16)

tcmin
ij ≤ tcij ≤ tcmax
ij ; i; j ∈ ΩB : (17)

The decision variables of conventional ST‐AC‐OPF are indicated in (1). The objective function of this OPF including
three components is presented in (2) where Cost1, Cost2, and Cost3 indicate the cost of purchased power from the upstream
6 of 17 TAVAKKOLI AND AMJADY

transmission grid, cost of load shedding, and cost of DG generation, respectively. Active and reactive power balance con-
straints at boundary busses connected to the upstream transmission grid are given in (3) and (4), respectively. Active and
reactive power balance constraints at other busses are presented in (5) and (6). Active and reactive power flows of branches
considering tap setting of tap changing transformers are calculated in (7) to (10). For transmission lines, tap setting is set to
1. Apparent power flows of branches are limited in (11). Voltage magnitude limits are shown in (12). Equation (13) bounds
the active and reactive powers generated by DGs. It is assumed that DGs work with constant power factor.36 Powers pur-
chased from upstream transmission grid are limited in terms of apparent power capacity of boundary substations in (14).
Constraints of shunt compensators' reactive powers, load sheds, and tap settings are given in (15), (16), and (17), respec-
tively. From (1), it is seen that the decision variables of conventional ST‐AC‐OPF do not include primary distribution
switching actions. Accordingly, the variability of Pdi values, occurred in practice, is ignored.

2.3 | Proposed ST‐AC‐OPF model based on LT

The proposed ST‐AC‐OPF model is as follows:

Min OF Pi ; Qi ; Pgi ; LSi ; Qci ; tcij ; Sn ; (18)

ð2Þ; ð7Þ − ð15Þ; ð17Þ; (19)

Pi þ Pgi − PdSi þ LSi ¼ ∑j∈ΩB Pbij ; i ∈ ΨB ; (20)

     
Qi þ Pgi × tg cos−1 ðPFgi Þ þ Qci − PdSi − LSi × tg cos−1 ðPFdi Þ ¼ ∑j∈ΩB Qbij ; i ∈ ΨB ; (21)

Pgi − PdSi þ LSi ¼ ∑j∈ΩB Pbij ; i ∈ ðΩB − ΨB Þ; (22)

     
Pgi × tg cos−1 ðPFgi Þ þ Qci − PdSi − LSi × tg cos−1 ðPFdi Þ ¼ ∑j∈ΩB Qbij ; i ∈ ðΩB − ΨB Þ; (23)

0 ≤ LSi ≤ PdSi ; i ∈ ΩB : (24)

In the proposed ST‐AC‐OPF, primary distribution switching actions Sn have been added to the decision variables.
Another important difference with conventional ST‐AC‐OPF is considering variables PdSi instead of constant values
Pdi. To introduce the impact of decision variables Sn on the loads PdSi , consider the practical example presented in
Figures 1 and 2, but with k switches and k + 1 loads in general (instead of three switches and four loads in Figure 2).
In this general case, active loads of feeders A and B, considering distribution switching actions, can be computed as
follows:
i − 1 
k k
PdSA ¼ L1 þ ∑i¼2 Li × ∏ Sm × ∑n¼i ð1 − Sn Þ ; (25)
m¼1

 k

k i−1
PdSB ¼ Lkþ1 þ ∑i¼2 Li × ∏ Sm × ∑n¼1 ð1 − Sn Þ ; (26)
m¼i

Si þ Sj ≥ 1; i ¼ 1; ⋯; k − 1; j ¼ i þ 1; ⋯; k; (27)
TAVAKKOLI AND AMJADY 7 of 17

k
∑i¼1 Si ≤ k − 1; (28)

jSA j ≤ Smax
A ; jSB j ≤ SB :
max
(29)

Equations (25) and (26) apply the impact of primary distribution switching actions S1, …,Sk on the loads of feeders PdSA
and PdSB . To better illustrate the performance of these equations, consider the example of Figures 1 and 2 with k = 3, for
which (25) and (26) become as (30) and (31), respectively, as follows:

PdSA ¼ L1 þ L2 × fS1 × ½ð1 − S2 Þ þ ð1 − S3 Þg þ L3 × fS1 × S2 × ð1 − S3 Þg; (30)

PdSB ¼ L4 þ L3 × fS3 × ½ð1 − S1 Þ þ ð1 − S2 Þg þ L2 × fS3 × S2 × ð1 − S1 Þg: (31)

Based on Figure 2, to include L2 in the load of feeder A, ie, PdSA , switch S1 should be closed (S1 = 1) and switch S2 or
switch S3 should be open (S2 = 0 or S3 = 0). Note that both S2 and S3 cannot be open due to (27). In this case, the second
term in the right‐hand‐side of (30) yields the following:
8 9
< 1 × ½ð1 − 0Þ þ ð1 − 1Þ >
> =
L2 × fS1 × ½ð1 − S2 Þ þ ð1 − S3 Þg ¼ L2 × or ¼ L2 × 1 × 1 ¼ L2 : (32)
>
: >
;
1 × ½ð1 − 1Þ þ ð1 − 0Þ

Moreover, to include L3 in PdSA , S1 and S2 should be closed (S1 = 1 and S2 = 1) and S3 should be open (S3 = 0). In this
case, the third term in the right‐hand side of (30) yields the following:

L3 × fS1 × S2 × ð1 − S3 Þg ¼ L3 × f1 × 1 × 1g ¼ L3 : (33)

Thus, (30) correctly applies the impact of switching actions S1, S2, and S3 on PdSA . Similarly, it can be shown that (31)
correctly applies the impact of switching actions S1, S2, and S3 on PdSB . Constraints (27) avoid disconnecting loads L2,
…,Lk, or any combination of them (these loads should be supplied from one side). Constraint (28) avoids ring connection
between primary distribution feeders A and B. Constraints (29) restrict loading of feeders A and B to their allowable
limits. Formulations (25) to (29) correctly model the impact of feasible distribution switching actions on the loads of
feeders A and B (with any number of switches and loads between them) while avoiding infeasible distribution switching
actions. By obtaining the loads of primary distribution feeders, the loads of primary distribution busses (eg, the loads of
Busses 6 and 7 in the example of Figure 1) can be determined. By determining these loads, the loads of sub‐transmission
busses (eg, the loads of Busses 4 and 5 in the example of Figure 1) can be calculated. Thus, PdSi variables, to be used in
proposed ST‐AC‐OPF models (18) to (24), are obtained. Accordingly, to incorporate the impact of distribution switching
actions S1, …,Sk as well as the LT between feeders A and B into the proposed ST‐AC‐OPF, binary variables S1, …,Sk
should be added as Sn to the decision variables in (18). Also, (25) to (29) should be added to the constraints (19) to
(24). Similarly, we can incorporate LT between any other pair of primary distribution feeders and the associated distri-
bution switching actions into the proposed ST‐AC‐OPF. It is noted that in practice, many switches may exist between
two primary distribution feeders. However, only switches with these three characteristics are considered as decision var-
iables in the proposed sub‐transmission OPF: (a) switch should be interruptible under load, (b) switch should have
remote control from the associated dispatch center, and (c) switch should not be “must closed” or “must open” (eg,
because of maintenance programs). For instance, in the practical example of Figures 1 and 2, in fact, there are 14
switches between feeders A and B. However, only three switches have these three characteristics, which have been con-
sidered as S1, S2, and S3 in Figure 2.
Since we are proposing a sub‐transmission AC OPF tool, only the distribution maneuvers that can be performed in a
sub‐transmission dispatch center (and not all distribution network maneuvers) are included in the proposed ST‐AC‐
OPF model. In this regard, the constraints (25) to (28) can model load transferring between any pair of feeders with
any number of switches between the feeders. In other words, the constraints (25) to (28) can model any primary distri-
bution maneuver performed in a sub‐transmission dispatch center. Even if a primary distribution feeder has electrical
connections with several other primary distribution feeders and it is needed to transfer the load between several feeders,
8 of 17 TAVAKKOLI AND AMJADY

in practice, the dispatchers transfer the load between one pair of these feeders, then between the next pair, and so on.
This pairwise load transferring between primary distribution feeders continues until all required load transfers are com-
pleted. Thus, even in such cases, the constraints (25) to (28) can correctly model the primary distribution maneuvers.
Here, other major differences between the proposed ST‐AC‐OPF model and previous transmission switching and dis-
tribution switching research works can be described as follows:

1. While transmission switching works do not consider LT in the downstream grid, LT between primary distribution
feeders yielded by distribution switching actions is modeled in the proposed ST‐AC‐OPF to optimize the operating
point of both sub‐transmission and primary distribution networks.
2. Distribution switching works use distribution switching actions to optimize distribution network operating state.
Only Bernardon et al25 has taken into account sub‐transmission loss in addition to distribution loss, while only dis-
tribution switching actions have been considered in Bernardon et al.25 In other words, distribution switching works
mostly have the viewpoint of distribution optimization without evaluating its impact on the upstream sub‐
transmission network. However, the proposed ST‐AC‐OPF has the viewpoint of sub‐transmission optimization. In
addition, the proposed ST‐AC‐OPF uses primary distribution switching actions along with the sub‐transmission
control actions to optimize the operating state of both sub‐transmission and primary distribution networks.

2.4 | Convexification of the proposed ST‐AC‐OPF model

In order to obtain the global solution of the nonconvex ST‐AC‐OPF model, signomial convexification method presented
in Attarha and Amjady 4 is utilized. This method first converts the highly nonconvex trigonometric functions of the
nonconvex ST‐AC‐OPF model to signomial functions using the second order approximation of Taylor series as follows:

x3
sinx≅x − ; (34)
3!

x2
cosx≅1 − : (35)
2!

Considering the analytic behavior of these trigonometric functions and practical variation ranges of their arguments δij
in sub‐transmission systems, the approximations (34) and (35) have negligible error for the proposed ST‐AC‐OPF model.
Substituting (34) to (35) in (7) to (10) leads to an AC model that is still nonconvex. However, this AC OPF model is for-
mulated in terms of signomial functions. A signomial function f (.) of n variables x1, …, xn is defined as37

n p
m
f ðx 1 ; …; x n Þ ¼ ∑j¼1 aj ∏ x i ji ; (36)
i¼1

where m is the number of signomial terms; aj and Pji are constant coefficients and constant powers, respectively. Based
n p
on (36), the signomial function f (.) is the sum of m signomial terms aj ∏ x i ji . Under specific conditions on the coeffi-
i¼1
cients aj and powers Pji, and using this property that the sum of convex terms is convex, the signomial function f (.)
becomes convex.38 In the signomial convexification approach, shift transformation, power transformation, and sign
transformation are applied to the arguments x1, …, xn and powers pji to make the signomial representation of the ST‐
AC‐OPF model convex. Using the shift transformation, the arguments x1, …,xn become positive so that these variables
do not fluctuate between positive and negative values. Afterward, the power transformations are applied to the
signomial terms of the ST‐AC‐OPF model with positive arguments to make these terms convex. The inverse power
transformations, which are required for returning to the original solution space, are made convex using the sign trans-
formation. The details of these transformation techniques can be found in Attarha and Amjady.4
The proposed ST‐AC‐OPF model always converges, since it is made convex and convex optimization problems can be
solved to obtain their global optimum.39 In the next section, it will be shown that this convexification method signifi-
cantly decreases the computation burden of the proposed ST‐AC‐OPF model.
TAVAKKOLI AND AMJADY 9 of 17

3 | N U M E R I C A L RE S U L T S

The proposed ST‐AC‐OPF model has been tested on the practical example of Figures 1 and 2 as well as on a real‐world
STN. The proposed ST‐AC‐OPF model has been implemented within GAMS programming environment and solved
using KNITRO solver.40 In the simulation studies of this section, the required economic data has been obtained from
Ahmadigorji and Amjady.41 The results obtained for the two case studies are presented in the next subsections.

3.1 | Practical example of Figures 1 and 2

The resistance, reactance, and capacity of sub‐transmission branches in Figure 1 are given in Table 1. Also, four loads
L1, L2, L3, and L4 in Figure 2 are 4 MW, 2.5 MW, 2.85 MW, and 3.6 MW, respectively, with 0.98 lagging power factor.
The results obtained from the current ST‐AC‐OPF and the proposed ST‐AC‐OPF for this practical example are given in
Table 2. The current ST‐AC‐OPF tool has been produced by the ALSTOM Company of France30 and used in Tehran
sub‐transmission dispatch center of Iran. The same technical and economic data has been used for both the current
ST‐AC‐OPF and proposed ST‐AC‐OPF in all numerical tests of this paper. Table 2 shows that the cost of purchased
power from the upstream transmission grid (ie, Cost1) and total cost (ie, OF) obtained by the proposed ST‐AC‐OPF
are lower than those obtained by the current ST‐AC‐OPF. Its reason can be described as follows.
Current ST‐AC‐OPF considers the statuses of switches S1, S2, and S3 as given in the snapshot taken by the sub‐
transmission SCADA* system (which are S1 = 1, S2 = 1 and S3 = 0) without changing these switches. However, the pro-
posed ST‐AC‐OPF optimizes these switch statuses along with other OPF decision variables as given in (18) to (24). The
proposed ST‐AC‐OPF determines these switch statuses as S1 = 1, S2 = 0 and S3 = 1. Thus, the loads of feeders A and B
in the current sub‐transmission OPF are PdA = L1+L2+L3= 4 + 2.5 + 2.85= 9.35 MW and PdB = L4=3.6 MW. However,
the loads of feeders A and B in the proposed sub‐transmission OPF are PdSA ¼ L1 þ L2 ¼4 + 2.5=6.5 MW and
PdSB ¼ L3 þ L4 ¼ 2.85 + 3.6=6.45 MW. Thus, a more balanced loading condition for feeders A and B is obtained in
the proposed sub‐transmission OPF compared to current sub‐transmission OPF using the additional decision variables
of switch statuses. This more balanced loading condition leads to lower losses and so lower purchased power of the pro-
posed ST‐AC‐OPF than the current ST‐AC‐OPF as given in Table 2. However, the cost reductions in this small practical
example, given in the last row of Table 2, are small values. Since, in this practical example, two 20 kV feeders belong to
the same 230 kV/63 kV substation (as shown in Figure 1), these two feeders have the same electricity price. In addition,
the reduction in purchased power is only related to the reduction in power loss, which is a small value. Moreover, there
is no load shedding and DG in this example and thus their costs in both the current and proposed ST‐AC‐OPF tools are
zero in Table 2. In the next subsection, a larger test case with DG and load shedding is considered in which greater dif-
ferences between the current and proposed ST‐AC‐OPF tools are observed.

TABLE 1 Sub‐transmission branch data for the practical example of Figures 1 and 2

From Bus To Bus R (Ω) X (Ω) Capacity (MVA)

2 4 1.54791 1.11921 51
3 5 1.13692 0.49974 51
4 6 0.10502 2.62537 60
5 7 0.20381 5.23511 30

TABLE 2 Results obtained from the current ST‐AC‐OPF and the proposed ST‐AC‐OPF for the practical example of Figure 1 and 2

OPF model Cost1[$] Cost2 [$] Cost3 [$] Total cost (OF) [$]

Current 648.5 0 0 648.5

Proposed 647.8 0 0 647.8
Cost reduction (%) 0.11% 0 0 0.11%
Abbreviation: ST‐AC‐OPF, sub‐transmission AC optimal power flow.

*SCADA: Supervisory Control and Data Acquisition

10 of 17 TAVAKKOLI AND AMJADY

3.2 | A real‐world STN

The second test case of this paper, illustrated in Figure 3, is a real‐world STN that is a part of Tehran STN in Iran. In Figure 3
, upstream 400 kV and 230 kV transmission substations are indicated by purple and red colors, respectively. The 63 kV STN
is shown by blue color in this figure and 20 kV distribution side of 63 kV/20 kV transformers is shown by yellow color. This
is the coloring standard used in the Iran's power system dispatch centers. Dashed and solid lines in Figure 3 show under-
ground cables and overhead lines, respectively. Because of limited space, 20 kV primary distribution feeders as well as their
connections and switches cannot be shown in this figure, since this test case has 738 primary distribution feeders with
thousands of connections and switches among them. However, these primary distribution feeders as well as their connec-
tions and switches in this test case are similar to the typical ones shown in Figure 2. Likewise, DG units and shunt com-
pensators, which are connected to 20 kV busses in this test case, cannot be shown in Figure 3. Indeed, this figure shows
the map of STN used in the topology processor of the sub‐transmission dispatch center, which is the network map typically
observed by the sub‐transmission dispatchers. The number of transformers and their capacities for each substation are
shown in Figure 3. For instance, 2 × 30 indicates two 30MVA transformers. The network data of this test case is provided
in Table 3. The total active and reactive power consumptions of this test case (based on the snapshot taken from the sub‐
transmission SCADA system) are 2538.171 MW and 516.797MVAR, respectively.
The OPF results of this test system are presented in two numerical experiments: without and with considering
branch and substation capacity limits, ie, without and with considering constraints (11) and (14). The results of the cur-
rent and proposed ST‐AC‐OPF tools in these two numerical experiments are shown in Tables 4 and 5, respectively. To
evaluate the impact of different decision variables on the objective function components, each of these two numerical
experiments has been performed in four cases: (a) only purchased power is considered; (b) purchased power and DG
units are considered; (c) purchased power, DG units, and tap changers are considered; and (d) purchased power, DG
units, tap changers, and shunt compensators are considered. These four cases are indicated in the first column of
Tables 4 and 5. This test system has only five small DG units with a total capacity of 41 MW. Since DG costs are lower
than the cost of purchased power in the considered snapshot, full capacity of DG units is used in Cases 2 to 4 which
include DG. For this reason, the cost of DG (ie, Cost3) in all Cases 2 to 4 in Tables 4 and 5 is the same. As branch
and substation capacity limits are not considered in the numerical experiment of Table 4, no overload and so no load
shedding occurs in the four cases of Table 4. Thus, the cost of load shedding (ie, Cost2) is zero in Table 4. The other
component of the objective function, ie, the cost of purchased power or Cost1, changes in Table 4. This table shows that
this component as well as the total cost of the both current and proposed ST‐AC‐OPF tools decrease by extending the
decision variables from Case 1 to Case 4 since a better operating point can be obtained. In addition, it is seen that the

FIGURE 3 A part of Tehran transmission/sub‐transmission grid in Iran

TAVAKKOLI AND AMJADY 11 of 17

TABLE 3 Network data of the real‐world test case of Figure 3

Type of Equipment Number

Transmission substation (400 kV/63 kV) 1

Transmission substation (230 kV/63 kV) 11
Sub‐transmission substation (63 kV/20 kV) 78
Transmission transformer (400 kV/63 kV) 2
Transmission transformer (230 kV/63 kV) 30
Sub‐transmission transformer (63 kV/20 kV) 170
Sub‐transmission line (63 kV) 137
Bus (63 kV and 20 kV) 306
Primary distribution feeder (20 kV) 738
Distributed generation (20 kV) 5
Shunt compensator (20 kV) 59

proposed ST‐AC‐OPF tool has lower cost of purchased power and lower total cost than the current ST‐AC‐OPF tool in
all four cases of Table 4, which indicates higher effectiveness of the proposed ST‐AC‐OPF.
By considering branch and substation capacity limits in the numerical experiment of Table 5, the current ST‐AC‐OPF
tool requires load shedding to cope with overloads. Thus, Cost2 and total cost of the current ST‐AC‐OPF significantly
increase in all four cases of Table 5. However, the proposed ST‐AC‐OPF by using LT can produce a more balanced load-
ing condition and avoid load shedding. Table 5 shows that the load shedding cost of the proposed ST‐AC‐OPF is zero in
all four cases of Table 5. Although, by extending the decision variables from Case 1 to Case 4, the load shedding cost of
the current ST‐AC‐OPF slightly decreases (except for Case 2 with respect to Case 1, since DG units are not in the busses
encountering load shedding), the total cost of the proposed ST‐AC‐OPF is significantly lower than the total cost of the

TABLE 4 OPF results without considering branch and substation capacity limits

Case OPF model Cost1[$] (×105) Cost2 [$] (×104) Cost3 [$] Total cost (OF) [$] (×105)

1 Current 1.2980 0 0 1.2980

Proposed (convex) 1.2544 0 0 1.2544
2 Current 1.2766 0 1476 1.2914
Proposed (convex) 1.2394 0 1476 1.2542
3 Current 1.2729 0 1476 1.2877
Proposed (convex) 1.2330 0 1476 1.2478
4 Current 1.2725 0 1476 1.2873
Proposed (convex) 1.2317 0 1476 1.2465

Abbreviation: OPF, optimal power flow

TABLE 5 OPF results with considering branch and substation capacity limits

Case OPF Model Cost1[$] (×105) Cost2 [$] (×104) Cost3 [$] Total cost (OF) [$] (×105)

1 Current 1.2539 7.8887 0 2.0428

Proposed (convex) 1.2616 0 0 1.2616
2 Current 1.2325 7.8887 1476 2.0361
Proposed (convex) 1.2463 0 1476 1.2611
3 Current 1.2304 7.6962 1476 2.0148
Proposed (convex) 1.2403 0 1476 1.2551
4 Current 1.2326 7.3459 1476 1.9820
Proposed (convex) 1.2387 0 1476 1.2535

Abbreviation: OPF, optimal power flow

12 of 17 TAVAKKOLI AND AMJADY

TABLE 6 Cost reduction of proposed OPF with respect to current OPF

Numerical Experiment Case Cost1 (%) Cost 2 (%) Cost 3 (%) Total Cost (OF) (%)

Without considering branch and substation capacity limits 1 3.36 0 0 3.36

2 2.91 0 0 2.88
3 3.14 0 0 3.10
4 3.21 0 0 3.17

With considering branch and substation capacity limits 1 0.61 100 0 38.24
2 1.12 100 0 38.06
3 0.81 100 0 37.71
4 0.50 100 0 36.76

Abbreviation: OPF, optimal power flow

current OPF tool in all four cases of Table 5. Therefore, in the more realistic test conditions of Table 5, where the capac-
ity limits are considered, higher effectiveness of the proposed ST‐AC‐OPF is highlighted.
To better illustrate the difference between the current and proposed ST‐AC‐OPF tools, the cost reductions of the pro-
posed ST‐AC‐OPF with respect to the current ST‐AC‐OPF in all four cases of the both numerical experiments are pre-
sented in Table 6. In this table, downward green and upward red arrows indicate the cost decrease and increase of the
proposed ST‐AC‐OPF with respect to the current ST‐AC‐OPF, respectively. Without considering the capacity limits,
Cost1 and total cost of the proposed ST‐AC‐OPF are 2.91% to 3.36% and 2.88% to 3.36% lower than Cost1 and total cost
of the current ST‐AC‐OPF, respectively. With considering the capacity limits, Cost1 of the proposed ST‐AC‐OPF slightly
increases (from 0.50% to 1.12%) with respect to the current ST‐AC‐OPF, since the proposed ST‐AC‐OPF slightly pur-
chases more power from the upstream grid in these four cases. However, by creating a more balanced loading condition
and purchasing slightly more power, the proposed ST‐AC‐OPF can eliminate the entire load shedding (ie, 100%)
required by the current ST‐AC‐OPF in all four cases of this numerical experiment. Thus, the total cost of the proposed
ST‐AC‐OPF is significantly lower (from 36.76% to 38.24%) than the total cost of the current ST‐AC‐OPF in these four
cases. It is noted that the STN has radial structure, as shown in Figure 3, to limit the short circuit level. Hence, if a
sub‐transmission substation or its lines/transformers become overloaded, the current ST‐AC‐OPF should apply load
shedding to save the STN security. However, the proposed ST‐AC‐OPF can transfer load from an overloaded sub‐
transmission substation to another one with free capacity through the primary distribution switching actions and supply
the additional load by purchasing more power from the upstream grid without applying load shedding.
The results presented so far for the proposed ST‐AC‐OPF have been obtained with the signomial convexification.
The computation times of the proposed ST‐AC‐OPF without/with the convexification are significantly different. The
computation times of the current ST‐AC‐OPF as well as the computation times of the proposed ST‐AC‐OPF
without/with the convexification are shown in Table 7 for the four cases with considering branch and substation
capacity limits. All computation times of Table 7 have been measured on a windows‐based PC with CPU clocking
at 2.4 GHz and 8 GB RAM. Table 7 shows that the computation times of the proposed ST‐AC‐OPF with convexification
are significantly lower than the computation times of the proposed ST‐AC‐OPF without convexification. In addition,
the difference between the computation times of the proposed ST‐AC‐OPF without/with convexification increases by
increasing the problem complexity from Case 1 to Case 4. A similar trend is observed between the computation times

TABLE 7 Computation time (min:sec)

Proposed OPF
Current
Case OPF Nonconvex Convex

1 2:05 3:00 2:19

2 2:32 3:21 2:22
3 4:05 7:38 2:27
4 6:26 8:25 2:55

Abbreviation: OPF, optimal power flow

TAVAKKOLI AND AMJADY 13 of 17

TABLE 8 Error analysis of the signomial convexification method

Case OPF model NRMSE (%)

1 Active power (Pbij) 0.175

Reactive power (Qbij) 0.715
2 Active power (Pbij) 0.158
Reactive power (Qbij) 0.682
3 Active power (Pbij) 0.148
Reactive power (Qbij) 0.675
4 Active power (Pbij) 0.150
Reactive power (Qbij) 0.701

Abbreviation: OPF, optimal power flow; NRMSE, normalized root mean square error.

of the proposed ST‐AC‐OPF with convexification and the current ST‐AC‐OPF. Except for Case 1 that the computation
time of the proposed ST‐AC‐OPF with convexification is slightly higher than the computation time of the current ST‐
AC‐OPF, for the next three cases, the computation time of the proposed ST‐AC‐OPF with convexification is lower than
the computation time of the current ST‐AC‐OPF. Although the proposed ST‐AC‐OPF has additional decision variables
of primary distribution switching actions, the applied signomial convexification makes it more computationally tracta-
ble than the current ST‐AC‐OPF. In addition, it is seen that the computation time of the proposed ST‐AC‐OPF with
convexification for the most complex test, Case 4, on the real‐world STN is less than 3 minutes, which is a reasonable
computation time.
In Table 8, the accuracy of the signomial convexification method is evaluated. For this purpose, after solving the pro-
posed ST‐AC‐OPF model with convexification and obtaining the optimal values of all variables, the original nonconvex
AC power flow equations are used to calculate the actual values. More specifically, Vi, Vj, and δij values obtained from
the ST‐AC‐OPF model with convexification are substituted in the original nonconvex Equations (7) to (10) to obtain the
actual Pbij and Qbij values. The difference between these actual Pbij and Qbij values and the Pbij and Qbij values obtained
from the convexified model specifies the convexification error. These convexification errors in terms of normalized root
mean square error (NRMSE) criterion for the active and reactive powers in the four cases of Table 7 (which consider
branch and substation capacity limits) are reported in Table 8. The well‐known NRMSE criterion is defined as42
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  act cal 2
u 1 ∑NB l¼1 x l −x l
NRMSEð%Þ ¼ 100 × t ×   ; (37)
act 2
max −x min
NB x act

where NB is the number of branches (ie, the number of Pbij or Qbij variables); x act l and x cal
l are the lth actual value
obtained from the original nonconvex model and the lth calculated value obtained from the convexified model, respec-
tively; and x act act
min and x max indicate the range of the actual values. From Table 8, it is seen that the convexification errors
of the active powers are in the range of 0.148% to 0.175% and the convexification errors of the reactive powers are in the
range of 0.675% to 0.715%. Thus, all convexification errors are significantly less than 1%, which indicate the high accu-
racy of the signomial convexification. This high accuracy along with the high computational efficiency (reported in
Table 7) illustrate the effectiveness of the signomial convexification and justify its application to solve the proposed
ST‐AC‐OPF model.

4 | CONCLUSION

In this paper, a new AC OPF tool for STNs has been presented. Unlike currently available sub‐transmission OPF tools
that consider distribution feeders' loads constant, the proposed OPF tool can model and optimize LT between primary
distribution feeders. Using this concept, the proposed ST‐AC‐OPF tool can optimize feasible distribution switching
actions along with optimizing sub‐transmission control actions while avoiding infeasible distribution switching actions.
In this way, the proposed OPF tool can obtain a more optimal operating point for an STN than current sub‐transmission
OPF tools. Primary distribution switching actions have been considered as additional decision variables in the proposed
OPF tool, since primary distribution feeders usually have several sections, several connected loads, and several control-
lable switches (unlike sub‐transmission feeders), which provide high flexibility for the proposed ST‐AC‐OPF. In
14 of 17 TAVAKKOLI AND AMJADY

addition, a signomial convexification has been applied to the proposed ST‐AC‐OPF model to make it computationally
tractable. Higher effectiveness of the proposed ST‐AC‐OPF tool compared with a practical widely used ST‐AC‐OPF tool
in terms of operation cost, load shedding, and even computation burden has been illustrated. The numerical compari-
sons of this paper have been presented for both a practical example as well as a real‐world STN. The main reason of the
higher effectiveness of the proposed ST‐AC‐OPF tool is optimizing distribution switching actions along with optimizing
sub‐transmission control actions to create a more balanced loading condition and to avoid overload in both sub‐
transmission and primary distribution networks.
The LT concept as well as optimizing primary distribution switching actions introduced in this research work are
conceptually different from transmission switching capability recently presented in some transmission operation and
planning research works. In addition, the proposed ST‐AC‐OPF tool, which can provide higher flexibility for optimizing
the operating point of both sub‐transmission and primary distribution grids, has a different point of view and modeling
approach compared with distribution switching works.

NOMENCLATURE

Indices and sets

i,j Indices of busses

n Index of switches
ΨB Set of sub‐transmission boundary busses connected to transmission grid
ΩB Set of all sub‐transmission and primary distribution busses

Parameters

Bij Susceptance of branch i − j

Cgi Generation cost of DG located at bus i
CLS Cost of load shedding
EPi Electricity price at sub‐transmission bus i
Gij Conductance of branch i − j. i − j
Pdi Active power demand at bus i without considering primary distribution switching
i ; Pgi
Pgmin max
Minimum and maximum limits of Pgi
PFdi,PFgi Power factor of load and DG at bus i, respectively
Qcmini ; Qci
max
Minimum and maximum limits of Qci
Qgi ; Qgmax
min
i Minimum and maximum limits of reactive power generated by DG located at bus i
Smax
A ; S max
B Max apparent power capacity of primary distribution feeders A and B, respectively
Smax
i Max apparent power capacity of sub‐transmission boundary bus i
Sbmax
ij Max apparent power flow limit of branch i − j
ij ; tcij
tcmin max
Minimum and maximum limits of tcij
V i ; V max
min
i Minimum and maximum limits of Vi

Variables

LSi Load shedding in bus i

OF Objective function of OPF
Pi Active power purchased from the upstream grid (ie, transmission network) at bus i
PdSA ; PdSB Active loads of primary distribution feeders A and B, considering distribution switching, respectively
PdSi Active power demand at bus i with considering primary distribution switching
Pbij Active power flow of branch i − j
TAVAKKOLI AND AMJADY 15 of 17

Pgi Active power generated by DG located at bus i

Qi Reactive power received from the upstream grid (ie, transmission network) at bus i
Qbij Reactive power flow of branch i − j
Qci Reactive power of capacitor located at bus i
|SA|, |SB| Apparent power flow of primary distribution feeders A and B, respectively
Sn Binary variable indicating status of switch n (1: Closed, 0: Open)
tcij Tap setting of tap‐changing transformer i − j
Vi Voltage magnitude at bus i
δi Voltage angle at bus i
δij Phase angle difference between busses i and j

L I S T O F S Y M B O LS A N D A B B R E V I A T I O N S

AC‐OPF Alternating Current ‐Optimal Power Flow

CB Circuit Breaker
DG Distributed Generation
GAMS Generalized Algebraic Modelling System
LMP Locational Marginal Price
LT Load Transferring
MINLP Mixed‐Integer Non‐linear Programming
NRMSE Normalized Root Mean Square Error
OPF Optimal Power Flow
SCADA Supervisory Control and Data Acquisition
ST‐AC‐OPF Sub‐transmission‐ Alternating Current ‐Optimal Power Flow
STN Sub‐transmission Network

ORCID
Nima Amjady https://orcid.org/0000-0003-1308-1738

R EF E RE N C E S
1. Pandya KS, Joshi SK. A survey of optimal power flow methods. J Theor Appl Inf Technol. 2008;4(5):450‐458.
2. Maskar MB, Thorat AR, Korachgaon I. A review on optimal power flow problem and solution methodologies. ICDMAI Conf; 2017: 64‐70.
3. Amjady N, Ansari M. Non‐convex security constrained optimal power flow by a new solution method composed of Benders decomposi-
tion and special ordered sets. Int Trans Electr Energ Syst. 2013;24(6):842‐857.
4. Attarha A, Amjady N. Solution of security constrained optimal power flow for large‐scale power systems by convex transformation tech-
niques and Taylor series. IET Gener Transm Distrib. 2016;10(4):889‐896. https://doi.org/10.1049/iet‐gtd.2015.0494
5. AlRashidi MR, El‐Hawary ME. Applications of computational intelligence techniques for solving the revived optimal power flow prob-
lem. Electr Pow Syst Res. 2009;79(4):694‐702. https://doi.org/10.1016/j.epsr.2008.10.004
6. Reddy SS. Optimal power flow using hybrid differential evolution and harmony search algorithm. Int J Mach Learn Cyber. 2018;10:1‐15.
7. Glavitsch H. State of the art review—switching as means of control in the power system. Int J Elect Power Energy Syst. 1985;7(2):92‐100.
https://doi.org/10.1016/0142‐0615(85)90014‐6
8. Shao W, Vittal V. Corrective switching algorithm for relieving overloads and voltage violations. IEEE Trans Power Syst.
2005;20(4):1877‐1885. https://doi.org/10.1109/TPWRS.2005.857931
9. Granelli G, Montagna M, Zanellini F, Bresesti P, Vailati R, Innorta M. Optimal network reconfiguration for congestion management by
deterministic and genetic algorithms. Electr Pow Syst Res. 2006;76(6‐7):549‐556. https://doi.org/10.1016/j.epsr.2005.09.014
10. Fisher EB, O'Neill RP, Ferris MC. Optimal transmission switching. IEEE Trans Power Syst. 2008;23(3):1346‐1355. https://doi.org/10.1109/
TPWRS.2008.922256
16 of 17 TAVAKKOLI AND AMJADY

11. Hedman KW, O'Neill RP, Fisher EB, Oren SS. Optimal transmission switching‐sensitivity analysis and extensions. IEEE Trans Power Syst.
2008;23(3):1469‐1479. https://doi.org/10.1109/TPWRS.2008.926411
12. Hedman KW, O'Neill RP, Fisher EB, Oren SS. Optimal transmission switching with contingency analysis. IEEE Trans Power Delivery.
2009;24(3):1577‐1586.
13. Hedman KW, Ferris MC, O'Neill RP, et al. Co‐optimization of generation unit commitment and transmission switching with N‐1 reliabil-
ity. IEEE Trans Power Syst. 2010;25(2):1052‐1063. https://doi.org/10.1109/TPWRS.2009.2037232
14. Khanabadi M, Ghasemi H, Doostizadeh M. Optimal transmission switching considering voltage security and N‐1 contingency analysis.
IEEE Trans Power Syst. 2013;28(1):542‐550. https://doi.org/10.1109/TPWRS.2012.2207464
15. Hedman KW, Oren SS, O'Neill RP. Optimal transmission switching: economic efficiency and market implications. J Regul Econ.
2011;40(2):111‐140. https://doi.org/10.1007/s11149‐011‐9158‐z
16. Dehghanian P, Kezunovic M. Impact assessment of transmission line switching on system reliability performance. 18th Int Conf Intell
Syst Appl Power Syst (ISAP); 2015: 1‐6.
17. Reddy S, Bijwe PR. Multi‐objective optimal power flow using efficient evolutionary algorithm. Int J Emerg Electr Power Syst.
2017;18(2):1‐21.
18. Bacher R, Glavitsch H. Loss reduction by network switching. IEEE Trans Power Syst. 1988;3(2):447‐454. https://doi.org/10.1109/59.192895
19. Fliscounakis S, Zaoui F, Simeant G, et al. Topology influence on loss reduction as a mixed integer linear programming problem. IEEE
Power Tech Conf; 2007: 1987‐1990.
20. Gomes FV, Carneirio S, Pereira JLR, et al. A new distribution system reconfiguration approach using optimum power flow and sensitivity
analysis for loss reduction. IEEE Trans Power Syst. 2006;21(4):1616‐1623. https://doi.org/10.1109/TPWRS.2006.879290
21. Khodr HM, Crespo J, Matos MA, et al. Distribution systems reconfiguration based on OPF using Benders decomposition. IEEE Trans
Power Delivery. 2009;24(4):2166‐2176.
22. Oliveira LW, Carneirio S, Oliveira J, et al. Optimal reconfiguration and capacitor allocation in radial distribution systems for energy losses
minimization. Int J Elect Power Energy Syst. 2010;32(8):840‐848. https://doi.org/10.1016/j.ijepes.2010.01.030
23. Zidan A, El‐Saadany EF. Multi‐objective Network Reconfiguration in Balanced Distribution Systems with Variable Demand. 2nd Int.
Conf. (EPECS), Sharjah, United Arab Emirates; 2011: 1‐6.
24. Alonso FR, Oliveira DQ, Zambroni de Souza AC. Artificial immune systems optimization approach for multiobjective distribution system
reconfiguration. IEEE Trans Power Syst. 2015;30(2):840‐847. https://doi.org/10.1109/TPWRS.2014.2330628
25. Bernardon DP, Garcia VJ, Ferreira ASQ, Canha LN. Multicriteria distribution network reconfiguration considering subtransmission anal-
ysis. IEEE Trans Power Delivery. 2010;25(4):2684‐2691. https://doi.org/10.1109/TPWRD.2010.2041013
26. Lee C, Huang C. Analysis of load transfer on radial distribution systems with tie lines. Eur T Electr Power. 2007;17(3):268‐284. https://doi.
org/10.1002/etep.136
27. Liao H, Liu D, Huang Y, et al. Load transfer capability analysis considering interconnection of distributed generation and energy storage
system. Eur T Electr Power. 2012;24(2):166‐177.
28. Ma J, Ma W, Qiu Y, Yan X, Wang Z. Load transfer strategy based on power transfer capability for main‐transformer fault. Int Trans Electr
Energ Syst. 2015;25(12):3439‐3448. https://doi.org/10.1002/etep.2044
29. Jalali M, Zare K, Hagh MT. A multi‐stage MINLP‐based model for sub‐transmission system expansion planning considering the place-
ment of DG units. Int J Elect Power Energy Syst. 2014;63:8‐16. https://doi.org/10.1016/j.ijepes.2014.05.044
30. ALSTOM T&D SA. TREC‐ IRAN SCADA manual. Paris, France; 2005.
31. Rahmani S, Amjady N. A new optimal power flow approach for wind energy integrated power systems energy. Energy. 2017;134:349‐359.
https://doi.org/10.1016/j.energy.2017.06.046
32. Sharifzadeh H, Amjady N. Stochastic security constrained optimal power flow incorporating preventive and corrective actions. Int Trans
Electr Energ Syst. 2016;26(11):2337‐2352. https://doi.org/10.1002/etep.2207
33. Sharifzadeh H, Amjady N, Zareipour H. Multi‐period stochastic security‐constrained OPF considering the uncertainty sources
of wind power, load demand and equipment unavailability. Electr Pow Syst Res. 2017;146:33‐42. https://doi.org/10.1016/j.
epsr.2017.01.011
34. Attarha A, Amjady N, Conejo AJ. Adaptive robust AC optimal power flow considering load and wind power uncertainties. Int J Elect
Power Energy Syst. 2018;96:132‐142. https://doi.org/10.1016/j.ijepes.2017.09.037
35. Karbalaei F, Shahbazi H, Mahdavi M. A new method for solving preventive security‐constrained optimal power flow based on linear net-
work compression. Int J Electr Power Energy Syst. 2018;96:23‐29. https://doi.org/10.1016/j.ijepes.2017.09.023
36. AlRashidi MR, AlHajri MF. Optimal planning of multiple distributed generation sources in distribution networks: a new approach. Energ
Conver Manage. 2011;52(11):3301‐3308. https://doi.org/10.1016/j.enconman.2011.06.001
37. Lundell A, Westerlund T. Optimization of transformations for convex relaxations of MINLP problems containing signomial functions.
Comput Aid Chem Eng. 2009;27:231‐236.
38. Lundell A, Westerlund J, Westerlund T. Some transformation techniques with applications in global optimization. J Glob Optim.
2009;43(2‐3):391‐405. https://doi.org/10.1007/s10898‐007‐9223‐4
TAVAKKOLI AND AMJADY 17 of 17

39. Boyd S, Vandenberghe L. Convex optimization. Cambridge University Press; 2004.

40. GAMS. (2018). Generalized Algebraic Modelling System. [online], Available at http:// www.gams.com.
41. Ahmadigorji M, Amjady N. A multiyear DG‐incorporated framework for expansion planning of distribution networks using binary cha-
otic shark smell optimization algorithm. Energy. 2016;102:199‐215. https://doi.org/10.1016/j.energy.2016.02.088
42. Wackerly D, Mendenhall W, Scheaffer R. Mathematical Statistics With Applications. Belmont, CA, USA: Cengage Learning; 2007.

How to cite this article: Tavakkoli MA, Amjady N. A new AC OPF tool for sub‐transmission networks
considering distribution switching actions and load‐transferring capability. Int Trans Electr Energ Syst. 2019;29:
e12029. https://doi.org/10.1002/2050‐7038.12029