Number 4 Volume 15 December 2009 Journal of Engineering

GENETIC ALGORITHM BASED LOAD FLOW SOLUTION

PROBLEM IN ELECTRICAL POWER SYSTEMS
By

Asst. Prof. Hassan A. Kubba Samir Sami Mahmood

Electrical Engineering Department M.Sc. Electrical Engineering
Baghdad University Baghdad University

ABSTRACT
In this paper, a proposed method based on real-coded genetic algorithm is presented and applied to
solve multiple load flow solution problem. Genetic algorithm is a kind of stochastic search algorithm
based on the mechanics of natural selection and natural genetics. They combine the concepts of
survival of the fittest with genetic operators such as selection, crossover and mutation abstracted
from nature to form a surprisingly robust mechanism that has been successfully applied to solve a
variety of search and optimization problems. Elitist method is also used in this research, and
blending models are implemented for crossover operator. In the proposed work, five busbars typical
test system and 362-bus Iraqi National Grid are used to demonstrate the efficiency and performance
of the proposed method. The results show that, genetic algorithm is on-line load flow solution
problem for small-scale power systems, but for large-scale power systems, it is recommended that
the load flow solution using genetic algorithm is for planning studies. The main important feature of
the purposed method is to give high accurate solution with respect to the conventional methods.

‫الخالصة‬
‫ تقدم طريقة مقترحة مبنية على اساس خوارزمية جينية مشفرة باالعداد الحقيقية لحل مسألة سريان الحمل متعددة‬، ‫في هذه البحث‬
‫ تعتبر الخوارزمية الجينية احدى طرق البحث العشوائية القائمة على تقنيات االنتخاب (االنتقاء) الطبيعي و الجينات‬.‫الحلول‬
‫ العبور و التغيار االحيائي‬،)‫ تجمع الخوارزمية الجينية مبادئ (بقاء االصلح) مع عوامل جينية كاالنتخاب (االنتقاء‬.‫الطبيعية‬
‫ تم في هذا‬.‫(الطفرة) المستخلصة من الطبيعة لتكوين تقنية متينة استخدمت بنجاح في حل مختلف مسائل البحث و ايجاد القيم المثلى‬
‫ لتبيان كفاءة و مدى فعالية الخوارزمية‬.‫البحث استخدام طريقة انتخاب االمثل باالضافة الى استخدام نماذج الخلط في عملية العبور‬
‫ تبين النتائج‬.‫ تم تطبيق الطريقة المقترحة على منظومة قدرة كهربائية قياسية‬،‫الجينية في حل مسائل سريان الحمل متعددة الحلول‬
‫كون الطريقة المقترحة مالئمة للحل اللحظي لمسائل سريان الحمل و بالتالي التطبيق العملي اثناء التشغيل لمنظومات القدرة صغيرة‬
‫ اهم‬.‫ فيوصي الباحث باستخدام الطريقة المقترحة الغراض التصميم و التخطيط‬،‫ أما بالنسبة لمنظومات الق درة كبيرة الحجم‬.‫الحجم‬
.‫خواص الطريقة المقترحة هي الحصول على نتائج و حلول لمسألة سريان الحمل بقيم عالية الدقة‬

KEYWORDS
Continuous Genetic Algorithm, Chromosome Crossover, Load Flow Analysis, Newton-
Raphson Method, Mutation, Multi-Objective Minimization.

1414
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

INTRODUCTION
With increasing computer speeds, researchers are increasingly applying artificial and computational
intelligence techniques, especially in power system problems. These methods offer several
advantages over traditional numerical methods. Among these techniques is that of genetic algorithm.
Genetic algorithms (GAs) are efficient stochastic search algorithms that emulate natural phenomena.
They have been used successfully to solve wide range of optimization problems. Because of
existence of local optima, these algorithms offer promise in solving large-scale problems. A genetic
algorithm mimics Darwin’s evolution process by implementing “survival of the fittest” strategy.
Genetic algorithm solves linear and nonlinear problems by exploring all regions of the search space
and exponentially exploiting promising areas through selection, crossover, and mutation operations.
They have been proven to be an effective and flexible optimization tool that can find optimal or near-
optimal solutions [Talib 2007]. In this study, an improved genetic algorithm solution of the load flow
problem is presented in order to minimize the total active and reactive power mismatches of the
given systems, a real-coded genetic algorithm has been implemented.

THE CONCEPTS OF LOAD FLOW ANALYSIS

The load flow studies are the backbone of the design of a power system. They are the means by
which the future operation of the system is known ahead of time. The load flow problem is one of the
basic problems in the power system engineering, and can be expressed as a set of non-linear
simultaneous algebraic equations, and thus it is to have multiple solutions [Woon 2004]. A load flow
study is the determination of voltage, current, power, and power factor or reactive power at various
points in an electrical network under existing or contemplated conditions of normal operation, so
power flow calculations provide power flows and voltages for a specified power system subject to
the regulating capability of generators, condensers, and tap changing under load transformers as well
as specified net interchange between individual operating systems. This information is essential for
the continuous evaluation of the current performance of a power system and for analyzing the
effectiveness of alternative plans for system expansion to meet increased load demand. The continual
expansion of the demand for electrical energy due to the growth of industries, commercial centers,
and residential sections requires never-ending additions to existing power systems. The systems
engineer must decide what components must be added to the system many years before they are put
into operation and he does this by means of power flow studies. The load flow solution usually
provides additional information, e.g. losses [Kubba 1987]. The load flow is the most frequently
carried out study by power utilities and is required to be performed at almost all the stages of power
system planning, optimization, operation, control, and contingency analysis.

ITERATIVE NUMERICAL (LF) SOLUTION METHODS

* NEWTON-RAPHSON METHOD
At each iteration of the Newton-Raphson method, the nonlinear problem is approximated by a linear
matrix equation (Jacobian matrix). The linearzing approximation can best be visualized in the case of
a single-variable problem as shown in figure (1).

1414
 Number 4 Volume 15 December 2009 Journal of Engineering

Fig. 1 Single-Variable Linear Approximation [Taylor 1967]

The Newton-Raphson load flow equation are

N
ΔPk  Pk  | Vk |  | Vm | ( g km cos  km  bkm sin  km )
sp

m 1
N
ΔQk  Qk  | Vk |  | Vm | ( g km sin  km  bkm cos  km )
sp

m 1

Vk | Vk | e j k I k  I k e j k
where , and  km   k   m (Kubba 1987).
Vm | Vm | e j m Ykm  Ykm e j km  g km  jbkm

The Jacobian matrix equation can be written as:

  1   P1 
    P 
 2  2
     
   
 J 1 J 2   N   PN 
 J J   V     Q 
 3 4
   
1 1

 V2   Q2 
     
   
V N  Q N 

where  k   kv 1   kv and Vk  Vkv1  Vkv when v is the iteration index and

 P P 
 J1 J 2    V 

J J 4   Q Q 
Newton-Raphson load flow method may have the final formulation of
 3 
  V 
(Grisby 2007)

1411
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

 P   J 1 J 2    
Q    J J 4  V 
   3

Using Newton-Raphson method to solve the mismatch powers equations either in polar coordinates
with (∆V) and (∆δ) as variables or in rectangular coordinates with (∆e) and (∆f) as variables [Al-
Shakarchi 1973].

* FAST DECOUPLED LOAD FLOW METHOD

Fast decoupled load flow method, possibly the most popular method used by utilities, is well known
for its speed of solution, reduced memory, and reliable convergence (Nanda 1987). The algorithm is
simpler, faster and more reliable than Newton’s method and has lower storage requirements. The fast
decoupled load flow method is based on Newton’s load flow method with the modifications of
neglecting the J2 and J3 Jacobian submatrices due to the weak coupling between “P-V” and “Q-δ”
quantities in power transmission system. Together with other approximations, the fast decoupled
load flow equations become:

 ΔP 
 V   B Δ 

 ΔQ 
 V   B ΔV 
1 1
where Bkm   for m≠k and Bkk   for m=k
x km mk x km

   Bkm for m≠k and Bkk   Bkm

Bkm for m=k (Stott and Alsac 1974)
mk

GENETIC ALGORITHM
Genetic algorithms (GAs) are adaptive methods which may be used to solve search and optimization
problems. Over many generations, natural populations evolve according to the principles of natural
selection and “survival of the fittest”. By mimicking this process, genetic algorithms are able to
“evolve” solutions to real world problems, if they have been suitably encoded [Holland 1975].
Genetic algorithms work with a “population of individuals”, each representing a possible solution to
a given problem. Each individual is assigned a “fitness score” according to how good a solution to
the problem it is. The highly-fit individuals are given opportunities to “reproduce”, by “cross
breeding” with other individuals in the population. This produces new individuals as “offspring”,
which share some features taken from each “parent”. The least fit members of the population are less
likely to get selected for reproduction, and so “die out”. A whole new population of possible
solutions is thus produced by selecting the best individuals from the current “generation”, and mating
them to produce a new set of individuals. This new generation contains a higher proportion of the
characteristics possessed by the good members of the previous generation. In this way, over many
generations, good characteristics are spread throughout the population. By favouring the mating of
the more fit individuals, the most promising areas of the search space are explored. If the genetic
algorithm has been designed well, the population will converge to an optimal solution to the
problem. There are some differences between genetic algorithms and traditional searching algorithms
(such as numerical techniques). They could be summarized as follows [Younes 2006]:

1414
 Number 4 Volume 15 December 2009 Journal of Engineering

 The algorithms work with a population of strings, searching many peaks in parallel, as
opposed to a single point.
 Genetic algorithms work directly with strings of characters representing the parameters set,
not the parameters themselves.
 Genetic algorithms use probabilistic transition rules instead of deterministic rules.
 Genetic algorithms use objective function information instead of derivatives or other
auxiliary knowledge (convexity, modality, continuity, differentiability).
 Genetic algorithms have the potential to find solutions in many different areas of the search
space simultaneously.

GENETIC ALGORITHM IMPLEMENTATION

A simple genetic algorithm is an iterative procedure, which maintains a constant size population of
candidate solutions. During each iteration step (generation), three genetic operators (reproduction,
crossover, and mutation) are performing to generate new populations (offspring), and the
chromosomes of the new populations are evaluated via the value of the fitness which is related to
cost function. Based on these genetic operators and the evaluations, the better new populations of
candidate solutions are performed [Younes 2006]. With the above description, the three steps in
executing the genetic algorithm operating on fixed-length character strings are as follows:
1. Randomly create an initial population of individual fixed-length character strings.
2. Iteratively perform the following sub steps on the population of strings until the termination
criterion has been satisfied:
A. Assign a fitness value to each individual in the population using the fitness measure.
B. Create a new population of strings by applying the following three genetic operations. The
genetic operations are applied to individual string(s) in the population chosen with a probability
based on fitness.
i. Reproduce an existing individual string by copying it into the new population.
ii. Create two new strings from two existing strings by genetically recombining substrings using
the crossover operation at a randomly chosen crossover point.
iii. Create a new string from an existing string by randomly mutating the character at one randomly
chosen position in the string.
3. The string that is identified by the method of result designation (e.g. the best-so-far individual)
is designated as the result of the genetic algorithm for the run. This result may represent a solution
(or an approximate solution) to the problem.

Now, we'll discuss briefly each step of the implementation of the genetic algorithm:

- CHROMOSOME REPRESENTATION
Genetic algorithms operate on representations of solutions to problems. Since they work with
encoded parameters of the optimization problem, the choice of a representation form has a large
impact on the performance. There are different ways of encoding solutions, and probably no single
best way for all problems. The performance of genetic algorithms depends on the choice of a
suitable representation technique. Most genetic algorithms applications use Holland’s fixed-length
simple binary coding. This is historically the most widely used representation. Each chromosome
is comprised of zeroes and ones, with each bit representing a gene [Abdul-Haleem 2005]. A
conceptually simpler technique would be the real-coded representation, in which each
chromosome vector is coded as a vector of floating point numbers of the same length as the
solution vector. Each element was forced to be within the desired range, and the genetic operators
were carefully designed to preserve this requirement [Michalewicz 1996].

1414
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

- POPULATION INITIALIZATION
In the genetic algorithm, populations of chromosomes are created randomly by generating the
required number of individuals using a random number generator that uniformly distributes numbers
in the desired range. The extended random initialization is a variation whereby a number of random
initializations are tried for each individual and the one with the best performance is chosen for the
initial population. Other users of genetic algorithms have seeded the initial population with some
individuals that are known to be in the vicinity of the global optimum. This approach is only
applicable if the nature of the problem is well understood beforehand or if the genetic algorithm is
used in conjunction with knowledge based system [Abdul-Haleem 2005].

- OBJECTIVE FUNCTION OR FITNESS FUNCTION

The objective function is used to provide a measure of how individuals have performed in the
problem domain. In the case of a minimization problem, the mostly fit individuals will have the
lowest numerical value of the associated objective function. This raw measure of fitness is usually
only used as an intermediate stage in determining the relative performance of individuals in a genetic
algorithm. Another function is the fitness function, is normally used to transform the objective
function value into a measure of relative fitness [Ibrahim 2005].

- REPRODUCTION
The selection, or competition, is a stochastic process in which the chance of an individual surviving
is proportional to its adaptation level. The adaptation is measured by the phenotype (search point,
solution) evolution, that is, the characteristics presented by an individual in the problem environment
(search space). The genetic algorithm, through selection, determines which individuals will go to the
reproduction phase. There are several selection methods, where the fittest individuals from each
generation are preferentially chosen for reproduction [Zamanan 2006]. Some of these methods are:
a. ROULETTE WHEEL SELECTION METHOD.
b. TOURNAMENT SELECTION METHOD.

- RECOMBINATION
Recombination produces new individuals in combining the information contained in two or more
parents (parents-mating population). This is done by combining the variable values of the parents.
Depending on the representation of the variables, different methods must be used. For the
recombination of binary valued variables, the name “crossover” is established. This has mainly
historical reasons. During the recombination of binary variables, only parts of the individuals are
exchanged between the individuals. Depending on the number of parts, the individuals are divided
before the exchange of variables (the number of cross points). The number of cross points
distinguishes the methods.
In single-point crossover, one crossover position a € [ 1, 2 , ……… , Nvar-1 ], where (Nvar) is the
number of variables of an individual, is selected uniformly at random and the variables exchanged
between the individuals about this point, then two new offspring are produced.
In double-point crossover, two crossover positions are selected uniformly at random and the
variables exchanged between the individuals between these points. Then two new offspring are
produced. For multi-point crossover, (c) crossover positions ai € [ 1 , 2 , ……… , Nvar-1 ]; i = 1: m,
where (Nvar) is the number of variables of an individual, are chosen at random with no duplicates and
sorted into ascending order. Then, the variables between successive crossover points are exchanged
between the two points to produce two new offspring. The section between the first variable and the
first crossover point is not changed between individuals. Uniform crossover generalizes this scheme

1414
 Number 4 Volume 15 December 2009 Journal of Engineering

to make every locus a potential crossover point. A crossover mask, the same length as the individual
structure is created at random and the parity of the bits in the mask indicate which parent will supply
the offspring with which bits [Pohlheim 2005].

- MUTATION
By mutation, individuals are randomly altered. These variations (mutation steps) are mostly small.
They will be applied to the variables of the individuals with a low probability (mutation probability
or mutation rate). Normally, offspring are mutated after being created by recombination [Pohlheim
2005]. In this process, randomly selected bits of randomly selected strings are changed from (0) to
(1) and vice versa. This process occurs according to pre-specified probability. Usually, less than 5%
of bits are changed in this process. Mutation process is used to escape from probable local optimum
[Zamanan 2006].

- TERMINATION OF THE GENETIC ALGORITHM

Termination criteria or convergence criteria for genetic process may be triggered by finding an
acceptable approximate solution and bring the search to halt. The termination criteria can be one or
more of the following criteria [Abdul-Haleem 2005]:
a. Using diversity measure.
b. After a specified number of generations.
c. Finding an acceptable approximate solution.
d. Repetition until no change in the solution.

LOAD FLOW SOLUTION USING GENETIC ALGORITHM

The binary genetic algorithm is conceived to solve many optimization problems that stump
traditional techniques. But, what if we are attempting to solve a problem where the values of the
variables are continuous and we want to define them to the full machine precision? In such a
problem, each variable requires many bits to represent it. If the number of variables is large, the size
of the chromosome is also large. Of course, ones and zeros are not the only way to represent a
variable. One could, in principle, use any representation conceivable for encoding the variables.
When the variables are naturally quantized, the binary genetic algorithm fits nicely. However, when
the variables are continuous, it is more logical to represent them by floating-point numbers. In
addition, since the binary genetic algorithm has its precision limited by the binary representation of
variables, using floating-point numbers instead easily allows representation to the machine precision.
This continuous genetic algorithm also has the advantage of requiring less storage than the binary
genetic algorithm because a single floating-point number represents the variable instead of (Nbits)
integers. The continuous genetic algorithm is inherently faster than the binary genetic algorithm,
because the chromosomes do not have to be decoded prior to the evaluation of the cost function
(objective function) [Haupt 2004].

COMPONENTS OF A CONTINUOUS GENETIC ALGORITHM

- THE VARIABLES AND COST FUNCTION
A cost function generates an output from a set of input variables (a chromosome). The cost function
may be a mathematical function or an experiment. The objective is to modify the output in some
desirable fashion by finding the appropriate values for the input variables. The goal is to solve some
optimization problem where we search for an optimum (minimum) solution in terms of the variables
of the problem. If the chromosome has (Nvar) variables (an N-dimensional optimization problem)
given by (b1, b2, ………, bNvar), then the chromosome is written as an array with (1×Nvar) elements
so that:

1414
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

chromosome = [b1, b2, b3, ………, bNvar]

In this case, the variable values are represented as floating-point numbers. Each chromosome has a
cost found by evaluating the cost function (f) at the variables (b1, b2, ………, bNvar).

cost = f (chromosome) = f (b1, b2, ………, bNvar)

Equations (10) and (11) along with applicable constraints constitute the problem to be solved [Haupt
2004].
Our primary problem in this research is the continuous functions introduced below. The two cost
functions are [Kubba 1987]:

n
ΔPk = Pksp – Vk ∑ Vm (Gkm cosδkm + Bkm sinδkm)
m=1
for “PV” and “PQ” busses

n
ΔQk = Qksp – Vk ∑ Vm (Gkm sinδkm – Bkm cosδkm)
m=1
for “PQ” busses only

Where δkm = δk – δm, and

(ΔPk) is the mismatch active power at bus (k) and (ΔQk) is the mismatch reactive power at bus (k).
(Vk, Vm, δk, δm) are the voltage magnitude and angle at busses (k) and (m) respectively, which are
the variables of the two cost functions.

* VARIABLE ENCODIND, PRECISION, AND BOUNDS

Here, the difference between binary and continuous genetic algorithms is shown. It is no longer
needed to consider how many bits are necessary to represent accurately a value. Instead, (V) and (δ)
have continuous values that are limited between appropriate bounds (which are in our problem, 0.95
≤ V ≤ 1.05 and -5˚ ≤ δ ≤ 5˚ for “5 busbars” typical test system and 0.9≤V≤1.1, -20o ≤δ≤20o for Iraqi
National Grid). Although the values are continuous, a digital computer represents numbers by a finite
number of bits. When we refer to the continuous genetic algorithm, it means that the computer uses
its internal precision and roundoff to define the precision of the value. Now, the algorithm is limited
in precision to the roundoff error of the computer [Haupt 2004].

* INITIAL POPULATION
The genetic algorithm starts with a group of chromosomes known as the population. We define an
initial population of (Nind) chromosomes. A matrix represents the population with each row in the
matrix being a (1×Nvar) array (chromosome) of continuous values. Given an initial population of
(Nind) chromosomes, the full matrix of (Nind×Nvar) random values is generated.
All variables are normalized to have values between (0) and (1), the range of a uniform random
number generator. The values of a variable are “unnormalized” in the cost function. If the range of
values is between (blo) and (bhi), then the unnormalized values are given by:

b = (bhi – blo) bnorm + blo

where:

1414
 Number 4 Volume 15 December 2009 Journal of Engineering

bhi : highest number in the variable range.

blo : lowest number in the variable range.
bnorm : normalized value of variable.
This society of chromosomes is not a democracy: the individual chromosomes are not all created
equal. Each one’s worth is assessed by the cost function. So at this point, the chromosomes are
passed to the cost function for evaluation [Haupt 2004].

- NATURAL SELECTION
Survival of the fittest translates into discarding the chromosomes with the highest cost. First, the
(Nind) costs and associated chromosomes are ranked from lowest cost to highest cost. Then, only the
best are selected to continue, while the rest are deleted. The selection rate, (Xrate), is the fraction of
(Nind) that survives for the next step of mating. The number of chromosomes that are kept each
generation is:

Nkeep = Xrate . Nind

Natural selection occurs each generation or iteration of the algorithm. Of the (Nind) chromosomes,
only the top (Nkeep) survive for mating, and the bottom (Nind – Nkeep) are discarded to make room for
the new offspring. Deciding how many chromosomes to keep is somewhat arbitrary. Letting only a
few chromosomes survive to the next generation limits the available genes in the offspring. Keeping
too many chromosomes allows bad performers a chance to contribute their traits to the next
generation. We often keep 50% (Xrate=0.5) in the natural selection process [Haupt 2004].

- SELECTION FOR MATING

In this process, two chromosomes are selected from the mating pool of (Nkeep) chromosomes to
produce two new offspring. Pairing takes place in the mating population until (Nind – Nkeep) offspring
are born to replace the discarded chromosomes. Pairing chromosomes in a genetic algorithm can be
as interesting and varied as pairing in an animal species. We’ll look at a variety of selection methods.
a. Weighted random pairing (roulette-wheel): which is divided into:
i. Rank weighting.
ii. Cost weighting.
b. Tournament selection.
Each of the parent selection schemes results in a different set of parents. As such, the composition of
the next generation is different for each selection scheme. Roulette-wheel and tournament selection
are standard for most genetic algorithms. It is very difficult to give advice on which selection scheme
works best. In our problem, we follow the rank-weighting roulette-wheel and tournament parent
selection procedures [Haupt 2004].

- RECOMBINATION
As for the binary algorithm, two parents are chosen, and the offspring are some combination of these
parents. Many different approaches have been tried for crossing over in continuous genetic
algorithm. The simplest methods choose one or more points in the chromosome to mark as the
crossover points. Then the variables between these points are merely swapped between the two
parents. The problem with real-valued crossover methods is that no new information is introduced:
each continuous value that was randomly initiated in the initial population is propagated to the next
generation, only in different combinations. Although this strategy works fine for binary
representations, there is now a continuum of values, and in this continuum we are merely
interchanging two data points. These approaches totally rely on mutation to introduce new genetic

1444
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

material. The blending models remedy this problem by finding ways to combine variable values
from the two parents into new variable values in the offspring.
In our problem, we want to find a way to closely mimic the advantages of the binary genetic
algorithm scheme. It begins by randomly selecting a variable in the first pair of parents to be the
crossover point:

c = roundup {random*Nvar}

We’ll let

parent 1 = [bm1, bm2, ……, bmc, ……, bmNvar]

parent 2 = [bd1, bd2, ……, bdc, ……, bdNvar]

Where (m) and (d) subscripts discriminate between the mom and dad parent. Then, the selected
variables are combined to form new variables that will appear in the children:

bnew1 = bmc – β (bmc – bdc)

bnew2 = bdc + β (bmc – bdc)

Where β is a random value between 0 and 1. The final step is to complete the crossover with the rest
of the chromosome as in binary genetic algorithm:

offspring 1 = [bm1, bm2, ……, bnew1, ……, bdNvar]

offspring 2 = [bd1, bd2, ……, bnew2, ……, bmNvar]

If the first variable of the chromosomes is selected, then only the variables to the right of the selected
variable are swapped. If the last variable of the chromosomes is selected, then only the variables to
the left of the selected variable are swapped. This method does not allow offspring variables outside
the bounds set by the parent unless β > 1 [Haupt 2004].

* MUTATION
Random mutations alter a certain percentage of the genes in the list of chromosomes. We can
sometimes find our method working too well. If care is not taken, the genetic algorithm can converge
too quickly into one region of the cost surface. If this area is in the region of the global minimum,
that is good. However, some functions, such as the one we are modeling, have many local minima. If
nothing is done to solve this tendency to converge quickly, it may end up in a local rather than a
global minimum. To avoid this problem of overly fast convergence (premature convergence), the
routine is forced to explore other areas of the cost surface by randomly introducing changes, or
mutations, in some of the variables. The basic method of mutation is not much more complicated for
the continuous genetic algorithm. Do we also allow mutations on the best solutions? Generally not.
They are designated as elite solutions destined to propagate unchanged. Such elitism is very common
in genetic algorithms. Why throw away a perfectly good answer? The equation of mutation used here
is [Haupt 2004]:

bmut = (bhi – blo) bnorm + blo ; bmut: variable under mutation.

1444
 Number 4 Volume 15 December 2009 Journal of Engineering

MULTIPLE OBJECTIVE OPTIMIZATION (MOO)

In many applications, the cost function has multiple, often times, conflicting objectives. The most
important approach to (MOO) are: weighted cost functions [Haupt 2004].
N
cost = ∑ wi fi
i=1

Where:

fi is the cost function (i).

N
wi is the weighting factor and ∑ wi = 1.
i=1
The problem with this method is determining appropriate values of (wi). Different weights produce different
costs for the same (fi). This approach requires assumptions on the relative worth of the cost functions prior to
running the genetic algorithm. Implementing this multiple objective optimization approach in a genetic
algorithm only requires modifying the cost function to fit the form of eq. (17) and does not require any
modification to the genetic algorithm [Haupt 2004]. Thus,

cost = w f1 + (1-w) f2

This approach is adopted in this research for its simplicity, easy of programming and gives us the required
accuracy. Here, (w) is chosen to be (0.5).

1444
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

Define cost function, cost variables

Select genetic algorithm parameters

Generate initial population

Find cost for each chromosome

Select mates

Mating and crossover

Mutation

Find cost for each chromosome

NO
YES Convergence Test

End

Fig. (2) Flowchart of a Continuous Genetic Algorithm

IMPLEMENTATION AND RESULTS

The proposed continuous (real-coded) genetic algorithm is demonstrated on two test systems namely,
5-buses test system [Stagg 1968] which lines data and buses data are present in appendix. The

1444
 Number 4 Volume 15 December 2009 Journal of Engineering

second system is the 362-bus Iraqi National Grid (ING) with 599 branches. The software is
implemented by using MATLAB version 7 with Pentium 4, 2 GHz CPU, and 2 Gbyte RAM. The
input data are the nodal admittance matrix, buses data (specified values), and the slack voltage
1. Five Buses Typical Test System Results
Table (1) shows the results of the application of the genetic algorithm to “5 bus-bars” typical test
system to find the active and reactive power mismatches described in eqs. (12) and (13) for accuracy
of 0.001 p.u (0.1 MW/MVAR) and the voltage magnitude and phase angle associated with each
busbar. Because of the stochastic nature of the genetic algorithm process, each independent run will
probably produce a different number of generations and consequently the computation time and the
best amongst these should be chosen. The best of the (10) implementation runs are shown in this
table. The base quantities are 132 KV and 100 MVA.

Table (2) Load Flow Solution for “5 busbars” Typical Test System with an Accuracy of (0.001)

Bus Active power Reactive power Voltage Voltage No. of

No. mismatch(p.u) mismatch(p.u) magnitude(p.u) angle(deg.) generations
1 slack slack 1.06 0.00 ‫ـــــ‬
2 0.00018139 0.00076184 1.0333 -0.2856 37
3 0.00094649 0.00028622 1.0349 -2.5267 27
4 0.00096724 0.0004207 1.0014 -1.4736 26
5 0.00091083 0.00096267 0.9786 -1.5042 35
Total Computational Time: 0.516 sec.

Figures below, the dotted curves show the evolution process of the active and reactive power
mismatches (cost function) versus the number of generations and the solid curves show the mean
cost of the individuals versus the number of generations at each busbar of the “5 busbars” typical
test system with an accuracy of (0.001). In the figures and due to the stochastic nature of the genetic
algorithm, we note that each busbar requires different number of generations to reach the required
accuracy (0.001 p.u.).

Evolution Process at Busbar 2 Evolution Process at Busbar 3

1441
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

Evolution Process at Busbar 4 Evolution Process at Busbar 5

In the table (2) shown, the active and reactive line flows (power flows) for the “5 busbars” typical
test system which consists of (7) lines are calculated, the calculation is done for an accuracy of the
active and reactive power mismatches of (0.001).

Table (2) Active and Reactive Power Flows for the “5 busbars” Typical Test System with an
Accuracy of (0.001)

Line Line Active power Reactive power

No. Term flow (MW) flow (MVAR)
1 1─2 22.3473 -39.7436
2 1─3 21.5944 -4.3321
3 2─3 20.7693 7.3872
4 2─4 16.2575 -13.0168
5 2─5 17.1768 -63.7962
6 3─4 -22.2970 -123.5797
7 4─5 -3.0479 -21.5199

* Iraqi National Grid Load Flow Solution Results:

The 362-bus (ING) consists of 30-generator bus and 332-load bus with 599 branches [Al-Bakri
1994]. Table (3) illustrates the load flow solution for Iraqi National Grid using the proposed
continuous (real-coded) Genetic Algorithm with a power tolerance (minimum active and reactive
power mismatches) of 0.001 p.u. with base quantities of 132 KV and 100 MVA. Also, the number of
iterations (generations) for each bus are tabulated. Due to the huge table, only the load flow solution
results of 96 bus are tabulated in this paper.

1444
 Number 4 Volume 15 December 2009 Journal of Engineering

Table (3) Load Flow Solution for “IRAQI NATIONAL GRID” with an Accuracy of (0.001)
Reactive
Bus Active power Voltage Voltage No. of
power
mismatch
No. mismatch (p.u) magnitude (p.u) angle (p.u) generations
(p.u)
1 slack slack 1.04 0 ‫ـــــ‬
2 0.0005 PV 1 18.3805 262
3 0.00021334 PV 1 2.8233 57
4 0.0008081 PV 1 -9.5400 3191
5 0.00011245 PV 1 13.6445 521
6 0.00043106 PV 1 -11.8520 34
7 0.0018487 PV 1 4.1875 5000
8 0.00066843 PV 1 7.5529 244
9 0.00023882 PV 1 12.3150 30
10 0.00016648 PV 1 4.0006 134
11 0.0003391 PV 1 -19.7704 884
12 0.00045458 PV 1 -6.3530 266
13 0.00013682 PV 1 4.5221 424
14 0.00058912 PV 1 -6.9794 76
15 0.00054176 PV 1 -8.1968 353
16 0.00021063 PV 1 13.5898 42
17 0.00078201 PV 1 4.5766 39
-6
18 2.4477*10 PV 1 11.1094 415
19 0.00090163 PV 1 7.0672 47
20 0.00089409 PV 1 -7.0275 9
21 0.00037127 PV 1 -3.2876 159
22 0.00014522 PV 1 -10.7986 24
23 0.00093387 PV 1 2.0421 216
24 0.00084462 PV 1 9.0268 472
25 0.00038532 PV 1 2.9669 17
26 0.00023586 PV 1 3.8338 527
27 7.2047*10-6 PV 1 -6.8666 88
28 0.00011686 PV 1 0.0252 50
29 0.00026843 PV 1 7.3612 333
30 0.0005791 PV 1 9.1833 1342

1444
H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

Reactive
Bus Active power Voltage Voltage No. of
power
mismatch
No. mismatch (p.u) magnitude (p.u) angle (p.u) generations
(p.u)
31 0.0003 0.0002 1.02247 12.3782 36
32 0.00031048 0.00013264 1.03356 -12.5347 57
33 0.00076127 0.00088425 0.915482 -18.8922 747
34 0.00048097 7*10-5 0.981021 -3.14901 60
35 0.00082233 0.00019414 0.909731 -7.5778 927
36 0.0007 0.0004 0.999866 0.40992 54
37 0.0009483 0.00059979 0.99142 -0.108361 3
-5
38 0.00072258 2.708*10 1.02766 2.41241 393
39 0.00010822 3.1427*10-5 0.911085 8.90456 21
40 7.3767*10-5 8.2306*10-5 0.978972 7.68687 21
41 3.9217*10-5 0.00014068 0.914069 -2.09315 30
42 0.00056231 0.00034232 0.931974 10.2677 117
43 0.00015186 0.00084486 0.916778 9.06423 69
44 0.00093275 0.00060546 0.935108 -13.2829 18
45 0.00085845 3.9338*10-5 1.00216 -8.97243 12
46 2.0578*10-5 0.00022034 0.935224 -7.08481 54
47 0.0005849 0.00025364 0.96548 -8.2154 120
48 0.0006 0.0009 0.96959 0.0359168 9
49 0.00046407 0.00068482 1.01733 8.11071 36
50 0.00064014 0.00040881 0.906882 16.9887 3
51 0.00081139 0.00073374 1.01725 12.1607 15
52 0.00063582 0.00036833 0.999238 0.015595 12
53 0.00064585 0.00073222 0.924709 7.0377 54
54 0.00039461 0.00031144 1.02049 4.97693 54
55 0.00050197 0.00051873 0.912678 18.5084 312
56 0.0006757 0.00066047 1.02961 13.4216 57
57 0.00048624 0.00046999 1.01989 13.0202 165
58 0.00026657 0.00078447 1.03154 -4.56727 54
59 0.00020212 0.00087574 0.998382 -0.133244 135
60 0.00099704 0.00040061 0.969316 -9.13924 336
61 0.0002 0.0005 1.09633 -3.3891 60

1444
 Number 4 Volume 15 December 2009 Journal of Engineering

62 0.00062866 0.00048481 1.01166 3.0274 69

-6
63 3.4536*10 0.00050729 1.00118 11.0695 144

Reactive
Bus Active power Voltage Voltage No. of
power
mismatch
No. mismatch (p.u) magnitude (p.u) angle (p.u) generations
(p.u)
64 0.00052304 6.1774*10-5 0.926323 0.126932 30
65 0.00039148 0.00027503 0.959269 -3.93418 234
66 0.0003338 0.00048038 0.939048 5.45821 24
67 0.00045369 0.00074015 1.03691 8.05995 12
68 0.00077681 0.00053799 0.962027 -0.556099 186
69 0.00027057 0.00061177 1.0712 5.11859 144
70 0.00040663 0.00054506 0.909477 -13.4194 18
71 0.00015602 0.00084029 1.0151 4.6136 60
-5
72 0.00085314 4.4098*10 0.9505 -9.4481 54
73 0.00082757 0.00056498 0.9587 14.8084 132
74 6.6466*10-5 0.00028761 1.0205 -0.5232 24
75 0.00045213 0.00089712 0.9106 16.254 450
76 0.00026312 0.00099152 1.0235 -0.4562 60
77 0.0003 0.0002 0.9989 -15.9269 1137
78 0.00060173 0.00081968 0.9069 9.3835 102
79 0.00097288 0.00093161 0.9115 -9.6373 399
80 0.00030775 0.00026214 1.0172 -11.3539 567
81 0.00028197 0.0003666 0.9567 10.958 72
82 0.0001 0.0007 0.9009 -9.6128 1215
83 0.00017501 0.00031678 0.9538 0.5662 36
84 0.00060463 0.00055344 0.9389 12.4227 51
85 0.00049479 0.00025939 0.9917 -17.9393 69
-5
86 0.00074739 6.1917*10 0.9976 11.2733 54
87 0.00073692 0.00062359 1.0459 3.6541 90
88 0.0002 0 0.9265 7.1323 69
-5
89 8.0049*10 0.00089968 1.0154 -0.9266 33
90 0.00051316 0.00093613 1.0007 -15.3443 189
91 3.1397*10-5 0.00072166 1.0017 18.7953 138
92 0.00062835 0.00047428 0.9977 12.0889 27

1444
 H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

93 4.7755*10-5 0.00047299 1.0034 11.0409 60

94 0.00031952 0.00085516 0.9855 1.9537 135
95 0.00028333 0.00050642 1.0368 18.9362 60
96 0.0008504 0.00096063 1.0821 -15.1242 96

- CONCLUSION
The proposed method (Real-Coded Genetic Algorithm) presented in this paper can be
implemented on-line for small and medium-scale power systems load flow solution and it can be
used for planning study for large-scale systems. The proposed method has reliable convergence and
high accuracy of solution. Whereas the traditional numerical techniques (Gauss-Seidel, Newton-
Raphson, Fast decoupled,…etc.) use the characteristics of the problem to determine the next
sampling point (e.g. gradient, linearity and continuity), genetic algorithm makes no such
assumptions. Instead, the next sampled point is determined based on stochastic sampling or decision
rules rather than on a set of deterministic decision rules. Also, whereas the traditional numerical
techniques mentioned above use a single point at a time to search the problem space, genetic
algorithm uses a population of candidate solutions for solving the problem. Thus, reducing the
possibility of ending at a local minima. Although
binary-coded genetic algorithm has been successfully applied to a wide range of optimization
problems, they suffer from disadvantage when applied to the real-world problems involving a large
number of variables. Thus, we use in our problem the real-coded genetic algorithm, where all
decision variables (unknowns) are expressed as real numbers. Explicit conversion to binary does not
take place. A reduction of computational effort is an obvious advantage of a real-coded genetic
algorithm. Another advantage is that, an absolute precision is now attainable by making it possible to
overcome the crucial decision of how many bits are needed to represent potential solutions. Blending
models are used in the crossover operator to remedy the problem of the crossover in the real-coded
genetic algorithm which is, no new information is introduced: each continuous value that was
randomly initiated in the initial population is propagated to the next generation, only in different
combinations. Thus, the blending methods combine variable values from the two parents into new
variable values in the offspring. At the same time, these methods do not allow offspring variables
outside the bounds set by the parent unless β > 1, where β is a random number on the interval [0,1].
Finally, solving the load flow problems by genetic algorithm gives high accurate results with respect
to the conventional methods, since load flow study is multiple solutions.

1444
 Number 4 Volume 15 December 2009 Journal of Engineering

REFRENCES

 Abdul-Haleem G. F., 2005, “A Genetic Algorithm for Manufacturing Cell Formation”,

M.Sc Thesis, Mechanical Department, University of Baghdad.

 Al-Shakarchi M. R. G., 1973, “Nodal Iterative Load Flow”, A dissertation submitted to

the Victoria University of Manchester.

 AL-BAKRI A. A., 1994, “A Study of Some Problems on the Iraqi National Grid and
Establishing a Method Algorithm for Load Flow”, M.Sc. Thesis, University of
Baghdad.

 Grisby Leonard L., 2007, “Power Systems”, CRC Press.

 Haupt R. L. and Haupt S. E., 2004, “Practical Genetic Algorithms”, A John Wiley &
Sons, INC., Publication, 2nd edition.

 Holland J., 1975, “Adaptation in Natural and Artificial Systems”, MIT Press.

 Ibrahim S. B. M., 2005, “The PID Controller Design Using Genetic Algorithm”, A
dissertation submitted to University of Southern Queensland, Faculty of engineering
and surveying, Electrical and Electronics Engineering.

 Kubba H.A., 1987, “Comparative Study of Different Load Flow Solution Methods”,
M.Sc Thesis, University of Baghdad.

 Michalewicz Z., 1996, “Genetic Algorithms + Data Structure = Evolution Programs”,

AI series, Springer-Verlag, New York, 3rd edition.

 Nanda J., Kothari D. P. and Srivastava S. C., 1987, “Some Important Observations on
Fast Decoupled Load Flow Algorithm”, Proceedings of the IEEE, VOL. 75, No.
5, 732-733

 Pohlheim H., 2005, “GEATBx: Genetic and Evolutionary Algorithm

Toolbox for Use with Matlab”, available at
http://www.geatbx.com/.

 Stott B. and Alsac O., 1974, “Fast Decoupled Load Flow”, IEEE Trans.
Power App. Syst., VOL. PAS-93, 859-869

 STAGG G. W. and AL-ABIAD A., 1968, “Computer Methods in Power

System Analysis”, Mc-Graw Hill Publishing Company.

 Talib A., 2007, “An Optimization Approach of Robot Motion Planning

Using Genetic Algorithm”, M.Sc Thesis, Mechatronics Department, AL-
Khwarizmi Engineering, University of Baghdad.

1444
H.A. Kubba Genetic Algorithm Based Load Flow Solution

S.S. Mahmood Problem In Electrical Power Systems

 Taylor D. G. and Treece J. A., 1967, “Load Flow Analysis by Gauss-Seidel Method”,
presented at the Symp. on power systems load flow analysis, University of Manchester
Institute of Science and Technology, Manchester, U.K.

 Woon L. C., 2004, “Genetic Algorithm for Load Flow Solution Techniques”, Abstract
of thesis, Master of engineering (electrical), http://www.sps.utm.

 Younes M. and Rahli M., 2006, “On the Choice Genetic Parameters with Taguchi
Method Applied in Economic Power Dispatch”, Leonardo journal of sciences, issue 9,
pp. 9-24.

 Zamanan N., Sykulski J., Al-Othman A. K., 2006, “A Digital Technique for Online
Identification and Tracking of Power System Harmonics Based on Real Coded Genetic
Algorithm”, Proceedings of the sixth IASTED international conference, European
power and energy systems, Rhodes, Greece.

LIST OF SYMBOLS AND ABBREVIATIONS

b : Any variable in the chromosome.

B : Imaginary part of admittance.
c : Crossover point.
e : Real part of bus voltage.
E : Complex bus voltage.
f : Imaginary part of bus voltage.
G : Real part of admittance.
GA : Genetic Algorithm.
LF : Load Flow.
n : Number of busses in the power system.
N : Number of objective functions.
Nvar : Number of variables in an individual.
Nkeep : Number of chromosomes that are kept each generation.
Nind : Number of individuals in the population.
p : Iteration index.
p.u : Per unit.
PQ : Load busbars.
PV : Generator busbars.
sp : Specified value.
V : Bus voltage magnitude.
w : Weighting factor.
Xrate : Selection rate.
Ykk : Self admittance of bus (k).
Ykm : Branch admittance between busses (k) and (m).
ΔP : Active power mismatch.
ΔQ : Reactive power mismatch.
δ : Bus voltage phase angle.
β : Random number on the interval [0,1].

1444
 Number 4 Volume 15 December 2009 Journal of Engineering

APPENDIX I

Nodal admittance matrix elements for “5 busbars” typical test system, per-unit quantity = 100
MVA, and 132 KV. The following data are all in (p.u), and buses data.

From Bus To Bus G (p.u) B (p.u)

1 1 6.25 -18.695
1 2 -5 15
1 3 -1.25 3.75
1 4 0 0
1 5 0 0
2 2 10.83334 -23.415
2 3 -1.66667 5
2 4 -1.66667 5
2 5 -2.5 7.5
3 3 12.91667 -38.695
3 4 -10 30
3 5 0 0
4 4 12.91667 -38.695
4 5 -1.25 3.75
5 5 3.75 -11.21

Bus Specified active Specified reactive Voltage

No. power (p.u) power (p.u) (p.u)
1 slack slack 1.06+j0
2 0.2 0.2 ‫ـــــ‬
3 -0.45 -0.15 ‫ـــــ‬
4 -0.4 -0.05 ‫ـــــ‬
5 -0.6 -0.1 ‫ـــــ‬

1444