Doctoral Dissertation Shibaura Institute of Technology
Doctoral Dissertation Shibaura Institute of Technology
Doctoral Dissertation Shibaura Institute of Technology
2014/SEPTEMBER
Supervisor
SEPTEMBER 2014
ii
And my supervisor
Prof. Dr. Goro Fujita
iii
ACKNOWLEDGEMENTS
In the name of Allah, most Gracious and Merciful. I am very grateful with his
permission; I was able to accomplish my Doctoral thesis.
ABSTRACT
TABLE OF CONTENTS
1 INTRODUCTION 1
1.1 Research Background 1
1.2 Research Questions 2
1.3 Problem Statement 3
1.4 Research Objectives 4
1.5 Scope of Work 5
1.6 Significance of the Research 5
1.6 Thesis Organization 6
2 LITERATURE REVIEW 7
2.1 Introduction 7
2.2 Network Reconfiguration 7
2.3 Installation of Distributed Generation in 8
Distribution System
2.4 Previous Research Works on Distributed 11
Generation Coordination
2.5 Previous Research Works on Network 14
Reconfiguration
vi
3.5 Conclusion 28
4.7 Conclusion 52
5.1 Introduction 53
5.2 Problem Formulation for Minimal System 53
Power Losses
vii
REFERENCES 69 -74
Appendices A – E 75-84
viii
LIST OF TABLES
LIST OF FIGURES
LIST OF SYMBOLS
FV - Fitness Value
Ob. Func - Objective Function
xijnew - New value of variable
xijold - Old value of variable
xkj - Neighbour value that is selected randomly
range(0,1) - Random value between 0 and 1
probi - Probability value
N - Total number of employed bees
x1 - DG output power
x2 - DG location
nbr - Number of lines
I - Current
i - Line number
R - resistance
n - Bus number
PDG,min - Lower bound of DG output power
PDG,max - Upper bound of DG output power
PDG - DG output power
Vn,min - Minimum allowable voltage
Vn,max - Maximum allowable voltage
Vn - Voltage at each bus
tdg - Total number of DG
tl - Total number of load
Pload - total amount of load consume at each bus
Psubstation - Power from substation
W - Unit for real power
xii
LIST OF ABBREVIATIONS
TS - Tabu Search
TS-SA - Tabu Seacrh-Simulated Annealing
X/R - Reactance/Resistor
xiv
LIST OF APPENDICES
INTRODUCTION
load demands, which increase significantly year by year. However, the progress in
enhancing the efficiency of the system is hindered by one major factor that is the
existence of high real power losses. Furthermore, the current trend of electrical
energy tariff is showing signs of increment and it is expected to escalate in the future;
hence, increasing the significance of power reduction. Consequently, many
researchers have devoted their effort to find the possible solution to minimize the
power losses whilst retaining the stability and security of the system.
There are several approaches suggested by researches to deal with the
reduction of power losses such as determining optimal DG coordination (output
power and location), capacitor coordination (size and location) and network
reconfiguration. However, the incorrect use of these approaches might deteriorate the
system’s performance and appear operational and planning problems of the
distribution system. The aim of this research is to find a solution that can harmonize
the network reconfiguration with DG coordination in order to further reduce the
power loss in the distribution network.
Several research questions that will arise when analyzing the effect of
combining the approaches on the distribution system are listed below:
ii. Which combination gives the best impact to the distribution system?
iii. What happens to the performance of voltage profile and stability index
when the coordination is performed on the system?
i. The total output of the DG unit must be less than the summation of total load
and power losses in the test system, in order to avoid reverse power flow to
the main grid.
ii. The maximum amounts of the DGs are only three units.
iii. The objective function for all case studies only focuses on single objective
which is total power losses.
iv. All the results obtained in this thesis is in the context of operational planning
division only, except the results obtained for DG coordination is a suggestion
to construction planning division for their references.
units of DG are used with five tie switches (for network reconfiguration process) in
the test network. Hence, the total variables involved are 11, Xi= [Location1 Location2
Location3 DG Output1 DG Output2 DG Output3 Switch1 Switch2 Switch3 Switch4
Switch5]. Thus, hybridization between the ABC and the AIS is proposed to solve the
problem and subsequently give minimum total power losses.
Overall, the analysis simultaneously provides a positive impact on
distribution system especially for the reduction of the power loss. Furthermore, the
use of optimization method assists and facilitates utilities in the planning division to
provide optimal coordination.
This thesis is organized in six chapters. The first chapter begins with an
overview of the research background and problem statement. It includes the most
important aspects for overall research including the objectives and scopes. Chapter
two presents a collection of literature from previous research work to tackle the
problems in coordination. Next, in the chapter three describes a new hybrid
optimization methods based on the ABC and the AIS algorithm.
Chapter four focuses on comparison between the ABC and the proposed
method in solving DG coordination. In addition, the proposed method will be
compared with a separate analysis that use single DG placement algorithm to
determine the location and using the AIBC to determine the output of the DG.
Chapter five discusses the effectiveness of simultaneous approach by combining the
DG coordination with network reconfiguration. Furthermore, a comparison with
other optimization methods will be performed.
Chapter six concludes of this research. Additionally, several ideas for future
works are also proposed.
CHAPTER 2
LITERATURE REVIEW
2.1 Introduction
In the first chapter, the overview of research background and other important
aspects to define this research have been touched. As previously discussed, there are
various approaches that have been used to reduce overall total power losses in the
distribution system. Hence, in this chapter, previous related research works on
network reconfiguration and DG coordination will be discussed.
Thick line is
Substation original branch Substation
1 (sectionalize switch) 1
2
8 2
8
12 2 12
9 2 9
3 9 3 9
13 10 13 10
15 3 10 15 3 10
13 4 13 4
14 11 14 11
14 4 11 14 4 11
15 5 12 15 5 12
5 5
6 6
16 Dotted line is 16
6 6
7 optional branch 7
7 (tie switch) 7
8 8
As the power system become more complex and challenging, the need for
advanced technology on the system become a priority for electrical utilities in order
to ensure power is supplied to the demands without any interruption. Furthermore,
9
with more concern to achieve high reliability and efficiency of distribution network,
introduction of the DG concept seems as an interesting solution by integrating small
scale of generator in the distribution system. Followed by the expected increase in
energy consumption in the near future, the importance of DG installations has
attracted the attention of many parties, especially electrical utilities.
A recent report conducted by Synapse Energy Economics, Inc. comprised of
several organizations comprised of several organizations estimate that DG capacity
in 2013 will increase by 191% at the end of 2021 in several states in New England,
United State of America [17]. Meanwhile, another report released by the California
public utilities commission on the 18th of March, 2013 stated that the state was
targeting a total DG capacity of 12 GW to be installed by 2020 [18]. Based on these
reports indicate that the installation of DG is indispensable in the future.
There are numerous definitions to describe the DG and until now there are
still long arguments to justify the meaning. Not only that, term and rating of the DG
also have various understanding depending on the region and the institutes [13].
Although there are different interpretations exist, the purpose of this small capacity
of DG is still same. It will act as an additional power support to the system instead of
depending solely on centralized generation; hence, power sharing between them can
benefit the entire power system. In refs. [5, 19-22], they explain in detail the impact
of the DG when connected to the network. Based from these studies, they pointed out
that installation of the DG offer many benefits, especially to the distribution network.
Some of the advantages as mentioned earlier are as follows:
Substation Substation
1 8 1 8
2 2
12 2 12
9 2 9
3 9 3 9
10 13 10
13 3 10 3 10
4 13 4
11 14 11
14 4 11 14 4 11
5 12 15 5 12
5 5
6 6
6 6
7 7
DG
7 7
8 8
The power loss is one of the constraints to achieve a more reliable and
economic distribution system. This problem exists due to the nature of the system
which only depends on the generation that located far away to transmit power
supply. In addition, characteristic of the distribution system in which has a low X/R
ratio makes further increment of power losses. One of available solution is by install
the DG in the existing problem where it is close to the load to help reduce
dependence on the generation; hence, minimize the power losses. The presence of
DG in the system without a proper sizing and location will actually worsen the
problem. Therefore, many researchers have studied the effects of optimizing the
location and size of the DG.
There are many methods published in the literature to determine location and
size of DG, which can be categorized into several groups namely analytical,
numerical, heuristic and meta-heuristic [6, 27]. Each of these groups has its own
advantages and disadvantages. In ref. [6], the author has briefly discussed the
comparison of each group in solving the DG coordination problem. Based on their
findings, they concluded that meta-heuristic and heuristic methods are more robust
and easier to implement in solving the complex problems.
Previously, calculations of location and size of DG are treated independently,
which means different method is applied to each of them. Candidate of DG location
can be determined either by using voltage sensitivity factor [28-30] or loss sensitivity
factor [31, 32] or analytical method [33-35] or suggested by the authors [36].
Basically, the voltage and the loss sensitivity factor are operate by placing a DG at
each node in the test system and choose the node which has a high sensitivity on the
changes of power losses. Meanwhile, computation of the DG size is using
optimization approach. However, when the analysis is done separately, the objective
function or goal of the study is not in best solution due to predetermined location and
considered as fixed parameter when to optimize the size of DG. Thus, it could lead
the solution trapped in local optimum.
In ref. [30], the authors have applied Evolution Programming (EP) method to
determine optimal sizing of DG, whereas for the DG location based on loss
sensitivity factor and voltage sensitivity index. From the simulation results, both of
12
the sensitivity gives different optimal location for installing the DG. This shows that
these two methods do not provide consistent results to determine the optimal location
of the DG. In addition, there is no further explanation given by the authors which
method should be preferred. In ref. [29] work has proposed a novel optimization
method based on EP, which is known as Quantum-Inspired Evolutionary
Programming (QIEP) to determine optimal size of DG. Conversely, determination of
optimal location is not included in the QIEP, but based on the voltage sensitivity
index. Based from the obtained result, proposed method showed better performance
compared to the EP. However, the analysis had only focused a single mode of DG,
which is PQ (Power-Reactive) mode. In ref. [34], the determination of the optimal
location of DG is based on analytical approach which is known as single DG
placement algorithm, whereas size of DG by using the PSO. The basic concepts of
the algorithm is by insert a DG at each bus one by one and select the bus that give
maximum value of power losses reduction as a candidate for best location. This
process is repeated for the next DG. In another study proposed by ref.[35], the same
concept in ref. [34] have been used to determine the location of DG, but using
different optimization method which is ABC to calculate the size of DG. Although
both of these references successfully provides better results compared to other
analytical approach, but it does not guarantee to give the best solution due to the
analysis in the DG coordination is done separately.
Recently, some authors have proposed a solution to determine both the
location and size of the DG in simultaneous that means using one method to solve
the DG coordination problem. Some of the aforementioned methods are Genetic
Algorithm (GA) [37-39], Particle Swarm Optimization (PSO) [40, 41], Artificial Bee
Colony (ABC) [42], Evolutionary Programming (EP) [43] and Firefly Algorithm
(FA) [44].
The work done by ref. [41] applies PSO to deals with DG coordination
problem. In this simulation analysis, the authors consider only one unit of DG to be
optimized. Despite the reduction in power loss can be achieved, there is no
comparison with other methods are done to show the effectiveness of the proposed
method. Ref. [43] suggested EP to determine optimal location and size of single DG
optimally. In this paper, the authors tested the effectiveness of the proposed method
with the AIS. Based on the comparison, the EP method gives better results in term of
power loss reduction, but the AIS method shows better voltage performance
13
compared to the EP. In ref. [44], the authors propose to solve coordination of DG by
using the FA. This method imitate behaviour of fireflies communicate with others
through the brightness of flashing light. In this study, they tested the proposed
method using two units of DG and compare with the GA method. From the analysis
of the simulation results, they concluded that the proposed method gives similar
results with the GA and shows that there is no improvement in the quality of
solutions. Another ref. [42], they propose ABC for the coordination of DG. The
authors consider two types of scenarios for the load conditions, which are normal
load and increase the load up to 50% of the normal value. Furthermore, in the
analysis, they optimize the power factor and consider the DG in the PQ mode.
Nevertheless, based on the proposed method, only one variable is chosen during the
“mutation” process and the selection is done at random (for example randomly
selects either DG location or DG size). Hence, in case of high dimension problems,
the ability of the ABC algorithm to search all possible solution area might be limited
due to the aforementioned construction.
Some authors have also proposed hybridized optimization methods by
combining two optimization methods to cancel out the discrepancies of each method
for achieving global or near global optimal solutions [45, 46].In ref. [45], the authors
propose a hybrid optimization method which combines the PSO and the GA. Each of
these optimization methods has different functions where PSO and GA are used to
determine optimal size and optimal location for DG, respectively. Although the
proposed method succeeded in giving the best answer compared to the PSO and GA
methods, but the solution is divided into two stages, and it is probable that the
solution is trapped in local minimum due to the calculation of the location depends
on the size determined by other method. Similarly, in ref. [46] the authors propose a
combination between Fuzzy logic and ABC. The solution is divided into two phases
where the focus on the first part is to solve the optimal location for DG by using
Fuzzy logic, while in the second phase; the ABC method is used to solve the DG
size. Based on the results obtained, it was found that the proposed method gives the
same power loss reduction with PSO method for all case studies that have been
conducted. In addition, the authors only consider the type of DG in PQ mode.
Despite there are various method that have been used to solve the DG
coordination problem, there are several disadvantages that needs to be considered
before selecting the appropriate methods. For example, the PSO consumes high
14
number of iterations, easy to get trapped in local optimum when solving the complex
problems [47] and weak at local search [48]. For the GA method, the basic operation
is similar to PSO, but the GA has a filtering process (crossover and mutation), thus it
makes the GA process a bit complicated than the PSO [49]. In addition, it stability
depends on the crossover and mutation rates [47]. Unlike the PSO, the AIS shows a
good performance to avoid the local optima [48], however, one of the disadvantages
of AIS, it have many parameter to tune [47]. Another method which is known as
ABC show better performance compared to other methods such as the PSO and the
GA [50]. One of the clear advantages of the ABC, it only have two parameter that
need to be tuned and thus make it easier to implement. However, one of drawbacks
of ABC comes when solving problems with high dimensions. Therefore, in this
thesis a new hybrid optimization based on ABC and AIS, known as Artificial
Immune Bee Colony (AIBC) was introduced to improve the ABC method.
Overall, the coordination of DG can be divided into two groups of analysis,
which are separate and simultaneous analysis. As previously discussed, the separate
analysis is performed by determine location and size of DG separately. On the other
hand, the simultaneous analysis is based on the location and size is determined
simultaneously by using single calculation. The methodology for both analyzes is
typically comprised of several groups of methods such as analytical, numerical,
heuristic and meta-heuristic. However, most of recent work focused on meta-
heuristic optimization either for single DG or multiple DG. In addition, simultaneous
analysis preferred over the separate analysis because it provides an opportunity for
the optimization method to search all possible solutions and thus quality of solution
can be achieved. Moreover, to obtain more reliable results and consistency in the
determination of DG coordination, some researchers have combine two types of
optimization to produce better results than the original optimization.
open and close positions respectively, whilst maintaining radiality of the system and
network constraints. The appropriate protection scheme for over current relays
demands the radial topology of the system.
There are several methods that have been used to solve the problem of
network reconfiguration. Same as the coordination of the DG approach, these
methods can be divided into several groups such as mathematics, heuristic, optimal
flow pattern and meta-heuristic [14, 15].Based on ref.[51], most of the methods
chosen by the researches to solve the network reconfiguration are heuristic and meta-
heuristic.
One of the early researchers who worked on network reconfiguration is in ref.
[52]. They proposed a simple formula to estimate the loss reduction when
transmitting a cluster of load from one feeder to another feeder. It is further followed
by the authors of the study [53] who introduced approximation of power flow based
on concept presented by the study [52]. In addition, the authors also take into account
the load balancing in their analysis. There are researchers that were based on the
paper [52] and published the studies in ref. [54-56]. All of these solutions are known
as heuristic algorithm.
Recently, modern optimization methods have been applied to network
reconfiguration by many authors. Ref. [57] implements GA to solve the
reconfiguration problem. Later on in ref. [58], they proposed Artificial Neural
Network (ANN) to solve network reconfiguration problem. In the analysis, the
authors use seven ANNs, which have a different topology for each of them.
However, this method has a problem because it depends on the size of the system. If
the size used is large then the amount of time taken to train the ANN takes time. To
solve the problem, the authors in ref. [59] determine the amount of ANN based on
number of sensitive switches where only four ANNs involved for the same test
system. Therefore, the amount of time needed to train the ANN can be greatly
reduced. Further investigation on implementation of ANN in network
reconfiguration has been done in ref. [60]. They use clustering technique to divide
the loads based on load level, which is used as inputs to the ANN. Furthermore, the
training set used does not depend on the size of the test system and only one ANN is
applied to solve the problem.
In ref. [61], the authors suggest a combination of methods between the SA
and Tabu Search (TS). Based on the results obtained, the hybrid method shows
16
similar performance in terms of loss reduction and minimum voltage, but faster in
computation time when compare with SA and TS, separately. Besides that, the
authors in ref. [62] consider Hybrid Differential Evolution (HDE) to solve the
network reconfiguration. One of the advantages of the proposed method compared
with Differential Evolution (DE) is be able to reduce the amount of population;
thereby reducing the computational time. To test the effectiveness of the proposed
method, they make a comparison with SA method and based on the results obtained
it was found that their method gives faster results. In another research work done by
the authors in ref. [63] has used Ant Colony Search Algorithm (ACSA). The
proposed method is inspired from the ants through the process of finding food. In the
study, they were compared with the GA and SA. Based on their findings, the
proposed method provides more consistent results compared with other methods
when the simulation is repeated 100 times.
In ref. [64], the authors have proposed method that combines between the EP
and Ant Colony Optimization (ACO). In the analysis, they take into account the
impact of changes in load on the network reconfiguration. The authors have divided
each day into several slots, where each slot comprises a maximum load current.
Therefore, information on the number of switch conversions that occur for each slot
can be obtained. Another ref. [65], Selective Particle Swarm Optimization (SPSO) is
proposed to solve the network reconfiguration problem. The proposed method is a
result of some modifications in the updating (mutation) process on Binary Particle
Swarm Optimization (BPSO) that enables it to search in selected space. Furthermore,
to facilitate the conversion process in finding the appropriate switches, they close all
the switches (tie and sectionalize) on the test system and as a result several loops are
formed. During the process of searching, only switches that are in these loops will be
considered for network reconfiguration process.
Overall, apart from using the DG approach to reduce power loss in the
distribution system, the network reconfiguration is another approach to help in
enhancing the performance of the system. This method works by controlling the
switches located on the branches and indirectly alter the original topology of the
system. Therefore, the solution by considering two approaches together maybe can
give positive impact, especially in terms of power loss reduction when compared to
the solution that performed separately (DG coordination or network reconfiguration).
17
Ref. [66] is among the earliest researchers examined the impact of DG on the
network reconfiguration. In the study, they used TS method to determine the
appropriate configuration when DG included in the test system. In addition, they also
investigate the effect of Reverse Power Flow (RPF) caused by the DG due to
amounts of power generated exceeds the load requirements. Later on in ref. [67],
they suggest PSO to solve distribution feeder reconfiguration with DG. In this paper,
the authors make a comparison between Dispatchable Distributed Generation (DDG)
and Non-Dispatchable Distributed Generation (NDDG). Based on the results
obtained, they concluded that DDG gives better results in terms of reduced power
loss compared to NDDG. Furthermore, it also managed to reduce conversion process
between switches.
In ref. [68], they propose a heuristic method that is based on sensitivity and
branch exchange to determine the best configuration of the test system. In this paper,
they consider four DGs unit with various capacities. Based on their analysis, it was
found that the reconfiguration of the network in the presence of DG can help to
reduce power loss almost half from the initial power losses. In another study carried
out by the ref. [69], they apply ACO method to determine optimal network
reconfiguration includes several DG units with different capacities and power factor.
Aside from the reduction of power loss, the authors also consider load balancing in
their analysis. In the meantime, the authors in ref. [70] also used ACO to resolve
network reconfiguration with DG. In addition, they also study the ability of network
reconfiguration to ensure that all the loads can be supplied by the DG when test
system is disconnected with main substation (islanding).
According to the refs. [66-70], determination of the location and size of DG is
determined either randomly or based on assumptions or considered as fixed at
particular location by the author. Nevertheless, those solution obtained in the
aforementioned references can be considered not the best solution due to the result
for network reconfiguration is not based from optimal coordination of DG. If the all
locations in the distribution system assumed eligible for the installation of the DG,
18
2.7 Conclusion
3.1 Introduction
heuristic is the best decision, due to simplicity of the algorithm in solving complex
system [72].
According to the Oxford dictionary, the word “meta” is referring to the
“beyond”, while the definition of the heuristic refer to engage in the process of
learning something for themselves. In some sense, meta-heuristic is a refinement
from heuristic algorithm in searching of near-optimal solutions [47]. In general,
meta-heuristic can be categorized into two types which are trajectory-based and
population-based. The difference between these types can be seen in the initial
solution where for the trajectory starts with single point and replaced with another
point (the best solution so far) during the iterative process, whereas the population-
based have multiple point during the initial solution [47].
Under the population-based, there are two classes of algorithm which are
evolutionary and swarm intelligence [50]. Among the examples for the population-
based optimization that has extensively used in wide range of engineering problems
are Genetic Algorithm (GA) [8], Particle Swarm Optimization (PSO) [9], Firefly
Algorithm (FA) [73], Ant Colony Optimization (ACO) [10], Artificial Immune
System (AIS) [74], Harmony Search Algorithm (HSA) [75] and Artificial Bee
Colony (ABC) [11].
In ref. [47], the authors have made brief overview on related to several meta-
heuristic methods. In addition, they also make a comparison for each method and list
all the advantages and disadvantages. Based from their observations, they conclude
that none of these methods can solve all engineering successfully. This conclusion is
known as “No Free Lunch Theorem” [76] which means that for certain types of
engineering problem, the optimization approaches may not gives similar results from
each others. Therefore, some improvement or modification must be made to the
original method for solving specific problem.
In the literature review, there are various recommendations that have been
suggested by researchers to improve performance of the original method in solving a
specific problem. Typically, this is done through some modifications on the original
formula or by integrating (hybrid) two methods to become a new method. Some
examples available in the literature review are QIEP [29], GA-PSO [45], Hybrid TS-
SA [61], HDE[62], EP-ACO [64], and SPSO [65].
In ref. [50], a comparative study between the ABC with other methods has
been carried out. They make comparisons with various kinds of numeric test
22
functions and simulation results showed that the ABC method gives better
performance compared with other methods. Therefore, in the thesis, Artificial Bee
Colony (ABC) will be the main focus of the study. Overview of theory for ABC and
mathematical formulation will be covered in the next section.
Another Food
Source
Food Source
Hive
Food Area
Employed Onlooker
Bees Bees
Dancing
area
Scout Bees
New Food
Area
In general, the ABC consists of three types of bees, which are employed,
onlooker and scout, where each type of bees represents different phases and task.
This algorithm can be divided into four phase which are initialization phase,
employed phase, onlookers phase and scout phase as shown in Fig. 3.2. There are
23
two important parameters involved to tune the ABC optimization which are size of
colony and limit. The size of colony is defined as total number of onlooker bees and
employed bees, whereas; for the limit is total numbers of onlooker’s bees multiply by
size of dimensional parameter to be optimized [11].
Start
Initialization
Employed Bees
Stage
Reached
Limit? No
Yes
No
Scout Bee Stage
Reached
Max. Cycle?
Yes
End
The analysis is started by the scout bees; they are deployed out randomly in
the possible search solution area to find the food source. After the acquisition of food
sources, there will be an information exchange between scout bees and employed
bees, according to Eq. 1. Specifically, the employed bees will find new food sources
that are positioned near to the location given by previous scout bees by using Eq. 2.
Based from this new location, they calculate the new fitness to be compared with the
24
old fitness (obtained during the initialization phase) by using greedy selection. In the
greedy selection process, only the higher of fitness value will be memorized by the
bees and used in the onlooker bee phase.
1
FVi
1 Ob.Funci (1)
where FVi is the fitness value and Ob.Funci is the objective formula of the study’s
problem.
xijnew xijold range(0,1) xijold xkj (2)
where xijnew and xijold are the new and previous value of variable, respectively. xkj is a
neighbor value that is selected randomly from jth dimension and range(0,1) is a
random value between 0 and 1.
The obtained information about the amount of food will be share with the
onlooker via dancing area. However, the possibility for the onlooker bees to be
chosen for extracting the food at the area where given by the employed bees is
dependent on the amount of nectar available as in Eq. 3. If the food in the area is
exhausted, the onlooker will be change to scout bees and search for new food source
locations. The process is repeated until the maximum cycle that has been set by the
user is reached. However, if the fitness value didn’t improve for a specified number
of times (until it reaches a certain predetermined limit), they will abandon the food
source and assign a scout bee to explore new food source location randomly. Figure
3.3 shows an overall process of bees finding the foods.
FVi
probi N (3)
FVi
i 1
where probi is probability value and N is the total number of employed bees.
25
Food Area 1
Food Area 1
Onlooker Food 2
Phase Onlooker Bees
Food 1
Employed Bees
Food Area 1
Hive
Scout Bees Food 3
Scout Phase Employed Bees Food 2
Food Area 2
Food 4 Food 5
Food 6
In original ABC, mutations process (Eq. 2) for new value of variables are
determined randomly for one variable at one time (for example either location or
output power of DG) for each cycle. However, for a problem that involved with
many variables, hence, the ability of the ABC algorithm to search all possible
26
Start
Read all
parameter
Generate parent
population
Duplicate
parent
population
Achieved convergence
criterion?
Yes
End
Start
Reached
Duplicate population based on Limit?
variables category
Yes
No
Generated new population (For each Generated new population for the Scout
population) for the Employed bees using bees and evaluate the fitness
Eq. (2) and calculate the fitness
Table 3.1 shows comparison of pseudo code for each algorithm. It can be
observed that the differences between the ABC and the AIBC are on duplication,
sorting and selection process taken from the AIS algorithm.
3.5 Conclusion
4.1 Introduction
Vi Vj
Ii Iij Ri X L,i
X C ,i X C ,j
n
Ii
YijV j (4)
j 1
Transform the Eq. 4 into polar form and thus the new equation can be
expressed as follows:
n
Ii
Yij V j ij j (5)
j 1
where ij is the angle of nodal admittance, whereas δj is voltage angle at bus j.
The complex power at the bus i can be compute as in Eq. 6 and substitute the
equation into the Eq. 4. The complex power at bus i can be expressed as below:
n
Pi jQi Vi i
Yij V j ij j (6)
j 1
Separating the real and imaginary and the equations can be expressed as
follows,
n
Pi
Vi V j Yij cos(ij i j )
j 1 (7)
31
n
Qi
Vi V j Yij sin(ij i j ) (8)
j 1
After the Eqs. 7 and 8 are computed; the values of voltage and phase angle
can be determined by using formula below.
Vi (k 1) Vi (k ) Vi (k )
(9)
Pi J1 J 2 i
Qi J 3 J 4 V (11)
The power losses at each line and total power losses can be determined as in
Eqs. 12 and 13, respectively.
nbr
PTotallossij
Ii 2 Ri (13)
i 1
where nbr is total number of line/branch.
Figure 4.2 shows the test system with DG. Similar as calculation of power
losses without the DG, the line losses and total power losses can formulated as
below:
nbr
PTotallossij
Ii ' 2 Ri (15)
i 1
where Iij’ and Ii’ are the new current flow at line between two buses after DG is
connected.
Vi Vj
Ii Iij’ Ri X L,i
I DG
DG X C ,i X C ,j
2
min f x1 , x2 I i Ri
nbr
(16)
i 1
where,
x1 continuous variables (DG output power),
x2 discrete variables (DG location),
nbr number of lines,
Ii current at line, i of distribution system.
Ri resistance at line, i of distribution system.
33
All DG units must operate within the tolerable value where PDG,min and PDG,max are
minimum and maximum limit of DG output power. In this study, all DGs are
assumed to have the same minimum and maximum output.
The voltage value for all buses in the network must be within the acceptable limit
where Vn,min and Vn,max are minimum and maximum allowable voltage at bus n.
tdg tl nbr 2
The total power output from all DG units in the system must less than the total load
and total power loss in the test system. Therefore, the reverse power flow to the main
can be avoided with this constraint. The tdg and tl are total number of DG and total
number of load, respectively. PDG is size of DG and Pload is total amount of load
consume at bus n.
34
tdg tl nbr 2
The total power injected into the system which is from the DG unit and main
substation, Psubstation must be same with total load and the total power loss (power
equilibrium concept).
In order to solve this problem, DG power output and DG location are selected
as variable x1 and x2, respectively. Total power losses of the system are selected as
objective function, Ob.Funci, whereas, the fitness value, FVi can be obtained from
the Ob.Funci by using the relation as in Eq. 1. At the first step of solution process,
initial population, xi, is randomly generated that consists of DG output power and
location with the size of the population is N number. Based from this initial
population, the fitness value is calculated. After that, the population will be
duplicated based on the number of categories involved (For example DG location
and DG output power) and new population will be generated by using Eq. 2 and then
new fitness is calculated. Since, x2 is considered as a discrete value, so the results
obtained from the Eq. 2 will be rounded off to the nearest real number. In this
“mutation” process, only variable that belong to each category will be modified by
using Eq. 2, whereas other variable retained at previous value.
Next, the new and old fitness will be compared with greedy selection method
where only high fitness is selected. From this selected value, the probability value
will be calculated by using Eq. 3 for selection in onlooker phase. This selection is
using the roulette wheel selection process that normally used in original ABC. Each
of the fitness value will be normalized to 1.0 and will be randomly selected by the
onlooker bees. After that, new population is obtained and new fitness value is
calculated. Again, greedy selection will be used to obtain best fitness value.
However, if the fitness value is not improved for a certain numbers (limit) during
iteration process, the current variables will be abandoned and replaced with the new
value that is selected randomly (scout bee phase). All fitness values for each
population will be combined together and next, sorting and selection process will be
35
done. This process is repeated until maximum cycle is reached. The proposed of DG
coordination by using AIBC is summarized as follows:
Figure 4.3 shows flow chart for AIBC to determine the optimal DG location
and DG output power in the distribution system whilst as a comparison, Fig. 4.4
shows flowchart for ABC in solving the problem.
36
Start
Read all
parameter
DG output
Yes category? No
Generated new population for the Generated new population for the
Employed Bees using Eq. (2) (DG Employed Bees using Eq. (2) (DG
output power only while DG location location only while DG output
maintain the same value) and calculate power maintain the same value) and
the fitness by using (1) calculate the fitness by using (1)
Evaluate the fitness by using Eq. Evaluate the fitness by using Eq.
(3) and allocated the Onlooker (3) and allocated the Onlooker
Bees Bees
Reached Reached
No No
Limit? Limit?
Yes Yes
Randomly generate a new output Randomly generate a new location
power of DG (Scout Bee phase) of DG (Scout Bee phase) and
and calculate the fitness calculate the fitness
Max. Cycle?
Yes
END
4.3 Related Data for 33-bus and 69-bus Test System for DG Coordination
All the case studies are applied to the 33-bus and 69-bus test system as shown
in Fig. 4.5 and Fig. 4.6, respectively. The 33-bus system consists of 33 buses and 32
branches, whereas for the 69-bus system have 69 buses with 68 branches. Both of the
test system connected to the main substation of 132/12.66 kV. All data related to the
test systems can be obtained in Appendix B [83] and C [84]. The power and voltage
base values are 100 MVA and 12.66 kV, respectively. In addition, all the DGs
assumed to function in Power-Voltage (PV) mode whilst the loads are presumed to
be power constant.
38
19 20 21 22
26 27 28 29 30 31 32 33
19 20 21
18 26 27 28 29 30 31 32
25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 2 3 4 5 6 7 8 9
132/12.66 kV 2 6 7 8 10 11 12 13 14
2 3
3
4
4
5
5
6 7 99 15 16 17 18
22
23 24
23 24 25
36 37 38 39 40 41 42 43 44 45 46
66 67
36 37 38 39 40 41 42 43 44 45 65 66
47 48 49 50 51 52
68 69
35
47 48 49 50 51 67 68
46
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
1 2 3 4 5 6 7 8 9
132/12.66 kV 6 10 11 12 13 14 15
2
2
3
3
4
4
5
5 6 77 8 99 16 17 18 19 20 21 22 23 24 25 26 27
27 52
28 29 30 31 32 33 34
8 8 8 53 54 55 56 57 58 59 60 61 62 63 64
28 29 30 31 32 33 34 35
53 54 55 56 57 58 59 60 61 62 63 64 65
test system is one and two, except the standard deviation of the AIBC is slightly
higher than the ABC. The difference of the standard deviation values is contributed
by the proposed method sometimes trapped at local optimum, but does not preclude
the proposed method to obtain best results. When the number of the DG increased to
three units, which means that the number of variables increases, it can be observed
that the power losses reduction acquired by the AIBC is slightly better than the ABC
at 11.6420 kW and 11.6433 kW, respectively. Even though this value is the same
value when the number is rounded to two decimal places, but it showed that the ABC
failed to obtain the same value as AIBC. Another significant difference that is
expected is calculation time to solve the problem. The proposed method takes a
longer time to solve the problem due to the AIBC have additional steps compared to
the original ABC.
Table 4.1: Summary of Results between ABC and AIBC for 33-bus system
Number of DG
Method Descriptions 1 2 3
Control Parameters
(L=Limit L=40 L=140 L=300
NoB=Number of NoB=40 NoB=70 NoB=100
Bees)
Optimal Location 6 13,30 14, 24, 30
Optimal Output 0.8381, 0.7515, 1.0589,
2.4736
Power (MW) 1.1194 1.0368
ABC
Best (kW) 61.5481 28.5127 11.6433
Mean (kW) 61.5482 28.5367 11.7215
Worst (kW) 61.5488 28.6751 12.4054
Standard Deviation
1.1861×10-4 4.0536×10-2 1.4531×10-1
(kW)
Calculation Time
48.84 87.92 145.30
(Seconds)
Control Parameters
(L=Limit L=40 L=140 L=300
NoB=Number of NoB=40 NoB=70 NoB=100
Bees)
Optimal Location 6 13, 30 14, 24, 30
Optimal Output 0.8381, 0.7496, 1.0644,
2.4736
AIBC Power (MW) 1.1194 1.0375
Best (kW) 61.5481 28.5127 11.6420
Mean (kW) 61.5481 28.5734 11.9189
Worst (kW) 61.5481 28.9612 13.3664
Standard Deviation 5.1500×10-6 1.0977×10-1 4.4385×10-1
(kW)
Calculation Time 57.30 97.17 170.35
(Seconds)
40
Further analysis was conducted to compare the convergence curves for ABC
and AIBC method in solving the problem. Figure 4.7 shows comparison of
convergence characteristic (50 runs at different random seeds) for both methods to
solve case for three DGs. It can be clearly seen that for all runs in the proposed
method converge faster than the original ABC and this indicates that the AIBC is
capable to find the optimal results at about half from total cycles compared to the
ABC that requires more cycles to converge.
100
ABC
90
80
70
Total Power Losses (kW)
60
50
40
30
20
10
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99
Cycle
80
70
Total Power Losses (kW)
60
50
40
30
20
10
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99
Cycle
Figure 4.8 illustrates voltage profiles based from the results for both methods.
Voltage profile of the system without DG included for comparison and to investigate
the impact of installed the DG on the test system. All the voltage values are
normalized to 1.0 p.u (Per Unit) where the voltage that closer to 1.0 p.u is better. It
can be observed that both voltage profiles show similar performance. Figure 4.9
shows the optimal location of all three units of DG on the 33-bus test system based
on the results obtained by the ABC and the AIBC.
1.02
0.98
0.96
Voltage (p.u)
0.94 ABC
AIBC
0.92 Without DG
0.9
0.88
0.86
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Bus Number
Fig. 4.8. Comparison of voltage profiles between ABC and AIBC on 33-bus
system
ABC &
19 20 21 22
AIBC
26 27 28 29 30 31 32 33
19 20 21
18 26 27 28 29 30 31 32
25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 2 3 4 5 6 7 8 9
132/12.66 kV 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
2 3 4 5 6 7 9
22 ABC &
AIBC
23 24
23 24 25
ABC &
AIBC
Fig. 4.9. Comparison of optimal location of DG between ABC and AIBC on 33-
bus system
42
For the 69-bus system, the results obtained are shown in Table 4.2. It can be
observed that AIBC gives better results than the ABC when the number of DG
increased to three, where the best results provided by the AIBC is 4.2730 kW and the
ABC is 4.9241 kW. In addition, the optimal locations for both methods are different
too. This proves that when number of variables is huge, the ABC fails to search
better results. Nevertheless, for the cases when the DG connected are one and two,
the ABC and the AIBC gives similar total power loss reduction.
Table 4.2: Summary of Results between ABC and AIBC for 69-bus system
Number of DG
Method Descriptions 1 2 3
Control
Parameter
L=40 L=140 L=450
(L=Limit
NoB=40 NoB=70 NoB=150
NoB=Number of
Bees)
Optimal Location 61 17,61 17, 50, 61
Optimal Output 0.5284, 0.5278, 0.7080,
1.8369
ABC Power (MW) 1.7392 1.7386
Best (kW) 23.2151 7.2115 4.9241
Mean (kW) 23.2151 7.2521 5.0192
Worst (kW) 23.2151 7.4979 5.3361
Standard
2.8368×10-6 7.2504×10-2 1.0446×10-1
Deviation (kW)
Calculation Time
208.24 470.11 888.41
(Seconds)
Control
Parameter
L=40 L=140 L=450
(L=Limit
NoB=40 NoB=70 NoB=150
NoB=Number of
Bees)
Optimal Location 61 17,61 11, 18, 61
Optimal Output 0.5284, 0.5023, 0.3789,
1.8369
AIBC Power (MW) 1.7388 1.6742
Best (kW) 23.2151 7.2115 4.2730
Mean (kW) 23.2151 7.2949 5.0317
Worst (kW) 23.2151 8.5087 6.9156
Standard
2.4472×10-6 2.0906×10-1 5.6626×10-1
Deviation (kW)
Calculation Time
234.42 506.59 907.60
(Seconds)
43
200
ABC
180
160
140
Total Power Losses (kW)
120
100
80
60
40
20
0
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
53
55
57
59
61
63
65
67
69
71
73
75
77
79
81
83
85
87
89
91
93
95
97
99
1
3
5
7
9
Cycle
70
60
Total Power Losses (kW)
50
40
30
20
10
0
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
53
55
57
59
61
63
65
67
69
71
73
75
77
79
81
83
85
87
89
91
93
95
97
99
1
3
5
7
9
Cycle
Comparison of voltage profile for ABC and AIBC is depicted in Fig. 4.11.
Both of these graphs show almost performance. Although there are several nodes that
have low voltage compared to other nodes, but still in the allowed range and stable
condition. This voltage differences is caused by two different locations of DG.
Figure 4.12 shows the optimal locations proposed by both methods. All locations for
DG is different except at node 61.
1.02
0.98
Voltage (p.u)
0.96
ABC
AIBC
Without DG
0.94
0.92
0.9
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69
Bus Number
Fig. 4.11. Comparison of voltage profiles between ABC and AIBC on 69-bus
system
36 37 38 39 40 41 42 43 44 45 46
66 67
36 37 38 39 40 41 42 43 44 45 65 66
47 48 49 50 51 52
68 69
35 ABC
47 48 49 50 51 67 68 AIBC
46
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
1 2 3 4 5 6 7 8 9
132/12.66 kV 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
2 3 4 5 6 7 9 24 25 26 27
28 29 30 31 32 33 34
8 8 8 53 54 55 56 57 58 59 60 61 62 63 64
28 29 30 31 32 33 34 35
53 54 55 56 57 58 59 60 61 62 63 64 65
AIBC
Based on all the results obtained from the comparison between ABC and
AIBC, it can be concluded that the proposed method (AIBC) shows better results
than ABC. Although both of these methods give the same results in determination of
one and two units of DG, but when the number of DG is three, significant differences
can be observed in the results for the location, output power of DG and total power
loss. In next section, comparison between the coordination of DG performed
simultaneously and separately will be conducted.
Table 4.4 shows summary of results for all three cases in term of DG
locations, outputs power and total power losses in the system. Without presence of
DG in the system, an initial power loss is 203.19 kW. Further power losses reduction
can be acquired when the DGs are consider in the system as shown in the results for
case 2 and 3. A total of 83.63% of power losses reduction can be obtained for case 2
compared to case 1. However, more power losses reduction can be achieved when
the DG location and size are determined simultaneously as shown in results of case
3. Based on this result, it indicates that the solution for case 2 trapped in a local
minimum due to the chosen for the installation of DG is not the best location and in
turn affects the selection of optimal output power. As a consequence, greater power
loss reduction could not be achieved. Optimal locations of the DG that obtained by
both case 2 and 3 are totally different as shown in Fig. 4.13. By using single DG
placement algorithm, the optimal locations are at bus 6, 16 and 25, whereas by using
the AIBC for the case 3 are at bus 14, 24 and 30.
19 20 21 22 Case 3
26 27 28 29 30 31 32 33
19 20 21
Case 2
18 26 27 28 29 30 31 32
25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 2 3 4 5 6 7 8 9
132/12.66 kV 2 6 7 8 10 11 12 13 14
2 3
3
4
4
5
5 6 7 99 15 16 17 18
Case 2
Case 3
22
23 24 Case 2
23 24 25
Case 3
Fig. 4.14 illustrates voltage profiles of the network for all cases. Based from
the results obtained, voltages at each bus for case 2 are between 0.96 p.u and 1 p.u
whereas for case 3, they vary from 0.99 p.u to 1.0 p.u. Although the performance of
case 2 and case 3 increased significantly compared to the case 1, but there are still
some buses showed a slight voltage reduction at 28, 29, 30 31, 32 and 33 as shown in
case 2. Overall, the determination of the location and output power of DG
simultaneously gives better voltage profile compared to the base case and case 2.
1
33 1 2
32 3
31 0.99 4
0.98
30 0.97 5
29 0.96 6
0.95
28 0.94 7
0.93
27 0.92 8
0.91
26 0.9 9
25 10
24 11
23 12
22 13
21 14
20 15
19 18 17 16
Fig. 4.14. Comparison of voltage profiles for all cases on 33-bus system
48
Figure 4.15 shows line loss at each branch for 33-bus system. The base case
shows worst performance of line losses compared to other case studies, especially at
branch number 2 and 5 which give high power loss as shown in Fig. 4.15(a).
Nevertheless, it can be clearly seen that by applying simultaneous approach of DGs
size and location, each branch gives lower power losses compared to the base case.
Although case 2 also gives low power losses pattern same as in case 3, but several
lines showed a significant increase of power losses which are at branch 25, 26, 27, 28
and 29 as illustrated in Fig. 4.15(b).
50
40
Case 1
Line Losses (kW)
30 Case 2
Case 3
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
Line Number
10
Case 2
8
Case 3
Line Losses (kW)
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
Line Number
Table 4.5 shows summary of results for 69-bus system. The initial power loss
in the test system without DG is 225.06 kW. Based on the results obtained, the
proposed method shows promising results in reducing the amount of power loss. A
total of 98.10% power loss reduction was achieved when the optimal location for DG
are 11, 18 and 61 with the optimal output power of 0.50, 0.38 MW and 1.67 MW,
respectively. However, when a separate analysis is used to determine the
coordination of DG, the percentage of reduction in power loss is at 97.81% where the
total power losses increase at about 15.46% from the proposed method. In addition,
the optimal location and output power of the DG obtained are quite different with the
case 3 where the locations are 17, 50 and 61 with the corresponding output power of
0.53 MW, 0.72 MW and 1.74 MW, respectively. Figure 4.16 depicts comparison of
optimal location of multiple DG for case 2 and case 3.
36 37 38 39 40 41 42 43 44 45 46
66 67
36 37 38 39 40 41 42 43 44 45 65 66
47 48 49 50 51 52
68 69
35 Case 2
Case 3
47 48 49 50 51 67 68
46
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
1 2 3 4 5 6 7 8 9
132/12.66 kV 6 10 11 12 13 14 15
2
2
3
3
4
4
5
5 6 77 8 99 16 17 18 19 20 21 22 23 24 25 26 27
Case 2 Case 2
27 52 Case 3
28 29 30 31 32 33 34
8 8 8 53 54 55 56 57 58 59 60 61 62 63 64
28 29 30 31 32 33 34 35
53 54 55 56 57 58 59 60 61 62 63 64 65
Case 3
Figure 4.17 shows a comparison of the voltage profiles of 69-bus system for
all cases. It can be clearly seen that cases 2 and 3 shows significant voltage
improvement compared to the case 1. Nevertheless, case 3 shows a better voltage
improvement. This can be observed at bus 5 to 15, 51 to 56 and 66 to 69 in which the
voltage increases slightly compared to the case 2. On the other hand, there are
several buses which are not experiencing voltage increments such as bus number 49
and 50, due to no DG connected at a nearby location. This can be proved as in case 2
there is one DG connected at bus 50 and thereby improving the voltage. Overall, case
3 shows better improvements in voltage profile compared to other cases.
1
68 69 1 2 3 4
66 67 5
65 0.99
6
64 7
63 0.98
8
62 9
61 0.97 10
60 0.96
11
59 12
58 0.95 13
57 0.94 14
56 0.93 15
55 0.92 16
54 0.91 17
53 0.9 18
52 19
51 20
50 21
49 22
48 23
47 24
46 25
45 26
44 27
43 28
42 29
41 30
40 39 32 31
38 37 36 35 34 33
Fig. 4.17. Comparison of voltage profiles for all cases on 69-bus system
Figure 4.18 shows line losses in each line for case 2 and case 3. Based on the
results obtained, it can be observed that all the lines losses are low. Although there is
a line at 48 gives a relatively high (due to no DG connected at a nearby location) fro
case 3, but overall simultaneous approach is better than in the case 2.
51
1.8
1.6
1.4
1.2
Line Losses (kW)
0.8 Case 2
Case 3
0.6
0.4
0.2
0
1 2 3 4 5 6 7 8 9 1011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768
Line Number
Table 4.6: Comparison with Other Methods for 33-bus and 69-bus test system
Test DG Total Power
Approach DG Output (MW)
System Locations Losses (kW)
Proposed 6 2.47 61.55
Analytical
Approach 6** 2.49 61.56
[33]*
33-Bus CSA [31]* 6** 2.28 63.26
Proposed 13,30 0.85,1.12 28.52
GA [32]* 6, 8** 1.72,0.84 55.67
Proposed 14,24,30 0.75,1.06,1.04 11.64
PSO[34] 6,15,25*** 1.76,0.58,0.78 33.19
Proposed 61 1.84 23.22
Analytical
Approach 61** 1.81 23.39
69-Bus [33]*
Proposed 17,61 0.53,1.74 7.21
CSA [31]* 59,61** 0.37,1.47 23.37
*The result had been simulated again due to different mode of DG and power flow in original manuscript.
**Locations are based on sensitivity index.
***Locations are based on single DG placement via differential equation technique.
****Locations are based on loss sensitivity factor.
4.7 Conclusion
5.1 Introduction
The network must maintain the radial configuration for the whole duration of
the optimization process. It means that, when a tie switch in the network is
closed; one of the lines (sectionalizes switch).in the network must be opened.
b) Isolation constraint:
The configuration process must ensure that all nodes (load and lines) in the
system are connected to the main supply (substation). In other words, there is
no load or line disconnected from the system after the configuration process.
Method for solving this problem is the same as in solving the coordination for
DG, except there are three categories of variables are involved, namely DG output
power, DG location and list of switches for network reconfiguration, where it can be
expressed as x1, x2 and x3, respectively. To facilitate the search for closing/opening
switches, the method in ref. [65] was used. In general, the search is performed by
isolating certain number of switches in each group where the total group depends on
the number of loops that exist when all tie switches are closed. In addition, only
switches that are in the loop is included in the search space. Therefore, this will
reduce the number of solutions that not feasible.
The proposed of DG coordination with network reconfiguration by using
AIBC is summarized as follows:
Figure 5.1 shows flow chart for AIBC to determine the optimal DG
coordination with network reconfiguration in the distribution system.
5.3 Related Data for 33-bus Test System for DG Coordination with Network
Reconfiguration.
Start
Read all
parameters
Yes
Duplicated the results based
on the category (DG size,
Location and switches)
DG sizing
Yes
category?
No
DG Location
No
category?
Yes
Generated new population for the Generated new population for the Generated new population for the
Employed Bees using Eq. (2) (DG Employed Bees using Eq. (2) (DG Employed Bees using Eq. (2) (DG
output power only while DG location output power only while DG location output power only while DG location
maintain the same value) and calculate maintain the same value) and calculate maintain the same value) and calculate
the fitness by using (1) the fitness by using (1) the fitness by using (1)
Evaluate the fitness by using Eq. Evaluate the fitness by using Eq. Evaluate the fitness by using Eq.
(3) and allocated the Onlooker (3) and allocated the Onlooker (3) and allocated the Onlooker
Bees Bees Bees
Calculate fitness by using Eq. Calculate fitness by using Eq. Calculate fitness by using Eq.
(1) (1) (1)
Reached Max.
Cycle?
Yes
END
132/12.66 kV
1 18
2
2 19
22 3 19
23 20
23 3 20
4
24
4 33 21
24 5
21
25 22
5 25
6
26
6
37 7 26 27
7
8 27 28
8
9 28 29
35
9 29
10 30
10 30 31
11
34 11 31 32
12
12 32
13 33
13
14
14
15
15 36
16
16
17
17
18
Five case studies are considered in this section as presented in Table 5.1.
Case 1 is set as a reference case that doesn’t imposed any coordination methods
(network reconfiguration or DG installations). For case 2, network reconfiguration is
applied on the case 1 (base case), whereas for the case 3, DGs are installed at
predetermine location and the outputs are optimize via AIBC. In Case 4, three DG
units are installed at a predetermined location and determination of optimal output
power and reconfiguration of the network will done simultaneously by using the
AIBC. On the other hand, for the case 5, both of DG coordination (locations and DG
outputs) and network reconfiguration will be executed simultaneously.
58
Table 5.2 shows the results for open switches, DGs’ outputs and locations,
total power loss, statistical analysis (best, average, worst and standard deviation),
saving percentage, increment of minimum voltage improvement and calculation time
achieved by all cases. In the case 3, since there is no reconfiguration technique
applied, the opened switch’s numbers are similar to the initial condition (case 1).
Furthermore, for the cases 3 and 4, the location of the DGs are based from results in
chapter 4 (separate analysis) and only the output and/or reconfiguration action are
optimized, whereas, for the case 5, all the variables are optimized by AIBC. All the
cases are runs independently for 50 times. In addition, all control parameters for
limit, maximum cycle and total number of bees are set at 120, 100 and 140,
respectively.
From the obtained results, the simultaneous reconfiguration and DG
coordination gives the lowest power losses value compared to the other cases. Nearly
to 95.53% of power loss reduction is achieved in the case 5 compared to original
total power losses value, 203.19 kW (Case 1). The power losses value that is given
by the case 5 is actually being influenced by combination of both factors: network
reconfiguration and DG coordination. This can be proved by referring to the results
for reconfiguration and DG sizing in the cases 2 and 3, respectively. In the case 2,
performing the reconfiguration process, the power losses in the network only reduced
59
up to 31.11 %, whereas, for the case 3 about 83.63 %. This shows that that the power
loss reductions for single approach (either optimal reconfiguration or optimal DG
output power) are not superior as in the power loss reductions as in the case 5. In the
case 4, savings of power loss reduction increases about 10.15 % from the saving in
the case 3. This increment is caused by DGs output power and reconfigurations are
determined, simultaneously. Figure 5.3 shows final result for case 5 when the
coordination of DG and network reconfiguration is carried out simultaneously.
Table 5.2: Summary of results of 33-bus system for DG coordination and network
reconfiguration
Case
Method Descriptions 1 2 3 4 5
33, 34,
Branch 7, 9, 33, 34, 3,23,28, 5, 11, 13,
35, 36,
Opened 14,28,32 35, 36, 37 34, 35 23, 27
37
Optimal
- - 6,16,25 6,16,25 8,25,32
Location
Optimal
1.70,0.53, 0.99,0.69, 1,1.14,
Output - -
0.77 1.44 0.79
Power (MW)
Total Power
203.19 139.98 33.26 12.63 9.08
Loss (kW)
Best (kW) - 139.98 33.26 12.63 9.08
Mean (kW) - 157.07 33.26 13.45 12.46
AIBC
Worst (kW) - 193.05 33.26 14.95 18.52
Standard
Deviation - 12.97 1.06×10-6 0.53 2.04
(kW)
Saving (%) 0 31.11 83.63 93.78 95.53
Minimum
Voltage
- 3.43 6.44 8.95 8.95
Improvement
(%)
Calculation
Time - 302.40 81.44 1194.94 1550.70
(Seconds)
60
19 20 21 22
DG
35
26 27 28 29 30 31 32 33
19 20 21
36
18 26 27 28 29 30 31 32
33 25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 2 3 4 6 7 8 9
132/12.66 kV 2 3 44 5
6
6 7
7
8
8 9
9 10 11 12 13 14 15 16 17 18
2 3
DG
34
22 23 24 37
23 24 25
DG
1
33 1 2
32 3
0.99
31 4
0.98
30 5
0.97
29 0.96 6
0.95
28 0.94
7
0.93
27 8
0.92
0.91
26 9
0.9
25 10
24 11
23 12
22 13
21 14
20 15
19 16
18 17
Fig. 5.4. Comparison of voltage profiles for all cases on 33-bus system
The Figure 5.5 presents results of the maximum C-VSI value after the
application of the reconfiguration and/or optimal DG coordination. It is mentioned
that the C-VSImax decreases for the case 2 to 5. Specifically, the C-VSImax equals to
0.1276, 0.0876, 0.0509 and 0.0339 for the cases 2, 3, 4 and 5, respectively. Thus, the
combination of DG and reconfiguration action implies the improvement of the
system stability. Consequently, the technique of the DG location, output power and
reconfiguration (case 5) provides the most stable system compared with other cases.
Case 1
Case 2
Case 3
Case 4
Case 5
Fig. 5.5. Comparison of stability index for all cases on 33-bus system
paper. The open branches for the case 2 are totally different with other reference,
with an exception of one branch open (branch number 28) is similar. In addition, all
DGs location for this case has different locations. While, for the last case which is
the case 5, there are no comparison due to the lack of available results in existing
literature. Ultimately, the proposed method produces the best results when compared
to the other methods.
From the whole analysis, optimization of DG location, output and
reconfiguration simultaneously gives the best option to the power system planner to
have the lowest power loss value in the existing network. However, this approach
can be applied only if the DG units are not being installed yet in the system. If the
DG unit has already existed, the simultaneous analysis between DG output power
and reconfiguration technique still can give superior results.
Table 5.4 shows variation of the best result of total power losses at different
parameters settings for solving the case 5. All simulation results are performed at 50
trials with different random seeds and each of trial is run for 100 cycles. Based on the
results obtained, it can be observed that the total number of bees and limit plays a
vital role to obtain better results. Increase the total number of bee’s means that the
algorithm will increase the exploitation process; meanwhile, for higher number of
limits mean exploration process will be reduced.
Parameter Setting
Total Total Power
Scenario
Number of Limit Losses (kW)
Bees
100 120 9.71
1 140 120 9.08
200 120 9.13
100 550 10.25
2 140 770 9.61
200 1100 9.51
64
5.5 Conclusion
The power loss is one of important issues need to be consider by the utility
especially to the operational planning division. Improper planning of power system,
especially distribution system can gives negative impacts such as high power losses
and unstable voltage condition. As previously discussed, there are several approaches
can be used to reduce the power losses, which are installed DG, installed capacitor
and change the original topology of the system (network reconfiguration). To deal
with any of this approach, special techniques or methods are required. In general,
these methods can be categorized into four main groups, which are analytical,
numerical, heuristic and meta-heuristic. Each of this group has its own advantages
and disadvantages.
Based from literature review, meta-heuristic is more popular among the
researchers due to easy to implement and robust. Therefore, in this thesis, a new
hybrid optimization, which is known as Artificial Immune Bee Colony (AIBC) was
introduced. The effectiveness of the proposed method has been tested at various
dimensional problems in solving the DG coordination and successfully gives better
quality of solutions compared to the original ABC.
Further investigation was carried to achieve better results (power loss
reduction) than the previous analysis. As discussed in the literature review section,
many researchers consider only one approach to reduce power loss, for example,
installing DG in the test system without regard to other approaches. This will cause
the process to further reduce the power loss may be difficult to achieved. Therefore,
67
6.2 Contributions
Some future works that can be done to improve this in the future are as
follows:
46. Padma Lalitha, M., V.C. Veera Reddy, and N. Sivarami Reddy, Application
of Fuzzy and ABC Algorithm for DG Placement for Minimum Loss in Radial
Distribution System. Iranian Journal of Electrical & Electronic Engineering,
2010. 6(4): p. 248-257.
47. Beheshti, Z. and S.M. Shamsuddin, A review of population-based meta-
heuristic algorithm. International Journal of Advances in Soft Computing and
its Applications, 2013. 5(1).
48. Moslemipour, G., T. Lee, and D. Rilling, A review of intelligent approaches
for designing dynamic and robust layouts in flexible manufacturing systems.
The International Journal of Advanced Manufacturing Technology, 2012.
60(1-4): p. 11-27.
49. Sumathi, S. and P. Surekha, Computational intelligence paradigms theory
and application using Matlab2010: CRC Press Taylor And Francis Group.
50. Karaboga, D. and B. Akay, A comparative study of Artificial Bee Colony
algorithm. Applied Mathematics and Computation, 2009. 214(1): p. 108-132.
51. K.Kiran Kumar, D.N.V.R., Dr. S. Kamakshaiah, Nishanth P M, State of Art
for Network Reconfiguration Methodologies of Distribution System. Journal
of Theoretical and Applied Information Technology, 2013. 57(1): p. 25-40.
52. Civanlar, S., et al., Distribution feeder reconfiguration for loss reduction.
Power Delivery, IEEE Transactions on, 1988. 3(3): p. 1217-1223.
53. Baran, M.E. and F.F. Wu, Network reconfiguration in distribution systems for
loss reduction and load balancing. Power Delivery, IEEE Transactions on,
1989. 4(2): p. 1401-1407.
54. Taylor, T. and D. Lubkeman, Implementation of heuristic search strategies
for distribution feeder reconfiguration. Power Delivery, IEEE Transactions
on, 1990. 5(1): p. 239-246.
55. Wagner, T.P., A.Y. Chikhani, and R. Hackam, Feeder reconfiguration for
loss reduction: an application of distribution automation. Power Delivery,
IEEE Transactions on, 1991. 6(4): p. 1922-1933.
56. Goswami, S.K. and S.K. Basu, A new algorithm for the reconfiguration of
distribution feeders for loss minimization. Power Delivery, IEEE
Transactions on, 1992. 7(3): p. 1484-1491.
57. Nara, K., et al., Implementation of genetic algorithm for distribution systems
loss minimum re-configuration. Power Systems, IEEE Transactions on, 1992.
7(3): p. 1044-1051.
58. Hoyong, K., K. Yunseok, and J. Kyung-Hee, Artificial neural-network based
feeder reconfiguration for loss reduction in distribution systems. Power
Delivery, IEEE Transactions on, 1993. 8(3): p. 1356-1366.
59. Kashem, M.A., et al., Artificial neural network approach to network
reconfiguration for loss minimization in distribution networks. International
Journal of Electrical Power & Energy Systems, 1998. 20(4): p. 247-258.
60. Salazar, H., R. Gallego, and R. Romero, Artificial neural networks and
clustering techniques applied in the reconfiguration of distribution systems.
Power Delivery, IEEE Transactions on, 2006. 21(3): p. 1735-1742.
61. Young-Jae, J. and K. Jae-Chul. Network reconfiguration in radial distribution
system using simulated annealing and Tabu search. in Power Engineering
Society Winter Meeting, 2000. IEEE. 2000.
62. Ching-Tzong, S. and L. Chu-Sheng, Network reconfiguration of distribution
systems using improved mixed-integer hybrid differential evolution. Power
Delivery, IEEE Transactions on, 2003. 18(3): p. 1022-1027.
73
63. Su, C.-T., C.-F. Chang, and J.-P. Chiou, Distribution network reconfiguration
for loss reduction by ant colony search algorithm. Electric Power Systems
Research, 2005. 75(2–3): p. 190-199.
64. Tsai, M.S. and C.C. Chu. Applications of hybrid EP-ACO for power
distribution system loss minimization under load variations. in Intelligent
System Application to Power Systems (ISAP), 2011 16th International
Conference on. 2011.
65. Tamer M. Khalil, A.V.G., Reconfiguration for loss reduction of distribution
systems using selective particle swarm optimization. International Journal of
Multidisciplinary Sciences and Engineering 2012. 3(6): p. 16-21.
66. Mishima, Y., et al., Method for minimum-loss reconfiguration of distribution
system by tabu search. Electrical Engineering in Japan, 2005. 152(2): p. 18-
25.
67. Olamaei, J., T. Niknam, and G. Gharehpetian, Application of particle swarm
optimization for distribution feeder reconfiguration considering distributed
generators. Applied Mathematics and Computation, 2008. 201(1–2): p. 575-
586.
68. Li, Q., et al. A New Reconfiguration Approach for Distribution System with
Distributed Generation. in Energy and Environment Technology, 2009.
ICEET '09. International Conference on. 2009.
69. Yuan-Kang, W., et al., Study of Reconfiguration for the Distribution System
With Distributed Generators. Power Delivery, IEEE Transactions on, 2010.
25(3): p. 1678-1685.
70. N.I. Voropai, B.B.-U., Multicriteria Reconfiguration of Distribution Network
with Distributed Generation. Journal of Electrical and Computer
Engineering, 2012. 2012.
71. Rao, R.S., et al., Power Loss Minimization in Distribution System Using
Network Reconfiguration in the Presence of Distributed Generation. Power
Systems, IEEE Transactions on, 2013. 28(1): p. 317-325.
72. Lee, K.S. and Z.W. Geem, A new meta-heuristic algorithm for continuous
engineering optimization: harmony search theory and practice. Computer
Methods in Applied Mechanics and Engineering, 2005. 194(36–38): p. 3902-
3933.
73. Yang, X.-S., Firefly Algorithms for Multimodal Optimization, in Stochastic
Algorithms: Foundations and Applications, O. Watanabe and T. Zeugmann,
Editors. 2009, Springer Berlin Heidelberg. p. 169-178.
74. Farmer, J.D., N.H. Packard, and A.S. Perelson, The immune system,
adaptation, and machine learning. Physica D: Nonlinear Phenomena, 1986.
22(1–3): p. 187-204.
75. Zong Woo Geem, Joong Hoon Kim, and G.V. Loganathan, A New Heuristic
Optimization Algorithm: Harmony Search. SIMULATION, 2001. 76(2): p.
60-68.
76. Wolpert, D.H. and W.G. Macready, No free lunch theorems for optimization.
Evolutionary Computation, IEEE Transactions on, 1997. 1(1): p. 67-82.
77. Adil Baykasoğlu, L.Ö., Pınar Tapkan Artificial Bee Colony Algorithm and Its
Application to Generalized Assignment Problem. 2007; Available from:
http://www.intechopen.com/books/swarm_intelligence_focus_on_ant_and_pa
rticle_swarm_optimization/artificial_bee_colony_algorithm_and_its_applicat
ion_to_generalized_assignment_problem.
74
78. Karaboga, N., A new design method based on artificial bee colony algorithm
for digital IIR filters. Journal of the Franklin Institute, 2009. 346(4): p. 328-
348.
79. Kavian, Y.S., et al., Routing and wavelength assignment in optical networks
using Artificial Bee Colony algorithm. Optik - International Journal for Light
and Electron Optics, 2013. 124(12): p. 1243-1249.
80. Akay, B. and D. Karaboga, A modified Artificial Bee Colony algorithm for
real-parameter optimization. Information Sciences, 2012. 192(0): p. 120-142.
81. Karaboga, D. and B. Akay, A modified Artificial Bee Colony (ABC) algorithm
for constrained optimization problems. Applied Soft Computing, 2011. 11(3):
p. 3021-3031.
82. Karaboga, D., et al., Artificial bee colony programming for symbolic
regression. Information Sciences, 2012. 209(0): p. 1-15.
83. Kashem, M.A., et al. A novel method for loss minimization in distribution
networks. in Electric Utility Deregulation and Restructuring and Power
Technologies, 2000. Proceedings. DRPT 2000. International Conference on.
2000.
84. Baran, M.E. and F.F. Wu, Optimal capacitor placement on radial distribution
systems. Power Delivery, IEEE Transactions on, 1989. 4(1): p. 725-734.
85. J.J. Jamian, H.M., M.W. Mustafa, H. Mokhlis, S.S. Adamu, Combined
Voltage Stability Index for Charging Station Effect on Distribution Network.
International Review of Electrical Engineering, 2011. 6(7): p. 3175-3184.
APPENDIX A
A comparison between ABC and AIBC method was carried out to see the
differences in the mutation process. To simplify this example, let assumed the total
number of variables are four (two for DG locations and another two for DG output
power) and number of Employed bees are two. For i=1, it can clearly been seen that
for the ABC, only one variable is changed after mutation process, whereas for the
AIBC, two variables changed.
ABC
Locations Output Power
x 2 3 0.5 0.7
1
5 0.4 0.8
i=1,j=3,k=2
x 2
4
After Mutation Process
x 2 3 0.6 0.7
1 Other variables retain
x 2 4 5 0.4 0.8
at previous values
AIBC
Locations Output Power
x 2 3 0.5 0.7
1
x 5 0.4 0.8
2 4
Duplicate process i=1,j=3,k=2
2 3 0.5 0.7 2 3 0.5 0.7
4 5 0.4 0.8 4 5 0.4 0.8
Mutation process
LIST OF PUBLICATIONS
A. List of Journals
1. Mohd Nabil Bin Muhtazaruddin, Mohd Wazir Mustafa, Goro Fujita “Short –
Term Load Forecasting Via Artificial Neural Network”, SEATUC
Symposium, Bangkok, Thailand, 5 – 7 March 2012.
83