Applied Energy: Sciencedirect
Applied Energy: Sciencedirect
Applied Energy: Sciencedirect
Applied Energy
journal homepage: www.elsevier.com/locate/apenergy
a
Zanjan Electric Distribution Company, Zanjan, Iran
b
Department of Electrical Engineering, Faculty of Engineering, University of Zanjan, P.O. Box: 45371-38791, Zanjan, Iran
HIGHLIGHTS
• AThenovel heuristic approach is proposed for optimal active and reactive power allocation in distribution systems.
• The proposed method is easy to implement, fast, accurate, and applicable to large distribution systems.
• The presented approach is applied to 33, 69, and 119-bus distribution networks.
• results are compared with Global Search, PSO, IA, hybrid analytical-PSO, TLBO, QOTLBO, and CTLBO algorithms.
Keywords: A wide range of solution methodologies has been presented in the literature for the siting and sizing of active and
Distribution network reactive power sources in radial distribution systems. The solution techniques are mainly classified into math-
Optimal siting and sizing ematical programming algorithms, heuristics, meta-heuristic methods, and the analytical approaches. This paper
Distributed generation proposes a novel heuristic method for optimal allocation of distributed generation (DG) and capacitor banks. The
Capacitor banks
conducted approach is easy for implementation and does not suffer from the complicacy of the other methods.
New heuristic method
Based on some mathematical calculations, two main formulations are derived. Then, the resulting formulations
are employed to site and size capacitor bank and DG unit in the 33-bus distribution network to verify its per-
formance by comparing the obtained results with those of global search method. Afterward, the proposed
method is used to allocate a specified number of DGs and capacitors in the 33 and 69-bus systems. Finally, the
conducted approach is employed to allocate multiple DGs and capacitors, and also multiple DGs with optimal
power factor in the 119-bus distribution network. The recently presented methodologies including particle
swarm optimization (PSO), improved analytical (IA), hybrid analytical-PSO, Teaching Learning Based
Optimization (TLBO), Quasi-Oppositional Teaching Learning based optimization (QOTLBO), Comprehensive
Teaching Learning-Based Optimization (CTLBO), and LSF-BFOA algorithms are used for investigating the effi-
ciency of the developed method. The simulation results demonstrate that the presented heuristic approach is
robust in finding the optimal results, very fast and easy to implement, applicable to large distribution systems
and it give better results than the algorithms presented so far.
1. Introduction techniques [6,7]. In the reconfiguration, the tie switches are closed or
opened to change the topology of the distribution network, and con-
Traditionally, distribution systems are operated in the centralized sequently, the direction of power flow, in order to reduce the power loss
and vertical manner due to the simplicity of control and protection and improve the voltage profile. In this process, the radial structure of
operations. These networks mainly include radial topologies or weakly- the network should be maintained. With the aim of power loss reduc-
meshed structures having high value of R/X ratio which results in im- tion and voltage profile improvement by the injection of reactive
proper voltage profile and high power losses [1–3]. The prevailing power, the shunt capacitor banks having proper sizes can be connected
actions taken by the network operators to settle this problem include to the appropriate nodes of the distribution system.
the “feeder reconfiguration” [4,5] and “capacitor banks allocation” By development of distributed generation (DG) in the power system,
⁎
Corresponding author.
E-mail address: amir_bagheri@znu.ac.ir (A. Bagheri).
https://doi.org/10.1016/j.apenergy.2018.10.030
Received 14 April 2018; Received in revised form 25 August 2018; Accepted 9 October 2018
0306-2619/ © 2018 Published by Elsevier Ltd.
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
there has been another alternative to tackle with the poor voltage the proposed method are justified by comparing them with those of
regulation and high power losses in distribution systems [8]. DG units, conventional grid search method based on successive load flows. This
with several technological types, have potential advantages such as research finds the location and size of only one DG unit. In addition, the
power loss reduction, voltage profile improvement, reliability en- considered DG has the unity power factor, i.e., it injects only the active
hancement, and investment deferral of the system equipment. If the DG power. Analytical expressions have been proposed in [43] to find the
units’ location and size are appropriately determined, they can play a optimal size and power factor of four types of DG units for the sake of
key role in minimizing the power loss and improving voltage profile of highest power loss reduction in distribution networks. The presented
distribution networks [9,10]. expressions are based on the modifications made on the method of [44]
The problem of DG placement and sizing is, in principle, a complex that was limited to the DG delivering only the active power. The four
non-linear optimization problem. The implemented researches in this types of considered DGs include those capable of injecting active power
field fall into several categories regarding the considered objectives, (P) only, injecting reactive power (Q) only, injecting both P & Q, or
constraints, and solution algorithms [11]. The objective functions in- injecting P and consuming Q. The obtained results confirm that the
tended by the researchers include power loss minimization [12–14], proposed method can find the optimal solution as verified by the
voltage profile improvement [15–17], network reinforcement cost comprehensive load flow method. This research has concentrated on
minimization [18,19], reliability enhancement [20,21], and reduction the allocation of just a single DG, not multiple DG units. A heuristic
of environmental emissions [22]. method based on UVDA (Uniform Voltage Distribution Algorithm) is
The owner of DG may be the private investors or the distribution developed in [45] which is used for the optimal DG siting and sizing
company [23,24]. In the case of utility-owned DGs, the utility has to simultaneous with the feeders’ reconfiguration for the aim of maximum
optimally plan the location and size of the DG units in order to improve loss reduction. The algorithm is very fast and robust in finding the
the network technical and economic benefits [25]. optimal solutions, but, it is not appropriate for the optimal allocation of
A wide range of solution methodologies has been proposed in the reactive power sources, i.e., the capacitor banks. In [46], a hybrid
literature for the utility-owned DG siting and sizing problem. The so- analytical and meta-heuristic methodology is developed for the optimal
lution techniques are mainly classified as mathematical programming placement of multiple DGs with different types for the sake of power
algorithms, heuristics, meta-heuristic methods, and the analytical ones. loss reduction in distribution network. The locations of DGs are de-
The methods of the first category use mathematical formulations like termined by the PSO algorithm, while the sizes are adjusted using an
mixed-integer linear and non-linear programming [26,27]. These analytical method. The optimal power factor for the DG units and the
methods have some limitations such as unsuitability in application to voltage profile are also addressed. The results indicate the superiority of
large-scale systems and introducing errors when linearization of the the hybrid approach compared to PSO and other analytical methods.
non-linear characteristics of the problems. The heuristic methods are Due to using PSO which is an iteration-based optimization method, the
based on engineering experience such as researches done in [28,29]. convergence speed and accuracy of the algorithm can be affected by
These methods always find feasible solutions, but the obtained solu- increasing the number of variables and extent of the network. A Bac-
tions may not be optimal. The meta-heuristic techniques, also known as terial Foraging Optimization Algorithm (BFOA) is presented in [47] to
population-based optimization methods, include the algorithms like simultaneously allocate DG and distribution static compensator (DST-
genetic algorithm (GA) [30], particle swarm optimization (PSO) [31], ATCOM) in order to minimize power loss and operational costs, and
improved PSO (IPSO) [32], harmony search (HS) [33], artificial bee also, enhance the voltage profile of the distribution system. To pre-
colony (ABC) [12], Big Bang-Big crunch (BBBC) algorithm [34], grey determine the location of DG and DSTATCOM, the loss sensitivity factor
wolf optimizer (GWO) [35], and hybrid algorithms such as GA-PSO (LSF) is used, and then, the BFOA is employed to determine the optimal
[36], genetic algorithm-tabu search (GA-TS) [37], PSO and shuffled size of DG and DSTATCOM for different load models. The obtained si-
frog leaping algorithms (PSO-SFL) [38] etc. The above-mentioned po- mulation results are better than the existing methods; however, the
pulation-based techniques have been widely applied to optimize the DG execution time of the algorithm is relatively high. A new quasi-oppo-
planning studies, and they have obtained satisfactory results. However, sitional teaching learning-based optimization (QOTLBO) has been
these methods require long running time due to their iterative solving presented in [14] for optimal placement of DG with three objective
process, and they need tuning a great number of parameters. On the functions: power loss reduction, voltage deviation minimization, and
other hand, the convergence trend of these algorithms worsens when voltage stability improvement. The presented method is superior to
applied to the large-scale networks. original teaching learning-based optimization (TLBO) algorithm in
The analytical approaches are the fourth type of the solution terms of convergence speed and optimality of the results. This method
methodologies used for the DG siting and sizing problem. Due to their is from the family of meta-heuristic approaches and retains their ad-
simplicity, easy implementation, and low computational time, the vantage and disadvantages. A more advanced algorithm named com-
analytical methods have attracted great attention in DG planning pro- prehensive teaching learning-based optimization (CTLBO) has been
blems [39–43]. presented in [17] to settle the deficiencies of QOTLBO method such as
The power loss minimization and voltage profile improvement of parameter dependency and trapping in the local optima. This algorithm
distribution networks are the most common objectives of the researches finds better results than the TLBO and QOTLBO.
done in the area of DG allocation. A genetic algorithm in [30] seeks for The review of the literature reveals that several methodologies of
the optimal locations of DGs for minimizing the total active power loss different types have been proposed for the optimal DG siting and sizing
of the radial distribution systems based on loss sensitivity to the change problem. Each methodology has its own advantages and disadvantages.
of the active power. The authors in [39] employed phasor current in- In total, the following drawbacks can be found in these methods:
jection method to derive an analytical approach for optimal DG pla-
cement in radial distribution networks in order to minimize the power • Some approaches consider only the active power injection, i.e., DG
loss. For simplification, it is assumed that the load distribution in the with unity power factor;
network is in central, uniform, and increasing manner. In this study, • Some methodologies are appropriate for allocation of single DG
only, the location of DG is optimized, and its size optimization is not unit;
taken into consideration. Gozel and Hocaoglu [42] proposed an ana- • The computational time of some methods is so high;
lytical approach for determining the location and size of DGs aiming at • The results found by some presented methods are not optimal.
the minimization of the total power loss. Based on equivalent current
injection, a loss sensitivity factor is formulated which does not require To settle the above-mentioned shortcomings, the current paper
the admittance matrix, its inverse, or the Jacobian matrix. The results of proposes a novel heuristic method for the optimal allocation of active
72
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
and reactive power sources, i.e., distributed generation (DG) and ca-
pacitor banks. As the optimal use of energy resources, this problem is an
important subject in “Applied Energy”. In this regard, the conducted
approach in this paper is easy to implement and program, and does not
require complicated mathematical formulations which are seen in the
other methods. Based on some mathematical calculation, two main
formulations are derived indicating the best location and capacity of
active & reactive power injection. Then, the obtained formulations are
employed to site and size capacitor bank and DG unit in the 33-bus
distribution network. For verifying the performance of the presented
algorithm, the obtained results are compared with those of the global
search method. Then, the proposed method is used to allocate a spe-
cified number of DGs and capacitors in the 33 and 69-bus systems.
Finally, the conducted approach is employed to allocate multiple DGs
and capacitors, and also multiple DGs with optimal power factor in the
119-bus distribution network. Comparing the simulation results of the
proposed method with the recent methodologies of PSO, IA, hybrid
analytical-PSO, TLBO, QOTLBO, CTLBO, and LSF-BFOA algorithms
demonstrate that the presented heuristic approach is very powerful in
finding the optimal results, very fast and simple, and applicable to large
distribution systems. The man contributions of this paper can be out-
lined as the following:
73
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 2. (a) Power flow results for 33 bus network with the duplicated active power of loads (2Pi) when the reactive parts are removed, (b) optimal location for the
injection of active power for the given range (from 0 up to 7828.7 kW).
Fig. 3. Power loss resulted from the proposed and global search methods for the active power injection from 100 kW to 8000 kW (for each 100 kW increment in
active power, the optimal locations obtained by the proposed and global search methods are shown, proposed: up, global search: down).
74
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 4. Power loss resulted from the proposed and global search methods for the reactive power injection from 100 kVAr to 4600 kVAr (for each 100 kVAr increment
in reactive power, the optimal locations obtained by the proposed and global search methods are shown, proposed: up, global search: down).
requires
j j
branches brings about some active power loss, and this may slightly
max ( Ploss ) max ri IP = max VP increase 2Pi of the nodes. These negligible differences are near the zero
i=1 i=1
in the ending nodes, and they increase gradually as we close to the
requires
= max (1 Vj ) min (Vj ) (8) source node.
From (4) and (8), it can be concluded that for the sake of power loss 2.2. Description of the proposed approach for the reactive power allocation
minimization, the DG should be located at the node having the highest
voltage drop resulted from the real part of Ii (the bus having minimum A procedure similar to the one described in Section 2.1 can be
voltage as (8)) provided that the DG’s size (P0) is smaller than the du- employed to extract two necessary conditions for the maximum power
plicated value of the active power (2Pi) flowing at its upstream branch loss reduction using the injection of the reactive power source (e.g.
(relation (4)). To calculate the voltage drop resulted from the real part using the installation of capacitor bank). The first condition is same as
of Ii , all the reactive parts of the loads are removed during the DG siting (8), and the second one is considered according to (9):
and sizing process.
0 < Q0 < 2Qi i {1, 2, , j} (9)
Based on the above explanations, the following algorithm can be
employed to determine the location of DG unit in radial distribution Consequently, the following steps can be implemented for optimal
networks: reactive power sizing and placement:
(1) Remove all the reactive parts of the loads from the network data (1) Remove all the active parts of the loads from the network data and
and perform load flow; perform load flow;
(2) Sort the nodes of the network according to their voltage magnitude; (2) Sort the nodes of the network according to their voltage magnitude;
(3) calculate the active power flowing at the upstream branch (Pi); (3) Calculate the reactive power flowing at the upstream branch (Qi);
(4) The node with the minimum voltage which satisfies P0 < 2Pi is the (4) The node with the minimum voltage which satisfies Q0 < 2Qi is the
best node for the DG installation. best node for the capacitor installation.
To verify the performance of the proposed approach in the optimal For different values of reactive power injection (from 100 kVAr to
placement of active power injection, it is applied to the 33-bus dis- 4600 kVAr), the optimal location of capacitor and the corresponding
tribution network to find the best location of DG units. This network power loss resulted from the proposed method is depicted in Fig. 4 and
whose data has been given in [48] has a total load of 3.72 MW and compared with the global search algorithm. As it can be seen, the
2.3MVAr. The four steps mentioned above are implemented on the 33- proposed method, except in few cases, has obtained the same solution
bus network in Fig. 2. of the global search method.
For different values of active power injection (from 100 kW to
8000 kW), the optimal location and its corresponding power loss re- 2.3. Optimal placement of multiple active power sources with the given sizes
sulted from the proposed method and also by using the global search
algorithm are depicted in Fig. 3. As shown, of the 80 cases for the DG For multiple DGs (active power sources) allocation, the DGs are
size, in 76 cases, the developed approach has obtained the same solu- placed one after another so that the steps 1 to 4, described in Section
tion of the global search method. Only, in four cases, there are negli- 2.1, are performed at each DG placement stage. To verify the effec-
gible differences between the two methods. For example, in 4500 kW tiveness of the proposed method, it is applied to the 33-bus network to
case, the difference is 2 kW, or in the other words, 1.33%. This minor site three DG units with the sizes of 900 kW. The three DGs’ allocation
difference may arise from the fact that the reactive parts of load cur- stages using the proposed method are illustrated in Fig. 5. Based on the
rents have been disregarded in the calculations while their flow at the described methodology, the nodes are sorted according to their
75
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 5. The procedure of siting three DG units on the 33-bus network, (a), (b), and (c) are respectively for the first, second, and third DG allocation.
voltages, and by considering the active power range of all buses, the third DG is located at bus 24 (Fig. 5(c)). The results for the three DGs
first DG is located at bus 13 of the network (Fig. 5(a)). Then, by con- allocation are the same as the results of Ref. [41] which have been
sidering the first located DG, the procedure is re-performed so that the determined by using the exhaustive load flow (ELF) method, i.e., the
second DG is placed at bus 30 of the network (Fig. 5(b)). In the end, the nodes 13, 30 and 24. This arrangement of DGs yields 74.27 kW active
76
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 6. (a) Single-line diagram for part of distribution network; (b) phasor diagram of voltages and currents.
2.4. Optimal placement and sizing of multiple active and reactive power
sources
(14)
Vj 2cos 2 = (1 x )2
new
Vj 2 k 2x 2 = x 2 2x + 1 (16)
yields
(1 + k 2)x 2 2x + (1 Vj 2 ) = 0 Ax 2 + Bx + C = 0
A B C (17)
By solving the quadratic equation of (17), the proper value of active
current (IA ) and the corresponding active power (P0 ) is determined as
(18):
P0 = Vjnew IA (18)
Eq. (18) should be used in the iterative process of AC power flow for
distribution network until the convergence condition is satisfied. In this
paper, the backward-forward method has been employed for the power
flow calculations [49]. Fig. 7 shows the power flow procedure con- Fig. 7. Power flow process of radial distribution network considering active
sidering the active power injection model for improving the voltage of power injection model.
bus j to the desired value (1 p.u).
77
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 8. Flowchart of the proposed method for siting and sizing of DG with optimal power factor.
2.4.2. Reactive power injection model for bus voltage compensation Step 3) Determine the ending node with the lowest voltage mag-
The procedure described above for the active power compensation nitude;
can be extended for compensating the bus voltage using the reactive Step 4) Determine the upstream nodes (the nodes located between
power injection. For this aim, the active parts of loads are removed the ending node and the source node);
from the network data, and the load flow calculation is performed just Step 5) Allocate DGs having unity power factor on the acceptable
for the reactive powers. nodes (Stage A) in order to enhance the voltage of acceptable nodes up
to 1pu.
2.4.3. Description of the proposed method for optimal active and reactive After performing the steps 1 to 5, the location and size of active
power allocation power injection (DGs with unity power factor), are determined, and the
Based on the explanations given previously, the following stages are active power loss of the network (Ploss
A
) is calculated.
performed to site and size a desired number of DG units with the op- A procedure similar to steps 1 to 5 is implemented to determine the
timal power factor: location and size of reactive power injection (capacitor banks). In this
Step 1) Remove the reactive part of loads; condition, active power loss of the network (PlossB
) is again calculated.
Step 2) Calculate 2Pk of nodes by performing the power flow; Afterward, the power losses corresponding to active and reactive
78
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Table 1
Comparison of the obtained results of the proposed approach with the existing methods on the 33-bus network.
Number of DG/capacitor Techniques Details Total capacity (kW/kVAr) Power loss (kW)
79
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 9. Convergence curve of bus voltages for 1-DG case in the 33-bus network.
The 69-bus system has a total load of 3.80 MW and 2.69 MVAr. The found by the PSO algorithm, i.e., 83.2 kW [31]. In 2DG case, the result
complete data for this system has been given in [49]. The effectiveness is the same for the hybrid method, but better than PSO. The similar
of the proposed approach is also verified in this study. In 1DG case, it is result of case 1 is also seen in case 3, i.e., 3DG in which TLBO, QOTLBO,
seen that by the lower size of installed capacity (1823 kW compared to and CTLBO methods are also compared. In 2C scenario, the proposed
1870 kW), the presented approach yields nearly the same power loss approach finds the minimum power loss by the lower size of the
Table 2
Comparison of the obtained results of the proposed approach with the existing methods on the 69-bus network.
Number of DG/capacitor Techniques Details Total capacity (kW/kVAr) Power loss (kW)
Original network 0 225
80
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 10. Description of the proposed method for allocation of one DG in 69-bus network.
Table 3
Detailed steps of the proposed method for allocation of one DG in 69-bus network.
Steps of the proposed method (according to Section 2.4.3) Implemented operations Consumed time from the
beginning (seconds)
Step 1) Remove the reactive part of loads. Performing the required commands for eliminating the reactive power from the —
network data
Step 2) Calculate 2Pk of nodes by performing the power flow; Performing some MATLAB commands and execution of a conventional load flow t = 0.0345
Step 3) Determine the ending node having the lowest voltage
magnitude;
Step 4) Determine the upstream nodes (the nodes located between the Performing some MATLAB commands (the results have been shown in Table 3. As
ending node and the source node); seen, the node no. 65 has the lowest voltage and it is selected as the first node to
calculate the required capacity of DG for adjusting its voltage to 1pu in load flow
program using the proposed method)
Step 5) Allocate DGs having unity power factor on the acceptable
nodes (Stage A) in order to enhance the voltage of acceptable
nodes up to 1pu;
5–1) Perform the load flow by calculating the required The required capacity is 1421 kW t = 0.074
power injected to node 65 in order to adjust its
voltage to 1pu.
5–2) If 2Pk of node 65 is higher than the obtained No (2Pk = 117.94 < 1421) —
capacity?
5–3) Select the first node among the upstream nodes Node no. 61 is selected t = 0.0811
whose 2Pk is higher than 1421 kW
5–4) Perform the load flow by calculating the required The required capacity is 1823 kW t = 0.176
power injected to node 61 in order to adjust its
voltage to 1pu
5–5) If 2Pk of node 61 is higher than the obtained Yes (2Pk = 3123.92 > 1823) t = 0.21
capacity?
5–6) Consider the reactive power of the loads and Performing the required commands for addition of the reactive power of loads to
execute a conventional load flow the network data and executing the conventional load flow
5–7) Output the results Print the results and the required figures
81
A. Bayat, A. Bagheri
Table 4
Comparing the results of the proposed approach with the existing methods in 119-bus network.
Method Base case Proposed method LSF-BFOA [47] Proposed method LSF-BFOA [47] Proposed method LSF-BFOA [47]
Value of active and reactive power to be injected to the given locations at each scenario — Q Bus Q Bus P Bus P Bus P Q Bus P Q Bus
(kW/kVAr) 2665 52 2514 38 1295 44 2750 38 1295 833 44 2696 2480 38
1521 77 1425 47 2769 52 2000 47 2769 2621 52 1945 1300 48
1658 83 1521 76 2422 76 2800 73 2422 1718 76 2412 1580 76
1098 100 1715 94 1812 100 2160 94 1812 1446 100 2186 1700 94
2560 115 2021 122 2780 115 2950 113 2780 2560 115 2984 2400 113
Total active & reactive power — 9502 9196 11,078 12,660 11,078 9178 – 12,223 9460 –
Power loss (kW) 1298 861.6 871.4 580.74 578.97 227.5 227.9
Reduction in power loss (kW) 0.00% 33.5% 32.9% 55.2% 55.4% 82.5% 82.4%
Minimum voltage (p.u.) 0.869 0.9094 0.9062 0.9558 0.9536 0.9662 0.964
Computational time (seconds) 0 2 23.21 2.5 23.24 5.4 24.65
82
Method Base case Proposed method CTBLO [17] QOTLBO [14] TLBO [14] LSF-BFOA [47] Proposed method LSF-BFOA [47]
Value of active and reactive power to be injected to the given locations at each — P Bus P Bus P Bus P Bus P Bus P Q Bus P Q Bus
scenario (kW/kVAr) 1802 20 1817 20 1246 24 1755 8 1545 32 1802 1120 20 1756 1300 19
1267 44 1276 44 732 42 591 10 1968 39 1267 822 44 2645 2510 39
2731 52 2767 52 3539 47 1536 36 2073 48 2731 2613 52 2042 1165 48
2287 77 2533 75 2679 74 2686 49 2670 74 2287 1521 77 2391 1520 72
2080 83 2095 83 1248 78 2501 71 1534 87 2080 1658 83 1536 1000 87
1667 100 1663 100 1086 94 2494 79 2106 94 1667 1098 100 2012 1640 94
2780 115 3120 114 3243 108 2663 110 3118 113 2780 2562 115 2815 2400 113
Total active & reactive power — 14,614 15,271 13,773 14,226 15,014 14,614 11,394 – 15,197 11,535 –
Power loss (kW) 1298 515.7 516.25 576 590.697 526.34 128.8 132.1
Reduction in power loss (kW) 0.00% 60.2% 60.1% 55.6% 54% 59.45% 90.1% 89.9%
Minimum voltage (p.u.) 0.869 0.9558 — — — 0.9497 0.9753 0.9761
Computational time (seconds) 0 4.1 — 20.83 25.292 24.96 6.2 25.96
Applied Energy 233–234 (2019) 71–85
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 11. Voltage profile of the 119-bus network in the case of five DGs installation.
capacitor (1580 kVAr). It is noteworthy to mention that the results of all results of the proposed and LSF-BFOA methods in the second and third
the scenarios of 33 and the 69-bus systems have been obtained in less scenarios are almost the same. For example, in 5-DG with optimal
than 1 s which is a short execution time. The simulations have been power factor scenario, the power loss and minimum voltage are
programmed and executed in MATLAB 2014a, on a PC Intel Core 2 Duo 227.9 kW and 0.964p.u. for the LSF-BFOA, and 227.5 kW and
3 GHz, 4 GB RAM. 0.9662p.u. for the proposed method. These results certificate the va-
In order to elaborate the operation of the proposed heuristic ap- lidity of the obtained results and the effectiveness of the developed
proach, the implementation steps required for the allocation of one DG approach. The robustness of the proposed method can be observed in
unit in 69-bus network (1-DG scenario in Table 2) have been detailed in the last two scenarios where the number of variables is high (there are 7
Fig. 10 and Table 3. The steps given in Table 3 are according to the DGs, 7 capacitors, and 7 locations, totally 21 variables). In 7-DG with
flowchart of Fig. 8. The steps have been executed one by one, and for unity power factor case, the power loss obtained by the proposed
each step, the consumed time from the beginning is calculated and method is 515.7 kW which is 10.64 kW less than the power loss of LSF-
given in the Table. For the allocation of one DG with the aim of max- BFOA algorithm. This is while the total size of DGs is 14614 kW which
imum power loss reduction, as it is described in this table, two con- is 400 kW less than LSF-BFOA. Also, the power loss reduction by the
ventional load flows and two load flows using the proposed approach proposed approach is more than TLBO, QOTLBO, and CTLBO methods.
have been executed to determine the required capacity of DG in nodes A similar behavior can be observed in the last scenario, i.e., 7-DG with
61 and 65 so that their voltage is adjusted to 1pu. As seen, the required optimal PF case.
time from the beginning to the end of the program is about 0.2 s. In the The voltage profile of the 119-bus network in 5-DG and 7-DG cases
other words, as the proposed heuristic method is not based on the for the unity and optimal power factors have been depicted in Figs. 11
creation of population and updating it in an iterative process, it quickly and 12. As seen, in both cases, the voltage profile has been appro-
finds the optimal solution in a very short period of time. priately improved compared to the case in which there is no DG in-
stalled in the network. In the optimal power factor case, the voltage
profile is more flat, and the voltages of DG buses have been enhanced to
2.4.6. Applying the proposed method to the 119-bus network and 1p.u.
determining the optimal power factor Fig. 13 compares the voltage profile of the proposed method with
In the previous case studies, only active (DG with unity power other algorithms in 7-DG with unity power factor case. As it can be
factor) or reactive power (capacitor bank) allocation was considered. seen, the proposed method (blue curve with circle markers) and CTLBO
By using the proposed approach demonstrated in the flowchart of algorithm (yellow curve with square markers), which is the most recent
Fig. 8, in this part, it is aimed to optimally allocate DGs injecting both algorithm, are like each other in most of the buses; this is while the total
active and reactive powers. Also, the optimal power factor of the allo- size of the installed DGs by the CTLBO algorithm is 15271 kW and it is
cated DG units is determined. This optimal power factor determines the 14614 kW for the proposed method. This means 657 kW lower installed
best value of the injected active (P) and reactive power (Q) for the sake capacity compared to CTLBO. Also, it is observed that the minimum
of acquiring the minimum power loss. For this purpose, the proposed voltage of the proposed method is more than LSF-BFOA, TLBO, and
approach is tested on a 119-bus radial distribution network as a large- QOTLBO algorithms.
scale system [14,17]. Five scenarios have been considered including As noted before, the execution time of the proposed method is sig-
sizing and siting of: (1) five capacitors, (2) five DGs with the unity nificantly lower than the other methods. As an example, in 7-DG case
power factor, (3) five DGs with optimal power factor (PF), (4) seven with the unity power factor, the simulation time is 4.1 s while it is
DGs with the unity power factor, and 5) seven DGs with optimal PF. The 24.96 s for the LSF-BFOA algorithm.
results of these scenarios are given in Table 4 to be compared with those
of TLBO and QOTLBO [14], CTLBO [17], and LSF-BFOA [47].
In the first scenario, siting and sizing of five capacitors by the 3. Conclusion
proposed method have reduced the power loss from 1298 kW to
861.6 kW and enhanced the minimum voltage of the network up to It has been presented a wide range of solution methodologies such
0.9094p.u. This is while the power loss obtained by the LSF-BFOA al- as mathematical, heuristic, meta-heuristic, and analytical algorithms
gorithm is 871.4 kW and the minimum voltage is 0.9062p.u. [47]. The for optimal siting and sizing of active and reactive power sources in
83
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
Fig. 12. Voltage profile of the 119-bus network in the case of seven DGs installation.
Fig. 13. Voltage profile of the 119-bus network in the case of 7-DG with unity power factor for different methods.
radial distribution systems. In this paper, by deriving two main for- presented approach is very accurate, fast and easy to implement, and
mulations, a novel heuristic approach was proposed for the sake of applicable to large distribution systems.
active power loss minimization using the optimal allocation of DG units
and capacitor banks. The significant features of the conducted approach References
were demonstrated by establishing several experiments on different test
systems, i.e., 33, 69, and 119-bus distribution networks. To verify the [1] Willis HL. Power distribution planning reference book. 2nd ed. New York:
efficiency of the developed algorithm, it was first tested on the 33-bus MarcelDekker; 2004. p. 1217.
[2] Kersting WH. Distribution system modeling and analysis. CRC Press; 2000.
distribution network to allocate capacitor bank and DG unit with the [3] Srinivas MS. Distribution load flows: a brief review. IEEE-PES Winter Meeting
unity power factor, and the obtained results were compared with those 2000;2(11):942–94.
of global search method. It was shown that the presented approach [4] Sultana B, Mustafa MW, Sultana U, Bhatti AR. Review on reliability improvement
and power loss reduction in distribution system via network reconfiguration. Renew
obtains the same results of the global search method. Then, the method Sust Energy Rev 2016;66:296–310.
employed to allocate multiple DGs and capacitors in the 33 and 69-bus [5] Rao RS, Narasimham SVL, Raju MR, Rao AS. Optimal network reconfiguration of
distribution systems, and the obtained results were compared with the large-scale distribution system using harmony search algorithm. IEEE Trans Power
Syst 2011;26(3):1080–8.
PSO, IA, hybrid analytical-PSO, TLBO, QOTLBO, and CTLBO methods
[6] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H, Karimi M. Optimum shunt capacitor
reported in the recent literature. The comparison of results indicated placement in distribution system- A review and comparative study. Renew Sust
the excellence of the presented approach over these algorithms. After Energy Rev 2014;30:429–39.
[7] Nojavan S, Jalali M, Zare K. Optimal allocation of capacitors in radial/mesh dis-
confirming the capability of the conducted method, in the final stage, it
tribution systems using mixed integer nonlinear programming approach. Electric
was utilized to allocate multiple DGs with optimal power factor on the Power Syst Res 2014;107:119–24.
119-bus distribution network. It was verified that the proposed heur- [8] Ackermann T, Andersson G, Soder L. Distributed generation: a definition. Electr
istic approach is superior to the previous methods in terms of total al- Power Syst Res 2001;57(3):195–204.
[9] Pecas Lopes JA, Hatziargyrou N, Mutale J, Djapic P, Jenkins N. Integrating
located size, power loss reduction, and computational time. In total, the
84
A. Bayat, A. Bagheri Applied Energy 233–234 (2019) 71–85
distributed generation into electric power systems: a review of drivers, challenges sources using heuristic optimization. IEEE Trans Power Syst 2015;30(5):2538–46.
and opportunities. Electr Power Syst Res 2007;77(9):1189–203. [30] Shukla TN, Singh SP, Srinivas V, Naik KB. Optimal sizing of distributed generation
[10] Willis HL, Walter GS. Distributed power generation: planning and evaluation. CRC placed on radial distribution systems. Electr Power Compon Syst
Press; 2000. 2010;38(3):260–74.
[11] Pesaran HAM, Huy PD, Ramachandaramurthy VK. A review of the optimal allo- [31] Ameli A, Bahrami S, Khazaeli F, Haghifam MR. Multiobjective particle swarm op-
cation of distributed generation: Objectives, constraints, methods, and algorithms. timization for sizing and placement of DGs from DG owner's and distribution
Renew Sustain Energy Rev 2017;75:293–311. company’s viewpoints. IEEE Trans Power Deliv 2014;29(4):1831–40.
[12] Abu-Mouti FS, El-Hawary ME. Optimal distributed generation allocation and sizing [32] Kanwar N, Gupta N, Niazi KR, Swarnkar A, Bansal RC. Simultaneous allocation of
in distribution systems via artificial bee colony algorithm. IEEE Trans Power Deliv distributed energy resource using improved particle swarm optimization. Appl
2011;26(4):2090–101. Energy 2017;185–2:1684–93.
[13] Li Y, Feng B, Li G, Qi J, Zhao D, Mu Y. Optimal distributed generation planning in [33] Nekooei K, Farsangi MM, Nezamabadi-Pour H, Lee KY. An improved multi-objective
active distribution networks considering integration of energy storage. Appl Energy harmony search for optimal placement of DGs in distribution systems. IEEE Trans
2018;210:1073–81. Smart Grid 2013;4(1):557–67.
[14] Sultana S, Roy PK. Multi-objective quasi-oppositional teaching learning based op- [34] Esmaeili M, Sedighizadeh M, Esmaili M. Multi-objective optimal reconfiguration
timization for optimal location of distributed generator in radial distribution sys- and DG (Distributed Generation) power allocation in distribution networks using
tems. Electr Power Energy Syst 2014;63:534–45. Big Bang-Big Crunch algorithm considering load uncertainty 2016;103:86–99.
[15] Muttaqi KM, Le ADT, Negnevitsky M, Ledwich G. An algebraic approach for de- [35] Sultana U, Khairuddin AB, Mokhtar AS, Zareen N, Sultana B. Grey wolf optimizer
termination of DG parameters to support voltage profiles in radial distribution based placement and sizing of multiple distributed generation in the distribution
networks. IEEE Trans Smart Grid 2014;5(3):1351–60. system. Energy 2016;11:525–36.
[16] Al-Abri RS, El-Saadany EF, Atwa YM. Optimal placement and sizing method to [36] Moradi MH, Abedini M. A combination of genetic algorithm and particle swarm
improve the voltage stability margin in a distribution system using distributed optimization for optimal DG location and sizing in distribution systems. Electr
generation. IEEE Trans Power Syst 2014;28(1):326–34. Power Energy Syst 2012;34(1):66–74.
[17] Quadri I, Bhowmick S, Joshi D. A comprehensive technique for optimal allocation [37] Gandomkar M, Vakilian M, Ehsan M. A genetic based tabu search algorithm for
of distributed energy resources in radial distribution systems. Appl Energy optimal DG allocation in distribution networks. Electr Power Compon Syst
2018;211:1245–60. 2007;33(12):1351–62.
[18] Mendez VH, Rivier J, De la Fuente T, Gomez,, Arceluz J, Marn J, Madurga M. [38] Gitizadeh M, Azizi Vahed A, Aghaei J. Multistage distribution system expansion
Impact of distributed generation on distribution investment deferral. Electr Power planning considering distributed generation using hybrid evolutionary algorithms.
Energy Syst 2006;28(4):244–325. Appl Energy 2013;101:655–66.
[19] EI-Khattam W, Bhattacharya K, Hagazy YG, Salama MMA. Optimal investment [39] Wang C, Nehrir MH. Analytical approaches for optimal placement of distributed
planning for distributed generation in a competitive electricity market. IEEE Trans generation sources in power systems. IEEE Trans Power Syst 2004;19(4):2068–76.
Power Syst 2004;19(3):1674–84. [40] Naik SNG, Khatod DK, Sharma MP. Analytical approach for optimal siting and
[20] Chowdhury AA, Agarwal SK, Koval DO. Reliability modeling of distributed gen- sizing of distributed generation in radial distribution networks. IET Gener Transm
eration in conventional distribution systems planning and analysis. IEEE Trans Ind Distrib 2015;9(3):209–20.
Appl 2003;39(5):1493–8. [41] Hung DQ, Mithulananthan N. Multiple distributed generator placement in primary
[21] Wang DTC, Ochoa LF, Harrison GP. DG impact on investment deferral: network distribution networks for loss reduction. IEEE Trans Ind Electron
planning and security of supply. IEEE Trans Power Syst 2010;25(2):1134–41. 2013;60(4):1700–8.
[22] Hung DQ, Mithulananthan N, Bansal RC. An optimal investment planning frame- [42] Gozel T, Hocaoglu MH. An analytical method for the sizing and siting of distributed
work for multiple distributed generation units in industrial distribution systems. generators in radial system. Electr Power Syst Res 2009;79(6):912–8.
Appl Energy 2014;124:62–72. [43] Hung DQ, Mithulananthan N, Bansal RC. Analytical expressions for DG allocation in
[23] Pandi VR, Zeineldin HH, Xiao W. Determining optimal location and size of dis- primary distribution networks. IEEE Trans Energy Convers 2010;25(3):814–20.
tributed generation resources considering harmonic and protection coordination [44] Khan H, Choudhary MA. Implementation of distributed generation (IDG) algorithm
limits. IEEE Trans Power Syst 2013;28:1245–54. for performance enhancement of distribution feeder under extreme load growth.
[24] Munoz-Delgado G, Contreras J, Arroy JM. Joint expansion planning of distributed Electr Power Energy Syst 2010;32(9):985–97.
generation and distribution networks. IEEE Trans Power Syst 2015;30(5):2579–90. [45] Bayat A, Bagheri A, Noroozian R. Optimal siting and sizing of distributed generation
[25] Lee SH, Song TE, Park JW. Power management and control for grid-connected DGs accompanied by reconfiguration of distribution networks for maximum loss re-
with intentional islanding operation of inverter. IEEE Trans Power Syst duction by using a new UVDA-based heuristic method. Electr Power Energy Syst
2013;28(2):1235–44. 2016;77:360–71.
[26] Rueda-Medina AC, Franco JF, Rider MJ, Padilha-Feltrin A, Romero RA. Mixed-in- [46] Kansal S, Kumar V, Tyagi B. Hybrid approach for optimal placement of multiple
teger linear programming approach for optimal type, size and allocation of dis- DGs of multiple types in distribution networks. Electr Power Energy Syst
tributed generation in radial distribution systems. Electr Power Syst Res 2016;75:226–35.
2013;97:133–43. [47] Devabalaji KR, Ravi K. Optimal size and siting of multiple DG and DSTATCOM in
[27] Atwa YM, EI-Saadany EF, Salama MMA, Seethapathy R. Optimal renewable re- radial distribution system using Bacterial Foraging Optimization Algorithm. Ain
source mix for distribution system energy loss minimization. IEEE Trans Power Syst Shams Eng J 2016;7(3):959–71.
2010;25(1):360–70. [48] Mohd Zin A, Ferdavani AK, Khairuddin AB, Naeini MM. Reconfiguration of radial
[28] Muthukumar K, Jayalalitha S. Optimal placement and sizing of distributed gen- electrical distribution network through minimum-current circular updating me-
erators and shunt capacitors for power loss minimization in radial distribution chanism method. IEEE Trans Power Syst 2012;27(2):968–74.
networks using hybrid heuristic search optimization technique. Electr Power [49] Baran ME, Wu FF. Network reconfiguration in distribution systems for loss reduc-
Energy Syst 2016;78:299–319. tion and load balancing. IEEE Trans Power Deliv 1989;4(2):1401–7.
[29] Sebastian W, Jose LR, Istvan E. Optimal allocation and sizing of dynamic VAr
85