Electric Power Systems Research 171 (2019) 26–35

Contents lists available at ScienceDirect

Electric Power Systems Research

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

AC load flow based model for transmission expansion planning T

a a b,⁎
Mahmoud Ali Farrag , Kareem Mohamed Ali , Shaimaa Omran
a
Electrical Power & Machines Department, Faculty of Engineering, Cairo University, 1 Gamaa Street, 12613, Cairo, Egypt
b
Systems and Information Department, Engineering Research Division, National Research Centre, El-Buhouth street, 12622, Dokki, Cairo, Egypt

ARTICLE INFO ABSTRACT

Keywords: This paper proposes an accurate model to solve the transmission network expansion planning (TNEP) problem.
Transmission network expansion planning The TNEP problem is solved using a linear DC model, a mixed integer linear DC model, and the proposed
Mixed integer linear programming accurate exact mixed integer non-linear AC model. A 19 bus system which is a part of the 220 kV Egyptian
Mixed integer non-linear programming transmission network is used for the verification of the proposed load flow models. The proposed exact non-
DC load flow model
linear AC load flow based model that considered control of active and reactive power flow in the network and
AC load flow model
considered the power losses in the system yielded promising results when compared to DC load flow based linear
and mixed integer linear models. It provided a planned network with least number of added lines and con-
secutively with the least cost.

1. Introduction the power flow over transmission lines. Reference [2] proposed a
methodology that is based on DC power flow to optimize the expansion
Electrical power systems face challenges to supply the energy de- planning problem of the transmission network under constrained La-
mand through sustainable, environment friendly, and affordable energy grangian multiplier and the locational marginal price. A mixed-integer
resources. To overcome such challenges the integration of renewable programming model is used to solve the problem. The New England 39
energy resources and implementation of demand response resources are bus test power system was used for verification. In Reference [3] the
adopted in smart grids. The transmission network expansion planning transmission expansion planning problem was represented as a mixed
(TNEP) problem is the determination of the transmission lines that integer linear programming DC model that took into consideration the
should be added to the existing power grid to face the increase in de- N-1 security criterion. Afterwards, it introduced equivalency techniques
mand level and associated generation level in an economic manner to reduce the size of the large power system and thus enhance the
while satisfying the different quality constraints imposed on the computational performance when solving the transmission expansion
transmission network. The TNEP optimization problem has been solved planning. These techniques are an admittance equivalency method
using classical optimization techniques [1–5], and heuristic techniques known as WARD method and the Radial equivalent independent (REI)
[6–12]. Some recent papers [6–8,11,12] have tackled the transmission method. The proposed algorithms were simulated on IEEE 24-bus
optimization problem considering renewable energy penetration and system. Reference [4] presented an approach for transmission expan-
demand side management. The approximated and simplified DC models sion planning problem considering short circuit currents constraints
are typically used to solve the TNEP problem. The DC based model is a and N-1 security constraints. A linearized DC power flow model was
simplification of the AC based model, where the transmission losses and proposed and the IEEE 118-bus system was used for verification.
reactive power are neglected. Moreover, in DC model the voltage The DC based models faced drawbacks in handling the TNEP pro-
magnitude of all buses is assumed 1 p.u. and the buses voltage phase blems. Using DC model in many studies has resulted in sub-optimal
angle is considered to be very small [13,14]. Many references applied solutions rather than optimal ones [15]. Thus, using the AC load flow
the DC based model; some of these are References [1–4]. Reference [1] based model is expected to yield optimal solutions for most planning
formulated the transmission network expansion planning problem as a problems, as the AC load flow model is capable of considering the effect
mixed integer linear programming problem, using a model that avoids of injected reactive power on the planning process as well as the power
the nonlinearity found in the KVL for candidate lines. In this work the system losses [1,5,12,14,16–18]. Most researchers utilized the AC load
FACTS and the distributed FACTS were added to the system to control flow in its linearized form to get better solution results for the TNEP

⁎
Corresponding author at: Systems and Information Department, Engineering Research Division, National Research Centre, El-Buhouth Street, 12622, Dokki,
Cairo, Egypt.
E-mail address: [email protected] (S. Omran).

https://doi.org/10.1016/j.epsr.2019.02.006
Received 9 August 2018; Received in revised form 1 January 2019; Accepted 6 February 2019
Available online 16 February 2019
0378-7796/ © 2019 Elsevier B.V. All rights reserved.
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Nomenclature K̄ Large positive number

w(l) Number of new lines added in a loop
F Capital cost function of the new added lines qki Real power flow on line k from bus i
Ck Specific capital cost of line k qki Reactive power flow on line k from bus i
Pk Real power flow on new line k gk Series admittance of line k (real part)
P100+k Real power flow on new line k in the opposite direction bk Series admittance of line k (imaginary part)
M1 Number of proposed lines vi Voltage magnitude at bus i
M2 Number of existing lines i Voltage angle at bus i
M Total number of lines (proposed and existing) Vmin Minimum voltage magnitude permitted
Lk Length of line k Vmax Maximum voltage magnitude permitted
PIi Real power injected at bus i min Minimum line angle permitted
QIi Reactive power injected at bus i max Maximum line angle permitted
xk Reactance of line k k1 (i) Set of lines connected to bus i
Pmaxk Maximum real power permitted on line k k2 (i) Set of existing lines found in basic loop i
Smaxk Maximum apparent power permitted on line k k3 (i) Set of existing and new lines found in loop i
Zk Zero-one integer variable related to line k

problem such as these References [5,14,16–18]. line type and size), the cost function may be represented by line length
The contribution of this paper is the presentation and formulation of instead of line cost as follows
an exact nonlinear AC load flow model to solve the TNEP problem. This M1
proposed AC load flow based model considers active and reactive power F= i=1
Lk (Pk + P100 + k ) (2)
flow and power losses in the system. The TNEP problem is solved using
a linear DC model, a mixed integer linear DC model, and the proposed After obtaining the solution for the network, the power flow value
exact mixed integer non-linear AC model. These models are im- and direction will follow one of these scenarios:
plemented on a 19 bus system deduced from the 220 kV Egyptian
transmission network for verification. The used load flow models are 1 Pk has a positive value and P100+ k has zero value. This means that the
applied for a given scenario of predicted load values as well as assumed power flow direction is in forward direction which is the same di-
optimal generation planning which means they are applied for a certain rection considered in nodes and loops equations.
load generation pattern at a target year. A comparison between the 2 P100+ k has a positive value and Pk has zero value. This means that the
three models is presented. The following sections are organized as power flow direction is in reverse direction which is the opposite
follows: The mathematical formulations of the three models adopted in direction considered in nodes and loops equations.
this study are presented in Section 2. The characteristics of the 19 bus 3 Pk and P100+ k have positive values. This means that the power flow
system are described in Section 3. The implementation of the proposed value is the difference between Pk and P100+ k and the flow is in the
techniques is conveyed in Section 4. The results and findings of this direction of the larger value.
work are discussed in Section 5. The paper is concluded in Section 6.
The linear constraints are the line limit equations, and the real
power flow equations. The real power flow equations are obtained from
2. Mathematical formulation
Kirchhoff's laws while neglecting the lines resistances. Thus the linear
equations describing the model are represented by these three con-
The TNEP problem is formulated as an optimization problem with
straints:
an objective function and a set of constraints. The objective function
represents the cost function of the transmission lines of the system. The
constraints are the equations that represent the physical, quality and • Power balance equations at bus i:
logical limitations imposed on the power transmission network. These M
PIi = (Pk P100 +k) , k= 1,2, …M (3)
constraints represent the power flow equations and the line limit k k1(i)

equations. The mathematical formulations of the three models used for

solving the TNEP problem; the linear DC load flow based model, the • Loop equation for loops having only existing lines:
mixed integer linear DC load flow based model, and the mixed integer
non-linear AC load flow based model are presented in the following For loop L
subsections. M2
k k2 (i)
xk (Pk P100+ k ) = 0 (4)
2.1. Linear DC load flow based model

For developing a linear programming based planning model, the

• Lines limit equations:
following points are considered: (Pk + P100+ k ) Pmaxk , k = 1,2, …M (5)

a line cost is assumed linear with power flow,

2.2. Mixed integer linear DC load flow based model
b all constrained equations should be linear,
c the set of utilized constraints should be correct for either adding or
The previous model suffers a serious drawback which is ignoring the
neglecting any new line [19,20].
loops having one or more new lines. Considering loops that have new
lines is necessary. In the previous model if a new line is deleted, the
The cost function for this model is represented as follows
loop equation will be incorrect and thus leads to incorrect results. In
F=
M1
Ck (Pk + P100+ k ) this model integer variables (0–1 variables) are introduced to overcome
(1)
i=1
this drawback as follows [19,20].
For constant cost of transmission line per unit length (assuming one The capital cost of new line is

27
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

M1 M1
F= k=1
Ck Zk (6) F= k=1
Ck Zk (14)

Again, for constant cost of transmission line per unit length, the cost This objective function is minimized subject to the following set of
function may be represented by line length instead of line cost as fol- constraints
lows
M1
2.3.1. AC load flow constraints
F= k=1
Lk Zk (7)
a Existing line equations
This objective function F is minimized subject to the following
linear constraints For each existing line k between buses i and j, four equations are
developed for active and reactive power flows
• Power balance equations at bus k: Pki = gk vi2 vi vj (gk cos( i j) + bk sin( i j )) (15)
M
PIi = k k1 (i)
(Pk P100 +k ) , k= 1,2, …M (8)
qki = bk vi2 vi vj (gk sin( i j) bk cos( i j )) (16)

• Loop equation for loops having only existing lines: Pkj = gk v j2 vj vi (gk cos( j i) + bk sin( j i )) (17)

For loop L qkj = bk v 2j vj vi (gk sin( j i) bk cos( j i )) (18)

M2
xk (Pk P100+ k ) = 0 k = M1 + 1, ……M
k k2 (i) (9)

• Loop equations for loops having both existing and w(l) new lines: • Proposed line equations
xk (Pk P100+ k ) + xk (Pk P100+ k ) K¯ (Zk w (l)) For each proposed line k between buses i and j, four equations are
k k2 (i) k k3 (i) (10) developed for active and reactive power flows

xk (Pk P100 + k ) + xk (Pk P100+ k ) K¯ (w (l) Zk ) Pki = gk Zk vi2 vi vj (gk Zk cos( i j) + bk Zk sin( i j )) (19)
k k2 (i) k k3 (i) (11)
qki = bk Zk vi2 vi vj (gk Zk sin( i j) bk Zk cos( i j )) (20)
where K̄ is a large positive value. This means that when line k is added
at Zk = 1 then the voltage drop across the lines of this loop is 0, which is Pkj = gk Zk v 2j vj vi (gk Zk cos( j i) + bk Zk sin( j i )) (21)
a correct loop equation. If line k is deleted at Zk = 0 then the voltage
drop across the lines of this loop will take a value between K̄ up to K̄ , qkj = bk Zk v 2j vj vi (gk Zk sin( j i) bk Zk cos( j i )) (22)
this means that there is no constraint on the voltage which is realistic as
k = 1,2, …M1
this will be an open loop.
The modification of loop equations having new lines is made pos-
sible, irrespective of line addition or deletion, by using the zero-one • Node equations
integer variables. When zero-one integer variables are used for loops
For injected active and reactive power at bus i
having any number of new lines, the deletion of any one line will make
it an open loop. PIi = Pki
k k1 (i) (23)

• Lines limit equations: QIi =

k k1 (i)
qki
(24)
(Pk + P100+ k ) Pmaxk , k = 1,2, …M2 (12)

(Pk + P100+ k ) Pmaxk Zk , k = 1,2, …M1 (13) 2.3.2. Line limits constraints
For a new added line k
Thus the cost function now has a value depending on the presence
or absence of the new line. The above model has been derived based on (Pki )2 + (qki )2 S 2maxk Zk (25)
the fact that one line type and size has been used for all new routes (cost
per km is fixed). Also, losses and reactive power of loads and generators (Pkj ) 2 + (qkj ) 2 S 2 maxk Zk (26)
are neglected. The bus voltage magnitude of all buses is assumed fixed For an existing line k
at 1 p.u.
(Pki )2 + (qki )2 S 2maxk (27)
2.3. Mixed integer non-linear AC load flow based model
(Pkj ) 2 + (qkj ) 2 S 2 maxk (28)
The reactive power flow on transmission networks affects largely
the voltage magnitudes which are governed to be within a narrow range 2.3.3. Voltage limits constraints
around the one per unit. In order to avoid voltage problems that are to The voltage at each bus i must be within permissible limits
be solved by addition of new lines or addition of new reactive power
Vmin Vi Vmax (29)
resources, the TNEP problem is to be solved in a comprehensive rea-
listic manner using the exact AC load flow equations [21–25]. The AC
load flow model considers the effect of injected reactive power and the 2.3.4. Stability limits constraints
power system losses on the planning process. The stability limit for an existing line k between buses i and j is
The objective function for the planning model using the exact AC given by
load flow equations is minimizing the capital cost of new lines as fol-
min i j max (30)
lows
Minimize F And for a new line

28
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Fig. 1. The 19 bus test network.

min + ¯ k
min (KZ K¯ ) i j max + ¯
max (K ¯ k)
KZ (31) considered for real power flow.

Where the values for the limits in (29) and (30) are Vmin = 0.9 p.u.,
4. Implementation of the proposed techniques
Vmax = 1.06 p.u., and the stability limit constraints angle values are set
between −30° and +30°.
The implementation of the proposed techniques presented in
Section 2 to the 19 bus system described in Section 3 will be presented
3. The 19 bus system characteristics and descriptions in this section. The LINGO® 17.0 optimization software tool [26] is used
to simulate the three proposed models on an Intel® Core™ i7 @ 1.73
The three mathematical models presented in the previous section processor with a 4 GB installed memory on a 64-bit MS Windows®
for solving the TNEP problem are implemented on a 19 bus system for operating system.
verification. This system is deduced from the 220 kV Egyptian trans-
mission network and it is depicted in Fig.1 with the existing lines 4.1. Linear DC load flow based model
presented as solid lines and all possible routes shown as dotted lines
[20]. It has 19 existing buses and no isolated buses. The number of As mentioned in Section 2.2 the DC load flow model suffers a serious
circuits at any new route is two (fixed). The number of proposed lines is drawback which is ignoring the loops having one or more new lines. To
21 from line 1 till line 21 and the number of existing lines is 26 lines overcome this drawback the new lines will be added iteratively where
from line 22 till line 47. The thermal limit of all lines is set to 0.96 pu. only one line will be added in each iteration. Once a new line is added it
Tables A1 and A2 in the Appendix A present the line data and the bus will be treated as an existing line and if a new loop is formed it will be
data for the 19 bus system. It is shown that the limit of all lines is 1.2 added to the model in the following iterations. Thus the power flow on
p.u. MVA, and in the DC load flow 0.8 of the maximum MVA limit is the new lines will be accurately calculated as the iteration process

29
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Table 1 Table 4
DC load flow results using existing lines at target year. The power flow over the lines for the last iteration.
Line number Power loading Power capacity of line Amount of overload Line number Power flow (p.u.)
(p.u.) (p.u.) (p.u.)
1 0
23 1.1670 0.96 0.207 2 0
24 1.4427 0.96 0.4827 3 0.722
25 1.7255 0.96 0.7655 4 0
33 1.2623 0.96 0.3023 5 0.538
41 1.4977 0.96 0.5377 6 0.220
44 1.3986 0.96 0.4386 7 0.146
8 0.180
9 0.439
10 0.312
Table 2
11 0.535
The power flow over the loaded lines in the first iteration. 12 0.571
Line number Power flow over the line 13 0
14 0.626
1 0 15 0.603
2 0 16 0
3 0.7064246 17 0
4 0 18 0
5 0 19 0
6 0 20 0.766
7 0 21 0
8 0 22 0.753
9 0.4294180 23 0.833
10 0 24 0.842
11 0.0127323 25 0.933
12 0.96 26 0.544
13 0 27 0.622
14 0.5216165 28 0.012
15 0.4043867 29 0.177
16 0 30 0.472
17 0 31 0.303
18 0 32 0.366
19 0 33 0.960
20 0 34 0.520
21 0 35 0.017
36 0.053
37 0.104
38 0.321
Table 3 39 0.417
The added lines in the iterative process of the DC load flow based model. 40 0.523
41 0.742
Iteration number Added line Line length in km (cost of 1 km is 1 p.u.)
42 0.387
1 12 35 43 0.159
2 14 35 44 0.681
3 3 50 45 0.203
4 9 75 46 0.051
5 8 52 47 0.560
6 15 40
7 11 90
8 20 55 due to its addition which have the lines 12, 29, and 23 will be added in
9 7 40 subsequent iterations.
10 10 75
This process is repeated iteratively, and the final added lines are
11 6 50
12 5 45 given in Table 3. It is worth noting that the type of conductor used for
the added lines is an ACSR 2 × 400 mm2. The final power loading over
Total cost 642
all the lines obtained from the last iteration is shown in Table 4.
The DC load flow based model solves the TNEP problem using linear
proceeds. As the new lines will be added one at a time till no more lines programming technique in an iterative manner. The final optimal net-
are needed, the most significant and effective lines should be added work can be obtained fast as the iteration takes few seconds, since this
first. The line having the largest power flow will be given the priority model size is relatively small. Although, there is no guarantee that the
for addition. least cost network has been obtained, the presented method can yield
Application of the DC load flow for the 19 bus network with existing optimal or suboptimal solutions depending on the problem itself and its
lines only and load generation pattern at target year has resulted in size. Yet, the simplicity of the planning method and its high speed en-
overloading some of the existing lines as shown in Table 1. courage the utilization of this method, at least to get initial estimate for
Afterwards, when solving the planning model with all new lines the final planned network.
associated with the network, it is observed that some new lines were
loaded; lines 3, 9, 11, 12, 14, and 15. Table 2 depicts the power flow
4.2. Mixed integer linear DC load flow based model
over these loaded lines in the first iteration, it is shown that line 12 has
the maximum power flow, thus it is given priority for addition. The
The mixed integer linear DC mathematical model discussed in
addition of this lines means that it will be eliminated from the cost
Section 2.2 is applied to the test system presented in Section 3 and
function and treated as an existing line afterwards. Also the loop formed
depicted in Fig. 1. The following was obtained for the model:

30
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Fig. 2. Optimum network obtained using the mixed integer non-linear AC model for a ± 10% voltage limit.

1 Number of integer variables is 21. network with the least number of lines added is depicted in Fig. 2.
2 Number of node equations (power balance equations) is 18. Moreover, the real power flow over the lines obtained by DC load
3 Number of equality loop equations having only existing lines is 8. flow is compared to that obtained by AC load flow to double check the
4 Number of inequality loop equations having new lines is 42. accuracy. First, the full reactive power demand is assumed to be sup-
5 Number of power limit inequality constraints is 47. plied (Qload = full load). As reactive power injection from different
6 Ohmic losses cost is neglected. generators can change without any limits set, the developed model is
applied such that to minimize:
The solution of the mixed integer linear DC mathematical model has
resulted in addition of the following seven lines; 2, 8, 11, 12, 14, 15, 1 The real power loss.
and 16. 2 The bus voltage deviation from 1 p.u. value.

4.3. Mixed integer non-linear AC load flow based model Second, the two minimization problems are solved but this time
assuming that there is no reactive power demand (Qload = 0). Power
The mixed integer non-linear AC model in Section 2.3 is applied to flow results of both cases are represented in Table 5 once for mini-
the 19 bus test system presented in Section 3. Two bus voltage mag- mizing the real power loss and another time for minimizing the bus
nitude limits cases are considered which are (1 ± 0.05) % in p.u. and voltage deviation. As the bus voltages in DC load flow are assumed
(1 ± 0.1) % in p.u. For the ± 5% voltage limits the optimum lines constant at 1 p.u., the voltages obtained with AC load flow and full
added are 2, 3, 11, 12, 14, and 15. Whereas, for the ± 10% voltage reactive power load as well as the reactive power injected from dif-
limit, optimum lines added are 2, 11, 12, and 15. This optimum ferent generators are given in Tables 6 and 7 respectively.

31
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Table 5 Table 7
Real power flow for the DC model vs the AC model for Qload = 0, and Generation reactive power from AC load flow.
Qload = full load.
Reactive power Reactive power for power Reactive power for voltage
Real DC At Qload = full load At Qload = zero injection loss minimization (p.u.) deviation minimization (p.u.)
power based
flow model QG1 0.067 0.472
AC based AC based model AC based AC based model QG1 0.701 1.007
model (OPF for min model (OPF for min QG1 0.604 0.568
(OPF for voltage (OPF for voltage QG1 0.863 1.185
min deviation from 1 min deviation from 1 QG1 1.229 1.549
power p.u. ∑(1 − Vi)2) power p.u. ∑(1 − Vi)2) QG1 1.354 1.424
loss) loss) QG1 1.094 0.957
QGen total 5.849 6.020
P02 0.960 0.671 0.674 0.639 0.679 QLoad total 5.266 5.266
P08 0.960 0.244 0.244 0.257 0.239 QLoss total 0.58 0.761
P11 0.452 0.582 0.582 0.591 0.581 PLoss total 0.112 0.119
P12 0.564 0.594 0.592 0.613 0.592
P14 0.748 0.486 0.490 0.477 0.480
P15 0.712 0.812 0.815 0.843 0.808 5. Results and discussion
P16 0.256 0.380 0.368 0.257 0.363
P22 0.950 0.712 0.723 0.666 0.713
P23 0.890 0.924 0.927 0.953 0.926 A comparison of the application of the three proposed models to the
P24 0.944 0.785 0.788 0.790 0.786 19 bus system is presented in Figs. 3 and 4. Fig. 3 shows the lines added
P25 0.938 0.980 0.973 1.036 0.970 for each mathematical model adopted; moreover, the plot in Fig.4 de-
P26 0.759 0.618 0.622 0.565 0.623
picts the total number of lines added for each model, and the cost as-
P27 0.408 0.557 0.554 0.811 0.552
P28 0.083 0.580 0.047 0.144 0.049
sociated with this addition in monetary unit. The cost is calculated
P29 0.522 0.500 0.511 0.438 0.508 according to the length of each line as shown earlier in Table 1.
P30 0.486 0.471 0.472 0.467 0.471 The mixed integer non-linear programming AC based model yields
P31 0.290 0.307 0.306 0.311 0.306 the least number of lines to be added to the 19 bus system and thus it
P32 0.239 0.368 0.358 0.409 0.369
gives the most optimal network plan. It is observed from the number of
P33 0.960 1.090 1.080 1.136 1.090
P34 0.127 0.577 0.574 0.628 0.567 lines mentioned for each model in Section 4 and depicted in Fig. 3 that
P35 0.258 0.129 0.150 0.233 0.114 there are three lines in common between all tests done which are lines
P36 0.193 0.057 0.056 0.052 0.053 11, 12, and 15. A closer look at the DC based model with its two
P37 0.230 0.328 0.329 0.377 0.321 techniques linear and mixed integer linear reveals that the two models
P38 0.068 0.272 0.275 0.310 0.258
P39 0.365 0.442 0.444 0.317 0.462
resulted in 5 similar lines which are 8, 11, 12, 14, and 15.
P40 0.576 0.506 0.498 0.627 0.482 The lines obtained iteratively by the linear DC method are 12
P41 0.690 0.880 0.895 1.130 0.877 whereas those obtained by the mixed integer linear DC method are 7
P42 0.639 0.641 0.650 0.746 0.650 which is almost half the number of lines and which means about 58.7%
P43 0.094 0.025 0.018 0.084 0.195
less cost savings. The simulation running time for the mixed integer
P44 0.635 0.770 0.770 0.729 0.772
P45 0.347 0.47 0.470 0.416 0.469 linear technique is slightly larger than the time of the linear technique.
P46 0.567 0.570 0.560 0.481 0.559 This implies that the simplicity and high speed of the linear planning
P47 0.933 0.802 0.814 0.871 0.803 DC based method encourage the utilization of this method just to get
initial estimate for the planned network. Furthermore, the mixed in-
teger DC based model is to be implemented to get a more accurate and
Table 6 more economic network plan bearing in mind that it consumes larger
Voltage magnitudes from AC load flow.
computation time.
Bus voltage Bus voltage magnitude for Bus voltage magnitude for The comparison presented in Table 2 showed that the real power
magnitude min P loss min V deviation flow obtained from the DC based model is not perfect when compared
with the power flow obtained from the AC based model even when the
V1 1 1
V2 1.001 0.997 Q of the loads are equal to zero.
V3 1.076 1.007 The proposed AC based model proved its ability to yield an op-
V4 1.087 1.002 timum network successfully compared to the DC load flow based
V5 1.064 0.998
planning model. The AC load flow based planning model can control
V6 1.053 0.997
V7 1.039 0.995
both active and reactive power flow on the lines through controlling the
V8 1.058 0.999 new added lines as well as the generation reactive power injection.
V9 1.062 1 Furthermore, it considers the power losses of the system in the for-
V10 1.059 1 mulations. For an optimum network with a certain voltage deviation;
V11 1.069 1
the higher voltage deviation permitted the lesser lines added.
V12 1 0.988
V13 0.981 0.985 Moreover, and more importantly the AC based model can plan an op-
V14 0.996 0.989 timum network more economically than the DC load flow based plan-
V15 1.02 0.992 ning model.
V16 1.083 1.009
V17 1.086 1.006
V18 1.087 1.005 6. Conclusion
V19 1.113 1.012
This paper solves the TNEP problem using three mathematical
models; a linear DC model, a mixed integer linear DC model, and a
novel proposed exact mixed integer non-linear AC model that considers
active and reactive power flow control and considers the power losses
in the system. A 19 bus system which is a part of the 220 kV Egyptian

32
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Fig. 3. The lines added for the DC and AC load flow based cases studied.

are reached in all study cases.

The mixed integer linear DC based model has considered the voltage
loop equations in a modified form such that the addition or deletion of
new lines does not disturb the loop equations, which guarantee lower
cost when compared with linear models on the expenses of larger si-
mulation time.
The novel proposed non-linear AC load flow based model presents
an exact representation of the power flow equation where the effect of
the reactive power injection from the generators, system losses, and bus
voltage limits, on the planning process are correctly considered. Also
the line power limit is correctly considered in MVA rather than the limit
in MW.
All these variables/points guarantee an accurate solution for
transmission systems planning problem. Although this paper has con-
sidered one line size, as made in Egypt, the developed model can be
easily modified to consider more than one line size.
Fig. 4. The number of lines added and their total cost for the DC and AC load The results obtained for the considered test example are promising
flow based cases studied. where further researches can be made considering the presented DC
and AC models.
transmission network is used for the verification of the proposed load
flow models. Funding
It was concluded that the linear programming DC based model,
although can handle problems with minimum running time, can only This research did not receive any specific grant from funding
add new lines iteratively and has no guarantee that optimal solutions agencies in the public, commercial, or not-for-profit sectors.

Appendix A

Line data

Table A1
The line data of the 19 bus system.
Line number From To L (km) X (p.u.) R (p.u.) Capacity per unit MVA

1 2 14 40 0.08250 0.00825 1.2

2 3 14 80 0.1650 0.01650 1.2
3 14 16 50 0.1030 0.01030 1.2
4 2 4 75 0.1550 0.01550 1.2
5 3 17 45 0.0930 0.00930 1.2
6 3 18 50 0.1030 0.01030 1.2
7 4 16 40 0.0825 0.00825 1.2
8 4 17 52 0.1075 0.01075 1.2
9 5 18 75 0.0415 0.00415 1.2
10 5 8 75 0.0790 0.00790 1.2
11 6 18 90 0.0950 0.00950 1.2
12 9 18 35 0.0720 0.00720 1.2
13 10 18 50 0.1600 0.01600 1.2
14 8 17 35 0.0720 0.00720 1.2
(continued on next page)

33
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

Table A1 (continued)

Line number From To L (km) X (p.u.) R (p.u.) Capacity per unit MVA

15 10 16 40 0.0825 0.00825 1.2

16 10 15 45 0.0930 0.00930 1.2
17 11 15 60 0.1200 0.01200 1.2
18 10 15 55 0.1075 0.01075 1.2
19 11 12 52 0.1040 0.01040 1.2
20 13 19 55 0.1075 0.01075 1.2
21 13 15 46 0.0950 0.00950 1.2
22a 2 4 0 0.0105 0.00105 1.2
23 1 5 0 0.0515 0.00515 1.2
24 2 3 0 0.0515 0.00515 1.2
25 1 2 0 0.0455 0.00455 1.2
26 11 19 0 0.1650 0.01650 1.2
27 11 10 0 0.1030 0.01030 1.2
28 10 9 0 0.0720 0.00720 1.2
29 9 8 0 0.0100 0.00100 1.2
30 8 7 0 0.0700 0.00700 1.2
31 7 6 0 0.0830 0.00830 1.2
32 6 5 0 0.0890 0.00890 1.2
33 5 4 0 0.0415 0.00415 1.2
34 4 18 0 0.0415 0.00415 1.2
35 4 3 0 0.0620 0.00620 1.2
36 18 17 0 0.0415 0.00415 1.2
37 17 16 0 0.0700 0.00700 1.2
38 16 3 0 0.0410 0.00410 1.2
39 3 2 0 0.1280 0.01280 1.2
40 2 1 0 0.0805 0.00805 1.2
41 1 14 0 0.1030 0.01030 1.2
42 14 13 0 0.0825 0.00825 1.2
43 14 15 0 0.0890 0.00890 1.2
44 16 15 0 0.1300 0.01300 1.2
45 15 12 0 0.0805 0.00805 1.2
46 12 13 0 0.0290 0.00290 1.2
47 12 19 0 0.1445 0.01445 1.2

a
L has zero value for existing lines (starting line 22) to realize cost equals to zero.

Bus data

Table A2
The bus data of the 19 bus system.
Bus Real power generated Load active power Load reactive power
(p.u.) (p.u.) (p.u.)

1 1.266 0 0
2 0 0.941 0.452
3 1.515 0 0
4 1.789 0 0
5 0 0.721 0.349
6 0 1.352 0.655
7 0 0.776 0.376
8 0 0.724 0.351
9 0 0.814 0.394
10 0 0.903 0.437
11 0 1.167 0.565
12 0 0.712 0.345
13 0 1.206 0.584
14 0 0.918 0.445
15 0 0.638 0.309
16 1.646 0 0
17 1.632 0 0
18 1.587 0 0
19 1.437 0 0
Sum 10.872 10.872 5.266

References English Version (Master Thesis), COPPE/UFRJ, Rio de Janeiro, Brazil, 2014.
[2] B.T. Tessema, Strategies for Electrical Network Expansion, (2011).
[3] J. Jia, Transmission Expansion Planning in Large Power Systems Using Power
[1] R.C. Perez, Power Flow Controllability and Flexibility in the Transmission System Equivalencing Techniques, (2014).
Expansion Planning Problem: A Mixed-integer Linear Programming Approach, [4] W. Jianxiao, H. Zhong, X. Qing, K. Chongqing, Transmission network expansion

34
M.A. Farrag, et al. Electric Power Systems Research 171 (2019) 26–35

planning with embedded constraints of short circuit currents and N-1 security, J. [16] H. Kim, S. Lee, S. Han, W. Kim, Integrated optimization of generation and trans-
Mod. Power Syst. Clean Energy 3 (2015) 312–320. mission expansion planning using decomposition method, 2015 6th International
[5] J.A. Taylor, F.S. Hover, Linear relaxations for transmission system planning, IEEE Conference on Intelligent Systems, Modelling and Simulation (ISMS), (2015), pp.
Trans. Power Syst. 26 (2011) 2533–2538. 158–163.
[6] S. Abbasi, H. Abdi, Multiobjective transmission expansion planning problem based [17] J. Mareček, M. Mevissen, J.C. Villumsen, MINLP in transmission expansion plan-
on ACOPF considering load and wind power generation uncertainties, Int. Trans. ning, 2016 Power Systems Computation Conference (PSCC), (2016), pp. 1–8.
Electr. Energy Syst. 27 (2017) e2312. [18] R. Bent, C. Coffrin, R.R. Gumucio, P. Van Hentenryck, Transmission network ex-
[7] P.V. Gomes, J.T. Saraiva, Transmission System Planning Considering Solar pansion planning: bridging the gap between AC heuristics and DC approximations,
Distributed Generation Penetration, (2017). 2014 Power Systems Computation Conference (PSCC), (2014), pp. 1–8.
[8] J. Qiu, Z.Y. Dong, K. Meng, Y. Xu, J. Zhao, Y. Zheng, Multi-objective transmission [19] S. El-Sobki, M. El-Metwally, M. Farrag, New approach for planning high-voltage
expansion planning in a smart grid using a decomposition-based evolutionary al- transmission networks, IEE Proceedings C-Generation, Transmission and
gorithm, IET Gener. Transm. Distrib. 10 (2016) 4024–4031. Distribution, (1986), pp. 256–262.
[9] S.P. Torres, C.A. Castro, Expansion planning for smart transmission grids using AC [20] M. Farrag, M. El-Metwally, New method for transmission planning using mixed-
model and shunt compensation, IET Gener. Transm. Distrib. 8 (2014) 966–975. integer programming, IEE Proceedings C-Generation, Transmission and
[10] S.P. Torres, R.A. de Araujo, C.A. Castro, J. Pissolato, Security constrained trans- Distribution, (1988), pp. 319–323.
mission expansion planning for smart transmission grids based on the AC network [21] H. Zhang, G.T. Heydt, V. Vittal, H.D. Mittelmann, Transmission expansion planning
model, Transmission & Distribution Conference and Exposition-Latin America (PES using an AC model: formulations and possible relaxations, 2012 IEEE Power and
T&D-LA), 2014 IEEE PES, (2014), pp. 1–6. Energy Society General Meeting, (2012), pp. 1–8.
[11] A.S. Zakeri, H. Askarian Abyaneh, Transmission expansion planning using TLBO [22] H. Zhang, G.T. Heydt, V. Vittal, J. Quintero, An improved network model for
algorithm in the presence of demand response resources, Energies 10 (2017) 1376. transmission expansion planning considering reactive power and network losses,
[12] C. Zhang, Q. Wang, Y. Ding, J. Østergaard, A multi-objective model for transmission IEEE Trans. Power Syst. 28 (2013) 3471–3479.
planning under uncertainties, 2014 IEEE, Electrical Power and Energy Conference [23] S. Hong, H. Cheng, P. Zeng, J. Zhang, J. Lu, Composite generation and transmission
(EPEC), (2014), pp. 42–47. expansion planning with second order conic relaxation of AC power flow, 2016
[13] M. Elmetwally, F.A. Aal, M.L. Awad, S. Omran, A hopfield neural network approach IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC), (2016),
for integrated transmission network expansion planning, MEPCON 2006 Power pp. 1688–1693.
Systems Conference, 2006 (2006) 371–376. [24] T. Gönen, Electric Power Distribution System Engineering, McGraw-Hill, New York,
[14] I. Alhamrouni, A. Khairuddin, A.K. Ferdavani, M. Salem, Transmission expansion 1986.
planning using AC-based differential evolution algorithm, IET Gener. Transm. [25] K.M.I. Ali, A New Transmission Expansion Planning Model Based on AC Load Flow
Distrib. 8 (2014) 1637–1644. (MSc), Electrical Power & Machines Dept., Cairo University, 2018.
[15] D. Van Hertem, J. Verboomen, K. Purchala, R. Belmans, W. Kling, Usefulness of DC [26] LINGO 17.0 — Optimization Modeling Software for Linear, Nonlinear, and Integer
Power Flow for Active Power Flow Analysis With Flow Controlling Devices, (2006). Programming, in LINDO SYSTEMS INC., 2017.

35