International Journal of Engineering Trends and Applications (IJETA) – Volume 4 Issue 2, Mar-Apr 2017

RESEARCH ARTICLE OPEN ACCESS

Energy Optimization using Particle Swarm Optimization based

on Clustering for Wireless Sensor Network
O. Deepa [1], Dr. J. Suguna [2]
Research Scholar [1], Research& Development Centre,
Bharathiar University, Coimbatore, TN, India.
Associate Professor [2], Department of Computer Science,
Vellalar College for Women, Erode, Tamil Nadu, India

ABSTRACT
Wireless Sensor Network (WSN) is a self-organizing network which formed with huge sensors which are located in an
application specific environment to monitor the physical pheromone like temperature, fire, and pressure. Mostly sensors are
equipped with battery power through which they can perform sufficient operations and communication among neighboring
nodes. Researchers for WSNs face challenge that arises from limited energy. Maximizing the lifetime of the WSN, energy
conservation measures are essential for improving the performance of WSNs. Particle Swarm Optimization (PSO) is used to
form clustering and select cluster head with respect to minimizing the power consumption in WSN. This paper discusses about
energy optimization using PSO based on clustering algorithms for WSN, which aims to extend network lifetime.
Keywords :– Wireless Sensor Network, Particle Swarm Optimization, clustering, power consumption, fitness function

I. INTRODUCTION which increases the lifetime of the sensor network by

In recent years, with the rapid development of wireless grouping a number of nodes into clusters. Then a high residual
technology, sensors are receiving important attention, which energy node is selected for each cluster known as cluster head
leads to the emergence of directional sensor network. to collect the data from its members and transmit to the sink
Wireless Sensor Network (WSN) is a self-organized network with a minimum cost of multi-hop transmission. Extended
with huge, intelligent and small sensors. These sensors can network lifetime, reliable data transfer, energy conservation in
perform the data transmission among themselves within their sensors and scalability are the main necessities for WSN
radio range and also they are organized in a way to sense, applications [5]. Because of the several limitations in the
observe and recognize the physical entity of the real world [1]. sensors, WSN is having various issues such as coverage area,
WSNs integrate into the environment, machinery and human, network lifetime and dynamic topology and data aggregation.
coupled with the efficient delivery of sensed information, To enhance the network lifetime appropriately many
could deliver tremendous benefits to society. Some of the routing protocols and cluster-based algorithms are used to
potential benefits are reinforced emergency response, fulfill the application requirements in WSN. From existing
preservation of natural resources, improved homeland security research methods, optimizing energy dissipation for
and enhanced manufacturing productivity. The significance communication becomes very critical. For maximizing
of sensor networks have low energy consumption, sufficient lifetime of the WSN, part of an energy consumption of each
intelligence for signal processing, low cost, self-organizing sensor has an important role while communicating among
capability, data gathering and querying ability [2]. Wireless other sensors. This paper focuses on energy conservation in
sensor is equipped with a radio transceiver, a microcontroller, each sensor by using PSO-based clustering algorithm. The
an interfacing electronic circuit and an energy source usually a cluster head is selected using PSO, based on the distance from
battery. The features of these sensors include tiny, low cost, the cluster member node to sink and the residual energy in
low computation power, multifunctional and easy that node.
communication within short distances [3]. WSN consist of
intelligent sensors, which has been arranged and installed The rest of this paper is organized as follows: optimization
based on the applications and a sink that is located very near methods, PSO and its relative advantages are briefly outlined
to or within the radio range. The sink transmits the queries to in Section 2.Cluster formation using PSO’s are discussed in
the neighboring sensors which perform the sensing task and Section 3. Energy-aware clustering optimization algorithms
return the data to the sink as a response to the transmitted for WSN based on PSO are discussed in Section 4.
query [4]. In WSN nodes utilized is balanced amount of Conclusion and scope for future work is given in Section 5.
energy for communication and the required energy in terms of
battery power to transmit the packet will differ among the II. PSO: A BRIEF OVERVIEW
transmissions with respect to the distance between the sender A. Optimization Methods
and receiver nodes; therefore multi-hop communication is For example to consider the following global optimization
recommended. Data transmission using hierarchical routing problem

ISSN: 2393-9516 www.ijetajournal.org Page 65

 International Journal of Engineering Trends and Applications (IJETA) – Volume 4 Issue 2, Mar-Apr 2017
by a particular particle, then the best value is chosen among
the generated population and that particular best value is the
best solution and it is known as gBest. The PSO always try to
(1) change the velocity of every particle towards its pBest and
where is a continuous variable vector with domain lBest. The velocity is determined by random terminologies,
defined by the bound constraint . which is having randomly generated numbers for velocity
The function is a continuous real-valued function. towards pBest and lBest localities. From the large deposit of
Many real-world problems, such as engineering and related generated solutions, the best one is selected to resolve the
areas, can be reduced to formulation (1). This problem problem. The PSO algorithm always stores and maintains a
usually has many local optima, so it is difficult to find its record of results for three global variables such as target value
global optimum. For solving such problem, researchers have or condition, gBest and stopping value. Every obtained
presented many methods during the past years, which can be particle of PSO contains: a data which can represent a global
divided into two groups: deterministic and stochastic solution, value for velocity which will indicate the amount of
algorithms. Most deterministic algorithms usually effective data to be changed, lBest value [8].
for unique modal functions have one global optimum and In the past several years, PSO has been successfully
need gradient information. However, stochastic algorithms do applied in much research andcan be used across a wide range
not require any properties of the objective function. Therefore, of applications, as well as for specific applications focused on
more attention has been paid to stochastic algorithms recently a specific requirement. Advantages of PSO over these
and many effective algorithms have been presented, including alternatives are ease of implementation on hardware or
Simulated Annealing (SA), Genetic Algorithm (GA), software, availability of guidelines for choosing its parameters,
Differential Evolution (DE), Particle Swarm Optimization high-quality solutions because of its ability to escape from
(PSO), Ant Colony Optimization (ACO), Artificial Bee local optima, availability of variants for integer, real and
Colony (ABC), and Harmony Search (HS). Among these binary domains and quick convergence.
stochastic algorithms, PSO is a population based and
intelligent method, which is inspired by the emergent motion C. PSO Algorithm
of a flock of birds searching for food [6]. In PSO, a PSO is a popular multidimensional optimization
population of potential solutions is evolved through technique [9], which explore an -dimensional hyperspace in
successive iterations. Since PSO algorithm has a number of search of the global solution, where represents the number
desirable properties, including simplicity of implementation, of optimal parameters to be determined. A particle occupies
scalability in dimension and good empirical performance, it position and velocity in the dimension of the
has been applied to solve many real-world problems. hyperspace, and .Each particle is
Although PSO algorithm has been applied successfully in evaluated through an objective function ,
solving many difficult optimization problems, it also has where . The cost (fitness) of a particle close to the
difficulties in keeping balance between exploration and global solution is lower (higher) than that of a particle that is
exploitation when solving complex multimodal problems. farther. PSO thrives to minimize (maximize) the cost (fitness)
B. Introduction to Particle Swarm Optimization function. In the global-best version of PSO, the position where
PSO is an optimization technique in which natural species the particle has its lowest cost is stored as .
social behaviors are considered for the purpose of Besides, , the position of the best particle. In each
computation, introduced by Kennedy and Eberhert in 1995 [7]. iteration , velocity and position are updated using (2)
It is a swarm intelligence technique which isa population- and (3). The update process is iteratively repeated until either
based optimization scheme with the objective of optimizing a an acceptable is achieved or a fixed number of
fitness function by updating the generations. The model of this iterations is reached.
algorithm is based on the social behavior of bird flocking. It
is observed from the nature of birds which always travel in a
group without colliding. The PSO technique uses several
particles, each represents a solution and finds the best particle
position with respect to a given fitness function. In PSO,
each single solution is a particle (bird) in the search space. (3)
Each particle have fitness values that are evaluated by the
fitness function to be optimized and stored record for all its where and are constants and and are
coordinates which are related to obtaining the better solution random numbers uniformly distributed in [0, 1]. The general
by following the current best particles. The fitness value of the pseudo code of PSO is shown in Algorithm I
current optimum particle is called pBest. PSO optimizes the
best population value that is obtained so far by any particle in
the neighbors and its location is called lBest. When all the
generated populations are considered as topological neighbors

ISSN: 2393-9516 www.ijetajournal.org Page 66

 International Journal of Engineering Trends and Applications (IJETA) – Volume 4 Issue 2, Mar-Apr 2017

PROCEDURE PSO( )

BEGIN Start
Start

T=0;

INITIALIZE PARTICLES P(T);

Initialize particles

EVALUATE PARTICLES P(T);

Calculate the fitness
WHILE (TERMINATION CONDITIONS ARE
value for each particle
UNSATISFIED)

BEGIN

T=T+1;
If fitness is
UPDATE WEIGHTS; Yes No
better than
SELECT PBEST FOR EACH PARTICLE; pBest?

SELECT GBEST FROM P(T-1);

Assign current Maintain previous
CALCULATE PARTICLE VELOCITY P(T);
fitness as pBest pBest value
UPDATE PARTICLE POSITION P(T);

EVALUATE PARTICLES P(T);

END Assign best among

pBest as gBest
END

Algorithm I – Procedure of PSO

Find velocity of each
III. CLUSTERING USING PSO particle
Clustering is an NP-hard optimization problem and is used
for several effective energy protocols in WSN. The common
operation of cluster-based protocol can be understood with the With each particle’s
help of the following phases. In cluster formation phase, velocity update its data
Cluster Head (CH) election takes place in a random manner.
Once the CH election is performed, each CH broadcast an
advertisement message toall the sensors, which arein the
transmission range of a particular CH receives the message. No
Whether maxi
The sensors or cluster nodes which then send a reply message
with request message to join under the particular CH. In value attained?
information processing phase, each sensor involves in sensing
the attributes depending on the application. The sensed data Yes
from each node is transmitted to their respective CH. Once
CHs receive the sensed data, it performs data aggregation and Stop
this process is repeated periodically. In data dissemination Start
phase, each CH transmits the aggregated data to the sink Fig.1. Flowchart of cluster formation using PSO
based on the data dissemination interval.
Particles are created by considering two parameters namely,
PSO is a popular choice for WSN clustering because its position and velocity. Once the sensors are deployed in the
optimization algorithm is simpler and the network efficiency sensor field, the sink broadcasts a request message to all the
is better. By using PSO for clustering, let us consider a sensors in the network to collect the node’s information. This
sample space and deploy the particles or sensors in it. Let be in turn sends reply message to the sink which contains:
the number of particles considered in the sample space. position, velocity and energy of the sensor. These values are
maintained and updated at the sink. Then the sink makes the
sensors to perform clustering. Cluster formation is performed
using PSO by allowing each sensor to find the nearest
neighbor within its radio range. Likewise all the sensors in

ISSN: 2393-9516 www.ijetajournal.org Page 67

 International Journal of Engineering Trends and Applications (IJETA) – Volume 4 Issue 2, Mar-Apr 2017
WSN are allowed to form clustering, it become a part of any
of the clustering. The same process is repeated until all the (4)
sensors become a member of any cluster in the network. The
fitness value is calculated for choosing a cluster particle
depends on the following factors namely, energy of the In PSO-TVIW, the inertia weight is decreased linearly
particle or sensor, energy of particles or sensors with in a in each iteration. In PSO-TVAC, inertia weight is set constant
radio range from a particular particle and distance of those and acceleration constants and are varied linearly in
particles within the radio range from a particular particle. every iteration. In HPSOTVAC, the particle update is not
Cluster particle is one in which more number of nodes are influenced by the velocity in previous iteration; but,
accessible in its radio range. The cluster particle can be reinitialization of velocity is done when the velocity stagnates
generally referred as cluster head. Then, fitness function of in the search space. Finally, in PSO-SSM,
each particle is calculated for each iteration [10]. Maximum
fitness value in each iteration is called as lBest and maximum (5)
value among all iteration is called as gBest. If suppose, gBest
value is obtained in the ith iteration, fitness values of all
particles in that particular iteration are taken into account for
where is a constant called momentum factor. Clustering is
cluster formation. Node with the maximum fitness value is
based on a simple idea that for a group of nodes that lie in a
taken as referenceand constructing the cluster by making the
neighborhood, the node closest to the sink becomes the cluster
nodes in its radio range as its cluster members. The value of
head. A detailed comparative analysis of the algorithms for
gBest is broadcasted to each cluster head so that, each head
optimal clustering is presented. This scheme considers only
may aware of the gBest node. With reference to the node id
the physical distances between nodes and their assigned
the information is being transmitted. This process is
cluster heads, but not the energy available to the nodes
illustrated in Fig. 1.
E. PSO-C
IV. ENERGY OPTIMIZATION USING PSO Latiff et al. consider both residual energy and physical
BASED ON CLUSTERING ALGORITHMS distances between the nodes and their cluster heads [13]. Each
FOR WSN particle represents a combination of cluster heads. The fitness
The most important issue in this type of networks is function for the centralized PSO (PSO-C) is defined as
energy constraints. In WSN, energy efficient routing protocols C where is the maximum average
should be designed in an efficient manner in order to improve Euclidean distance of nodes to their associated cluster heads
the network lifetime. The network lifetime is measured by the and is the ratio of total initial energy of all nodes to the total
effective usage of the sensor in the network. Sensor in the energy of the cluster-head candidates in current round. These
sensing region is used to perform sensing, processing and are expressed as (5) and (6), respectively.
communication. The overall network lifetime is based on the
above said factors. The network lifetime can be improved by
(5)
avoiding the sensor to transmit raw data. This can be achieved
by aggregating the sensed data to eliminate the data
redundancies, eliminating the control overhead messages and
(6)
avoiding the long distance transmission. If a network is
constructed by considering the above said factors, the overall
network lifetime can be improved [11]. In this section, where N is the number of nodes out of which K will be
various energy optimization using PSO based on clustering elected as the cluster heads. |Cp,k| is the number of nodes that
algorithms for WSN was discussed, which aims to extend belong to cluster Ck in particle p. This ensures that only the
network lifetime. nodes that have above-average energy resources are elected as
the cluster heads, and that the average distance between the
D. PSO-Clustering nodes and the cluster heads is minimum. They compare the
Guru et al. have proposed four variants of PSO, namely, results of the algorithm with those of LEACH and the
PSO with time varying inertia weight (PSO-TVIW), PSO with LEACH-C algorithms [14]. The PSO-based clustering
time varying acceleration constants (PSO-TVAC), algorithms outperform GA, LEACH and LEACH-C in terms
hierarchical PSO-TVAC (HPSO-TVAC) and PSO with of the network lifetime and the throughput.
supervisor student mode (PSO-SSM) for energy-aware
clustering[12]. PSO assigns nodes to each of the cluster- F. MST-PSO
Cao et al. have considered an interesting case in which a
heads, such that the total energy loss due to
node and its cluster-head engage in a multi-hop
physical distances is minimum. This is defined in (4),
communication [15]. The method computes a distance-based
where is the distance between cluster-head and the sink minimum spanning tree of the weighted graph of the WSN.
The best route between a node and its cluster-head is searched

ISSN: 2393-9516 www.ijetajournal.org Page 68

 International Journal of Engineering Trends and Applications (IJETA) – Volume 4 Issue 2, Mar-Apr 2017
from all the optimal trees on the criterion of energy [2] Romer K. and Mattern F., “The design space of Wireless
consumption. Cluster heads are elected based on the energy Sensor Networks”, Proc. IEEE Wireless Communications,
available to the nodes and the Euclidean distance to its vol.11(6), pp.54-61, 2004.
neighbor node in the optimal tree. The authors compare the
[3] K. Akkaya and M. Younia, “A Survey on Routing
performances of three mechanisms of cluster-head election
based on energy, auto-rotation and probability. Routing and Protocols for Wireless Sensor Networks”, Elsevier Ad
cluster-head rotation are treated as optimization problems and Hoc Network Journal, Vol. 3 (3), pp.325-349, 2005.
tackled through PSO. The results show that the PSO-based [4] Heinzelman W. R., Chandrakasan A., and Balakrishna H.,
clustering methods ensure longer network life. “Energy efficient communication protocol for wireless
G. PSO-MV microsensor networks,” in Proc. 33rd Annu. Hawaii Int.
PSO-MV protocol [16] is based on PSO method and the Conf. Syst. Sci., Jan. pp. 1–10, 2000.
residual energy of cluster heads is higher than other nodes, the [5] Al-Karaki J. N.,and Kamal A. E., “Routing techniques in
purpose of the approach is energy balance. In the PSO-MV Wireless Sensor Networks: a survey”, IEEE Wireless
method, to choose the best two nodes as cluster heads, namely Communications, vol.11(6), pp. 6–27, 2004.
Master Cluster Head (MCH) and other as Vice Cluster Head [6] Kennedy J., and Eberhart R., “Particle swarm
(VCH) and the tasks between the nodes can be categorized,
optimization”, in Proc. IEEE Int. Conf. Neural Netw., vol.
where the MCH is responsible for data collecting and
transmission and VCH is responsible for inter-cluster 4, pp. 1942–1948, 1995.
communications or intra –cluster communications to sink. [7] Del Valle Y., Venayagamoorthy G. K., Mohagheghi S.,
PSO-MV clustering algorithm is based on routing of clusters Hernandez J. C., and Harley R., “Particle swarm
included generation steps and transmission of data according optimization: Basic concepts, variants and applications in
to the set of nodes into clusters and threshold energy that power systems”, IEEETrans.Evol.Comput.,vol.12(2), pp.
shows to is chosen as follow 171–195, Apr. 2008.
[8] Hassan R., Cohanim B., and De Weck O., “A comparison
But the weakness of this algorithm is in selection the number of particle swarm optimization and the genetic
of cluster heads optimum. algorithm,” presented at the
AIAA/ASME/ASCE/AHS/ASC 46th Struct. Dyn.Mater.
V. CONCLUSIONS Conf., Austin, TX, Apr. 2005.
Scale and density of deployment, environmental [9] N. M. A. Latiff, C. C. Tsimenidis, and B. S. Sharif,
uncertainties, and constraints in energy, memory, bandwidth,
“Performance comparison of optimization algorithms for
and computing resources pose serious challenges to the
developers of WSNs. Issues of the node deployment, clustering in wireless sensor networks,” in Proc. IEEE Int.
localization, energy-aware clustering and data aggregation are Conf. Mobile Ad Hoc Sens. Syst., pp. 1–4, Oct. 2007.
often formulated as optimization problems. Most analytical [10] Kulkarni R. V., and Venayagamoorthy G. K., “Particle
methods suffer from lack of convergence to the final solutions. swarm optimization in wireless-sensor networks: A brief
PSO has been a popular technique used to solve optimization survey,” IEEE Trans. Syst., Man, Cybern. C, Appl. Rev.,
problems in WSNs due to its simplicity, high quality of vol. 41(2), Mar. 2011, pp. 262–267.
solution, fast convergence and reduced computational burden.
In this paper, an overview of optimization algorithms and a [11] Amin Rostamiand Mohammad Hossin Mottar, "Wireless
brief survey of recent PSO-based solutions to the WSN are Sensor Network Clustering using Particles Swarm
presented. The network performance of the WSNs is enhanced Optimization for reducing energy
by using PSObased on clustering algorithms in terms of consumption",International Journal of Managing
increasing the network lifetime, throughput, residual energy Information Technology (IJMIT), vol. 6(4), pp. 1-15, Nov.
and number of active nodes. From the current rate of growth 2014.
of PSO-based applications, it is envisioned that PSO will
[12] Vimalarani C., Subramanian R. and Sivanandam S.N.,
continue as an important optimization technique in the field of
WSNs. "An Enhanced PSO-based Clustering Energy
Optimization Algorithm for Wireless Sensor Network",
REFERENCE The Scientific World Journal, 2016, pp.1-11, 2016.
[1] Akyildiz I.F., Su W., SankarasubramaniamY., and [13] Guru S., Halgamuge S., and Fernando S., “Particle
CayirciE., “Wireless Sensor Networks, A Survey”, swarm optimizers for cluster formation in wireless sensor
Communication Magazine, IEEE, vol. 40(8), pp. 102-114, networks”, in Proc. Int. Conf. Intell. Sens.,
August 2002. Sens.Netw.Inf.Process., pp.319–324, 2005.
[14] Latiff N. M. A., Tsimenidis C. C., and Sharif B. S.,
“Energy-aware clustering for wireless sensor networks

ISSN: 2393-9516 www.ijetajournal.org Page 69

 International Journal of Engineering Trends and Applications (IJETA) – Volume 4 Issue 2, Mar-Apr 2017
using particle swarm optimization”, in Proc. 18th IEEE
Int. Symp. Pers., Indoor Mobile Radio Commun., pp. 1–5,
2007.
[15] Cao X., Zhang H., Shi J., and Cui G., “Cluster heads
election analysis for multi-hop wireless sensor networks
based on weighted graph and particle swarm
optimization”, inProc.4thInt.Conf.NaturalComput.,vol.7,
pp. 599–603, 2008.
[16] Yua H. and Xiaohuia W., “PSO-based Energy-balanced
Double Cluster-heads Clustering Routing for Wireless
Sensor Networks”, Procedia Engineering, vol 15, pp.
3073 – 3077, 2011.

ISSN: 2393-9516 www.ijetajournal.org Page 70