Telecommunication Systems

https://doi.org/10.1007/s11235-020-00659-9

Energy efficient protocol in wireless sensor network: optimized cluster

head selection model
Turki Ali Alghamdi1

© Springer Science+Business Media, LLC, part of Springer Nature 2020

Abstract
Energy efficiency has become a primary issue in wireless sensor networks (WSN). The sensor networks are powered by
battery and thus they turn out to be dead after a particular interval. Hence, enhancing the data dissipation in energy efficient
manner remains to be more challenging for increasing the life span of sensor devices. It has been already proved that the
clustering method could improve or enhance the life span of WSNs. In the clustering model, the selection of cluster head
(CH) in each cluster regards as the capable method for energy efficient routing, which minimizes the transmission delay in
WSN. However, the main problem dealt with the selection of optimal CH that makes the network service prompt. Till now,
more research works have been processing on solving this issue by considering different constraints. Under this scenario, this
paper attempts to develop a new clustering model with optimal cluster head selection by considering four major criteria like
energy, delay, distance, and security. Further, for selecting the optimal CHs, this paper proposes a new hybrid algorithm that
hybridizes the concept of dragon fly and firefly algorithm algorithms, termed fire fly replaced position update in dragonfly.
Finally, the performance of the proposed work is carried out by comparing with other conventional models in terms of number
of alive nodes, network energy, delay and risk probability.

Keywords Wireless sensor networks · Energy efficiency · Cluster head · Delay

1 Introduction cols to improve energy efficiency [5–7] and clustering [8–11]

for guaranteeing the proper network operation. Routing pro-
WSNs [1, 2] has a broad range of application over many tocols of WSN are of two main classes, namely hierarchical
fields. The sensor nodes are mostly battery-powered devices. routing, and flat routing. In the flat routing protocol, the entire
They are cheap and tiny; it employs for gathering helpful nodes have been deployed at a similar level and are consid-
information and transmit them via wireless links from the ered as simple. Further, the scalability is more reduced due to
physical area to sink nodes. These sensor node batteries the overhead packet control, and if the network size increases,
are unchangeable one. The charge ruining causes network then the energy consumption is also increased. Hierarchical
disconnection. The communication of these sensors is hap- routing [12] has split the areas on the network into clusters.
pening either within themselves or else straight to the sink. Every cluster is assigned to the cluster head node, where the
Hence the resource selection and communication need to data from the other cluster nodes are collected and trans-
optimize to enhance the lifetime of the network [3, 4]. fers them to Base station. This in turns extends the network
Besides, some issues are there in WSNs due to the restricted lifespan and balance efficiently the energy consumption in
computational resources and buffering characteristics. sensor nodes. On comparing the flat and hierarchical routing
The alternative ways concerned to extend network lifetime protocols, the hierarchical routing protocols pose superior
includes a selection of necessary constraints, routing proto- scalability and flexibility. Nowadays, the hierarchical rout-
ing protocols [13–15] are focused mainly on many of the
researchers.
B Turki Ali Alghamdi
Routing protocols are of two types, i.e., network structure
[email protected]
and protocol operation. The routing models were deployed
1 Department of Computer Science, College of Computer and for increased robustness, strong scalability and reduced data
Information Systems, Umm Al-Qura University, Mecca, retransmission. For enhancing the life span of the system,
Saudi Arabia

123
 T. A. Alghamdi

hierarchical routing protocols assist in network scalabil- Table 1 Table of abbreviation

ity, data collection and also minimizes the delay in data. Abbreviation Description
In numerous works, it is described that the routing mod-
els, especially the clustering systems, create a significant WSN Wireless sensor networks
development on WSNs. Hierarchical routing is exploited for CH Cluster head
minimizing the network performance and energy utilization CHS Cluster head selection
while the entire network sensor nodes are sending data to the DA Dragon fly
BS [16, 12]. The core elements of the cluster-based WSNs FF Firefly algorithm
are sensor nodes, CH’s, BS, and end-users. The controlling BS Base station
of every cluster is carried out by the CH that have direct FPU-DA FF replaced position update in DA
communication with the BS [17, 18]. However, the optimal MEACBM Mobile energy aware cluster-based multi-hop
CH selection process is still challenging. The frequent selec- ACO Ant colony optimization
tion of CH affects routing performance since the nodes are PSO Particle swarm optimization
busy in node selection rather than communication. The fac- NAN Number of alive nodes
tors affecting cluster formation and CH communication are NNE Normalized network energy
open issues [19, 20]. Hence it is considered as important in GECR Genetic algorithm-based energy-efficient clustering
the implementing model. and routing
The significant contribution of this paper is depicted GWO Grey wolf optimization
below. WOA Whale optimization algorithm
GA Genetic algorithm
1. This paper intends to present a new clustering model with
optimal CHS by considering four primary criteria like
energy, delay, distance, and security.
2. These node initializations are done by setting the con-
strains as zero. the CHs were selected. In this, only the sensor nodes were
3. Here, for selecting the optimal CHs, this work proposes chosen as CH that poses larger energy between other sensor
a new hybrid algorithm known as the FPU-DA model. nodes. Here, the hierarchical heterogeneous clustering was
4. After the optimal cluster head is selected, the nodes com- assumed with sensor nodes having three levels; sensor node
municate with each other using inter and intracluster connectivity inside the entire network area and multi-hoping
communication. The packet relay is held by considering for inter-cluster communication. Once the cluster formation
constraints like distance, energy, and security. and sensor node deployment were finished, this MEACBM
5. At last, the performance of the adopted model is carried has split the whole network area as sectors and within every
out by comparing over existing schemes in terms of a sector places the mobile sensor node that plays as MDC,
number of alive nodes, network energy, delay, and risk which gathers data from CHs. This approach considerably
probability. minimizes the sensor node’s energy consumption to trans-
fer the information over the Base station. The experimental
The abbreviations used in this work are given in Table 1. outcome has shown the betterment over the other models
The rest of the paper is set as: Sect. 2 presents the literature regarding the throughput, count of CHs, network lifetime
review of the benchmark protocols. Section 3 describes the and count of dead nodes.
proposed clustering model: energy and security awareness in In 2018, Kaur and Mahajan [22] have currently stated
the CHS process and Sect. 4 portrays the optimal selection Energy efficiency as the initial problem in WSNs. These
of secure and energy-aware CHS. Besides, Sect. 5 illustrates sensor nodes become lifeless over time as they were bat-
the results and discussions, and Sect. 6 concludes the paper. tery powered. Therefore, the enhancement in data dissipation
within energy efficiency was considered as the promising
issue in improving the sensor device’s lifespan. The network
2 Literature review lifespan of WSN has been enhanced using the tree-based and
clustering-based data aggregation methods for sensor net-
2.1 Related works works. This paper has implemented the tree-based routing
protocol and, Hybrid ACO and PSO based energy-efficient
In 2019, Toor and Jain [21] have proposed a narrative per- clustering model. At first, based on the residual energy,
ception for hierarchical heterogeneous WSNs with regards the clusters have been formed, and after that, the hybrid
to mobile sensor nodes named MEACBM routing protocol. ACOPSO termed data aggregation was developed for further
Further, based on this new implemented probability equation, enhancing the inter-cluster data aggregation. The prolonged

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

analysis has stated the considerable improvement in the net- efficient clustering for prolonging the life cycle of network
work lifespan among the proposed model. and enhancing energy efficiency. The optimal solution that
In 2018, Reem et al. [23] have examined an optimum node attained from the earlier network round has been added to
degree for minimum energy consumption. The node degree the prime population for the present round, by which the
of the Degree Constrained Tree (DCT) in homogeneous WSN search efficiency has been enhanced. Further, while model-
with only one BS was adopted for their study. The degree ing the fitness function, the load balancing factor was also
of the nodes, in turn, affects the lifetime of the network. considered that balances the energy consumption between
Consequently, the paper proposes a Collaborative Distributed the nodes. The simulation outcome has revealed that the pro-
Antenna (CDA) routing protocol in terms of transmission posed model has bettered in terms of load balancing having
energy to provide node distribution. The experimental results low variance and energy-efficient.
show that the optimal node degree doubles the lifetime of In 2019, Ennaciri et al. [28] have exploited a new load
the network. Further, DCT with CDA proves enhancement balancing protocol for managing the energy consumed by
in network stability. the sensor nodes in the WSN. The simulation was done
In 2017, Sharawi and Emary [24] adopted Grey wolf using MATLAB, the performance comparison was evaluated
optimization (GWO) to resolve the CH selection problem. using two protocols namely LEACH and SEP. Thus, finally
Suitable fitness function was employed to ensure coverage of it proved its superiority.
the WSN and is fed to the GWO to find its optimum. The pro- Problem statement In the literature, the cluster-based rout-
posed model results are compared with the LEACH routing ing has been reported with different contributions. They are
protocol. Four different like residual energy lifetime, network broadly categorized into, algorithm for cluster head selection
throughput and performance indicators are analysed. Finally, and constraints for cluster head selection. As per the catego-
the introduced model outperforms the LEACH in almost all rization, the constraints for cluster head selection process for
topologies using the different indicators. routing protocol have been reported in [21, 23, 26], whereas
In 2018, Jadhav and Shankar et al. [25] proposed an [22, 24, 25, 27] proposes metaheuristic optimization algo-
energy-efficient cluster head selection algorithm, which is rithm for cluster head selection process is named as set 1.
based on Whale Optimization Algorithm (WOA) called Since our research intends to develop an optimization algo-
WOA-Clustering (WOA-C) was proposed. Consequently, it rithm for cluster head selection process, we consider [22, 24,
helps in the selection of energy aware cluster heads based 25, 27] for further study and analysis. Moreover, our paper
on a fitness function that considers the residual energy of focuses on both the DA and FA concepts that take update
the node and the sum of energy of adjacent nodes. Finally, based on levy distribution, encircling and random solutions.
the implemented algorithm was evaluated in terms of energy It leads the algorithm to converge earlier than the traditional
efficiency, network lifetime, overall stability and throughput. algorithms. It is grouped to form set 2.
Moreover, the performance of WOA-C was assessed against
other standard contemporary routing protocols to show its 2.2 Reviews
superiority over other models.
In 2018, Yahiaoui et al. [26] have proposed delay and Table 2 shows the reviews on CHS in WSN. MEACBM [21]
energy-sensitive routing protocol to ensure enhanced quality has a superior performance by minimizing the consumption
of service. The main objective of the paper includes mini- of energy and raises the network lifetime, throughput and
mization of delay and energy consumption. The WSN and count of dead sensor nodes. However, needs simulation over
actuator network are considered. It composed of both sen- the real-time experiments and requires the consideration of
sor and actuator nodes. The actuators are responsible for scalability of sensor nodes within every cluster. ACOPSO
taking prompt decisions and react accordingly to the data [22] the algorithm selects a cluster head based on the hybrid
gathered by sensor nodes. These networks are organized in operation of PSO and ACO. ACO is a local search algorithm,
clusters, which are supervised using CH. The CH was elected while the cluster head selection problem is a multimodal
based on connectivity and energy capability. Moreover, the optimization problem. So, the probability of sticking with
latter metric ensures the distance among a number of hops the local optimal is high. DCT with CDA [23] has elongated
of sensor nodes relative to the actuator nodes. This metric network lifetime and network stability. However, the security
improves the network reliability by reducing the communi- condition is the major challenge of this method. In [24, 25],
cation delay when alerting the actuator nodes, and hence, Whale optimization and GWO take the updating principle
reduces the energy consumption. Finally, the simulation out- based on the encircling model. The model is not compet-
comes show sufficient gain in terms of communication delay itive enough to understand the cluster head characteristics
and energy consumption. because there is no swirling geography in the cluster head
In 2018, Tianshu et al. [27] have presented a rout- selection process. Delay and Energy sensitive routing proto-
ing approach GECR and genetic algorithm-based energy- col [26] has reduced the packet drop and energy consumption.

123
 T. A. Alghamdi

Table 2 Review on state-of-the-art of conventional energy efficiency routing in WSN

Protocols Features Challenges

MEACBM [21] Superior performance by minimizing the amount of Needs simulation over the real-time experiments
energy consumption of sensor node Requires the consideration of scalability of sensor nodes
Increases the network lifetime, throughput and count within every cluster
of dead sensor nodes
ACOPSO [22] Reduced packet size Critical to define the initial design parameters
Enhances the network lifetime Cannot define the issues of scattering
Conserves energy in effective manner
DCT with CDA [23] Elongates the network lifetime Security needs to be considered
Improves network stability
GWO [24] Better network lifetime and throughput Uncertain obtaining process
Simplified energy consumption Needs joint consideration of MAC and network layer
WOA-C [25] Extends the life span of network Network performance needs improvement
Improved scalability Cluster Head characteristics are not clearly defined
Delay and energy sensitive Improves the reliability Research never measured security and efficiency
routing protocol [26] Minimize the delay and energy consumption
GECR [27] Better network life cycle Future work focuses on verifying and applying the
Optimal energy utilization appropriate metaheuristic algorithms
Optimal selection is uncertain
Load balancing protocol Improved network stability, throughput, and packet Energy harvesting scheme needs to be considered
[28] drop rate Network lifetime further needs improvement
Significantly reduced energy consumption
RE-attempt [36] Better network lifetime and throughput Will focus on cross-layer design protocols
Simplified energy consumption Needs joint consideration of MAC and network layer
K-means algorithm [37] High throughput Hard to reach the optimal solution within a reasonable
Less energy consumption time
FCM [38] Maximises energy There is possibility to loss information
DBSCAN [39] It categorized cluster based on belong-to, not Less number of constraint are considered
belong-to and ambiguity
Reduce energy consumption
LEACH [40] It measure the level of life time of the sensor nodes Routing of packets is critical
Minimize intra-cluster distance

It still lacks in measuring energy efficiency and security. transmitting the sensed data to the sink node Bs . The pattern
GECR [27] genetic algorithm highly depends on random- of transmission of all sensor nodes depends on the uniform
ization and roulette wheel selection so that updating towards distribution, in which the data is distributed with utmost radio
the optimal point is uncertain. It poses better network life level with the dimensions of Mn and Mm in meters. The group
cycle, energy utilization, and load balancing. Load balanc- of sensor nodes turns into a cluster denoted by Cn .
ing protocol (EESAA) [28] has improved network stability, In each cluster, the term Dcn denotes the corresponding
throughput, and packet drop rate and further significantly CH. A new mechanism for transmitting the data from the
reduced energy consumption. Two major disadvantages are entire nodes to BS is known as CH-oriented routing. Here,
energy harvesting scheme needs to be considered and the the distance among BS and CHis given by d H B and distance
network lifetime further needs improvement. among i th normal node to jth CHis denotedas di j .The main
advantage of this model [29] is of structural data flow, and
lifespan improvement. However, it considered less number of
3 Proposed clustering model: energy constraints for optimal cluster head selection. To overcome
and security awareness in cluster head that issue, few more constraints are included in the proposed
selection process model (Fig. 1).

3.1 Network model 3.2 Security awareness

The WSN model is adopted from [29], of that comprises Here, the security mode, risky mode and γ -risky mode are
of m count of nodes. These sensor nodes are efficient of regarded as constraints that are defined below.

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

1, the selection process would be implemented, and if the

Cluster heads WSN Base station state is 1 < sd − sr ≤ 2, there will be a delay in the selection
Clustering
process. Still, the CHS process could not be completed, and
the corresponding function should be continued for the state
2 < sd − sr ≤ 5.

3.3 Energy awareness

Optimal Cluster Head Optimization using
FPU-DA model
Selection The adopted model implements the CHS model [29] by
employing the hybrid concepts of FF and DA, by consid-
ering or subjecting the constraints namely risk factor, delay,
energy, and the distance. Generally, the distance amongst
the CH and the normal node, risk factor and delay have to
Energy Delay be minimized while the energy of the CH has to be high
for attaining the improved CHS approach. Thus, depending
Distance Security on these parameters, Eq. (3) shows the minimized objective
function for CHS. In addition, the constant constraints of
energy, distance, risk factor, and delay, are specified by ψ1 ,
Fig. 1 Representation of the proposed CHS in WSN model ψ2 , ψ3 and ψ4 respectively. The clustering model maintains
trade-off by considering the weighted sum of all the objec-
Security mode This mode chooses the CH that satisfies the tives such as energy, delay, distance and security, as shown
needs of security. In Eq. (1), sr and sd denotes the security in Eq. (4). By adjusting the weights, the significance of these
rank and security needs related with the CHS, correspond- functions can be considered in the optimization process. A
ingly. If sd ≤ sr , the node will be considered as the required common way of maintaining trade-off is to set equal weigh-
CH. Further, the selection of CH from various nodes is a tage function in such a way that the sum of all the weights to
conservative model, which is claimed to be the secure mode. be equal to unity as given in Eq. (2).
Risky mode This mode chooses a prevailing CH for
enabling the optimal CHS, for which it takes all the risks. ψ1 + ψ2 + ψ3 + ψ4  1 (2)
Thus, the risky mode is regarded as an insistent mode during
the processing of CHS. Pobjective  τ g m + (1 − τ )g n ; 0 < β < 1 (3)
γ -risky mode The CH which could bear the maximum γ -
risk are elected depending on γ -risky mode. γ -risk is also g m  ψ1 ∗ gdis tan ce + ψ2 ∗ genergy + ψ3 ∗ gdelay + ψ4 ∗ grisk
named as grisk . Consequently, γ denotes the probability mea- (4)
sure with values, γ  0 and γ  1 (i.e., 100%) similar to n 
 
1  nor m 
the security and risky mode. gn  L p − Bs  (5)
On considering all the 3 nodes, the secure mode is iden- n
p1
tified as a challenging and inexpensive mode. On the other
hand, the most deployed one is the γ -risky mode. Five secu- The fitness function in terms of distance is given by Eq. (6),
(m)
rity levels are provided by the qualitative fuzzy scale namely, where, gdistance signifies the distance among the nodes that
“very low (assigned as 1), low (assigned as 2), medium (n)
lies between 0 and 1. Here,g(distance) specifies the distance
(assigned as 3), high (assigned as 4) and very high (assigned of packets transmitted from normal node to CH and from CH
as 5)”. If the appropriate CH is chosen during the selection to BS. When there is a high distance among normal node and
process, [30] it could be regarded as the secure process. The CH, it creates larger values of gdistance .
security constraint model probability is given by Eq. (1).
(m)
⎧ gdistance
⎪ i f sd − sr ≤ 0 gdistance  (6)
⎪0
⎪ (n)
⎪
⎨
(sd −sr ) g(distance)
1−e 2
i f 0 < sd − sr ≤ 1
grisk  (1)  Dn 
L  
q  q 
3(sd −sr )
⎪
⎪  nor m
− Dc  +  Dc − Bs 
(m)
⎪
⎪ 1 − e 2
i f 1 < s d − sr ≤ 2 gdistance  L p (7)
⎩
1 i f 2 < sd − sr ≤ 5 p1 q1

 L 
L  
In addition, if the chosen CH achieves the state sd > sr the (n)  nor m 
gdistance  L p − L qnor m  (8)
risk should be less than 50%. If the condition is 0 < sd −sr ≤ p1 q1

123
 T. A. Alghamdi

The fitness function in terms of energy is given by Eq. (9). tion of the present individual and N signifies the count of the
(m) (n)
If genergy and genergy of Eq. (9) attains more number of CH neighboring individuals.
and energy, then the value of genergy will be higher than one.
The deviations of node energy and unit value are the con- 
N
Hi  − Y − Yj (14)
ditions for achieving the reduction criterion which is given
j1
by Eq. (10). Here, the improved process of CHS provides a
reduced value of n E(q). In Eq. (12), E(L qnor m ) and E(L nor
p
m)
Alignment is measured, as revealed by Eq. (15), in which
specifies the energy of qth normal node and energy of the p th Q j signifies the velocity of the jth neighboring individual. In
normal node, correspondingly. addition, the formulation for cohesion is given by Eq. (16),
in which Y j specifies the position of the jth neighboring
(m)
genergy individual, Ne symbolizes the neighborcount and Y signifies
genergy  (n)
(9)
genergy the position of the current individual.
L 
 Ne
Qj
n E(q)  1 − E(L nor
p
m
) ∗ E(Dc ) ; 1 ≤ q < Dcn Bi 
j1
(15)
p1 Ne
p∈q Ne
(10) j1 Y j
Gi  −Y (16)
Ne
Dcn

(m)
genergy  n E(q) (11) Attraction to a food resource is computed by Eq. (17), in
q1 which Y + corresponds to the position of food source and Y
Dcn  Dcn  signifies the current individual position.
(n)
genergy  Dcn ∗ Max E(L nor
p
m
) ∗ Max E(L qnor m ) (12)
p1 q1
Fi  Y + − Y (17)
The fitness function in terms of delay is given by Eq. (13).
Distraction towards an enemy is specified by Eq. (18), in
The count of nodes in cluster is directly proportional to delay
which Y − describes the enemy’s position and Y denotes the
and therefore for minimizing the delay, the count of nodes in
present position of individuals.
cluster must be reduced.

Dcn
Ei  Y − + Y (18)
q
Max Dc
q1 For updating the position of dragonflies in an exploration
gdelay  (13) space and to accomplish their movements, two vectors are
L
evaluated and they are step (Y ) and position (Y ).
The value of f delay having to lie among 0 and 1. In The step vector exposes the direction of movement of the
Hcn dragonfly, as computed in Eq. (19). Here, Hi indicates the
q
Eq. (13), Max Hc encompasses the highest count of CH separation of ith individual, p denotes the separation weight,
q1 a denotes the alignment weight, G signifies the ith individ-
and L includes the entire count of clusters in the network. ual cohesion, c refers to cohesion weight, B points to the
alignment of the ith individual, Fi corresponds to the food
resource of the ith individual, f symbolizes the food fac-
4 Optimal selection of secure and energy tor, e corresponds to the enemy factor, w refers to the inertia
aware cluster head selection weight, E i signifies the enemy’s position of the ith individual
and t indicates the iteration counter.
4.1 Conventional dragonfly algorithm
Y (t + 1)  ( p Hi + a Bi + cG i + f Fi + eE i ) + wY (t)
The chief motivation of the DA model [31] emerges from (19)
the dynamic and the static swarming mechanisms. These
two mechanisms are associated with the two foremost stages Consequent to the assessment of the step vector, the posi-
namely: (i) exploration and (ii) exploitation. These two tion vectors are manipulated by Eq. (20), in which t indicates
phases are modeled as follows: The formulation for sepa- the current iteration.
ration is evaluated, as in Eq. (14), in which Y j reveals the jth
position of the neighboring individual, Y denotes the posi- Y (t + 1)  Y (t) + Y (t + 1) (20)

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

For improving the stochastic performance of the artificial addressing the solution S is proportional to the fitness value.
dragonflies, it is essential to fly around the exploration space, Here, according to Eq. (25), the intensity of light I (w)
where there are no any neighboring solutions. In such con- changes, where I0 indicates the light intensity achieved from
ditions, the dragonfly’s position is modified by Eq. (21), in the source, and light absorption is computed by deploying
which z signifies the dimension of the position vectors, and the light absorption coefficient γ .
t stands for the current iteration.
I (w)  I0 e−γ w
2
(25)
Y (t + 1)  Y (t) + Levy(z) × Y (t) (21)
The uniformity at w  0 in I  w2 is evaded by merg-
The Levy flight is evaluated by Eq. (22), in which β is a ing the inverse square law effects and Gaussian effects. That
constant factor and r1 and r2 are the random numbers that are specifies the attractiveness of fireflies is proportional to I
lying among [0, 1]. Further, δ is computed using Eq. (23), in (w).
which Γ (x)  (x − 1). Algorithm 1 depicts pseudo-code of Therefore Eq. (26) that is similar to Eq. (24) can be rep-
the conventional DA model. resented to portray the attractiveness,β, in which β0 denotes
the attraction level at w  0. The intensity of light I and β
r1 × δ
Levy(x)  0.01 × 1
(22) are equal under specific conditions.
|r2 | β
⎛  ⎞1 β  β0 e−γ w
2
(26)
πβ β
Γ (1 + β) × sin 2
⎜ ⎟
δ⎝ 
β−1
⎠ (23)
(1+β) The distance amongst two fireflies Si and S j is specified
Γ 2 ×β ×2 2
wi j as shown in Eq. (27), in which n signifies the dimension-
ality issues. The mobility of the ith firefly is fascinated by
another firefly j that is more gorgeous.
Algorithm 1 : Conventional DA algorithm

Initialization kn
While end condition is not attained   
wi j  Si − S j  
  2
Evaluate the objective value of the entire fireflies Sik − S jk (27)
Update enemy and food source k1
Update w , p , a , c , f and e
Compute H B , G , E and F using Eq. (14-18) In this manner, Eq. (28) can be deployed, in which εi
Update the neighboring radius
If a dragonfly involves one neighbor dragonfly, denotes a random number. The movement of fireflies includes
Update the velocity vector as per Eq. (19) three features, the current location of ith firefly, attractive-
Update the position vector as per Eq. (20) ness, and a random walk that involves a variable called α
else
Update the position vector as per Eq. (21) and the randomly created integer lying between [0, 1]. When
end If β0  0, the movement depends alone on the random walks.
Verify the novel positions, depending on the variable The variable γ has a great impact on the speed of conver-
boundaries
end While gence. Even if the γ value can achieve any value lying among
γ ∈ [0, ∞], its setting is based on the issues that have to be
optimized. Usually, it ranges from 0.1 to 10.
4.2 Conventional firefly algorithm
Si (t + 1)  Si (t) + β0 e−γ wi j S j (t) − Si (t) + αεi
2
(28)
Fireflies [32] are the most charismatic insects when com-
pared with others. The major features of fireflies are their 4.3 Proposed FPU-DA algorithm
flashing light that has two basic features. The first feature
concerns with the attracting of mating partners and second The novelty of the paper is the hybridization of DA and FF
feature concerns on warning the predators. In FF, the distance (Fig. 2).
w raises with the decrease in light intensity I as shown in For the selection of the optimal cluster head in each cluster,
Eq. (24). we propose a new hybrid algorithm that hybrids the concept
1 of DA and FF, respectively. Even though the conventional
Iα (24) DA algorithm poses various advantages, including accurate
w2
approximations, it also poses certain disadvantages such as
Two major issues namely, attraction and the changes in reduced internal memory and slow convergence. Similarly,
light intensity must be resolved for designing the FF model FF also seems to be little constricting by means of conver-
in an accurate manner. In FF algorithm, light intensity I gence rate. Hence, it is planned to mingle both the concept

123
 T. A. Alghamdi

3
Initialization 2 4 5 13
Distributed
1 10 12
sensor nodes 11
9 13
6 7 8 19 20

Wireless Sensor Network

14 18 21 22
Compute the objective value of the entire 15 17 25 23 24
dragonflies 26
16
Step 1

Update enemy and food source

Update w , p , a , c , f and e

Compute H , B , G , E and F CH-

Selection CH
using Eq. (14-18)
BS
1 19 13

Update the neighboring radius 3

2 4 5
1 10 12
CH 1 11
9 13
CH 2 CH 3
6 7 8 19 20
14 18 21 22
Step 2
If neighbor=1 15 17 25 23 24
26
16
Yes
No

Update the position vector based Update the velocity vector and
on FF model as per Eq. (28) position vector by means of Eq. Fitness
(19) and Eq. (20) respectively Evaluation

Evaluate based on energy consumed to transfer data from CH1 to

Step 3 BS, CH2 to BS and CH3 to BS

Moreover, it evaluate energy consumed to transfer data from

Else
sensor nodes of CH1 to CH1, sensor nodes of CH2 to CH2,
Verify the novel positions, sensor nodes of CH3 to CH3
depending on the variable Step 4
Update CH based on fitness function
boundaries

Fig. 2 Flow chart of the proposed model Fig. 3 Cluster head selection process

in a certain way that obviously solves the optimization prob- 5 Results and discussion
lems with better convergence. The procedure of proposed
algorithm is as follows, where the concept of FF is influ- 5.1 Simulation procedure
enced into DA algorithm. More particularly, if the dragonfly
involves no neighbourhood, conventionally levy update gets The simulation of FPU-DA based CHS in WSN was per-
executed as per Eq. (21), but in the proposed algorithm, for formed in MATLAB 2018a and the results were attained.
the same case, the FF position update gets evaluated that Here, the experimentations were carried out from 0 to 2000
is given in Eq. (28). As the modification is done based on rounds. The number of sensor nodes was distributed in the
the position update of FF, the presented model is termed as network of the area 100 m×100 m with centralized BS. Here,
FPU-DA model. Algorithm 2 depicts a pseudo code of FPU- E I was the initial energy set at 0.5 and E F was the energy
DA algorithm. An example of optimal clustering model is requirement of free space model that was set at 10n J/bits/m2 .
provided in Fig. 3. In addition, E p denotes the energy of power amplifier set
at 0.0013n J/bits/m2 and E T signifies the energy of trans-
Algorithm 2 : Proposed FPU-DA algorithm mitter set at 50n J/bits/m2 . E D denotes the energy of data
Initialization
While end condition is not attained aggregation set at 5n J/bits/signal. The implementation was
Evaluate the objective value of the entire fireflies performed with respect to constraints such as delay, risk
Update enemy and food source probability, and security. Moreover, the implemented CHS
Update w , p , a , c , f and e
in WSN was distinguished with other traditional methods
Compute H B , G , E and F using Eq. (14-18)
Update the neighboring radius namely, FF [32], GWO [33], DA [31] and WOA [34]. In
If a dragonfly involves one neighbor dragonfly, addition, the NAN and NNE in the network were also veri-
Update the velocity vector as per Eq. (19) fied. In addition, five set of experiments were carried out for
Update the position vector as per Eq. (20)
Else α, β and γ values of FF model namely, [0.1 0.3 1], [0.2 0.4
Update the position based on the FF algorithm 1], [0.3 0.5 1], [0.4 0.6 1] and [0.5 0.7 1] respectively.
as per Eq. (28)
end If
Verify the novel positions, depending on the variable
boundaries
end While

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

Fig. 4 Convergence analysis of

proposed model over
conventional schemes bay
varying iterations for a set 1,
b set 2

5.2 Convergence analysis model when compared over the other schemes. From Fig. 5a,
the NAN for the proposed FPU-DA model at 2000th round
The Fig. 4 shows convergence analysis of the proposed FPU- is 18.92%, and 24.32%, better than the GWO, and WOA,
DA model over other schemes. Figure 4 is splitting into two algorithms. Moreover, the distance among rounds of NAN
sets set 1 and set 2. Set 1 is defined by the optimization are evaluated and represented in Fig. 5b. In addition, from
model research reviewed in the literature [22, 24, 25, 27] and Fig. 5b, the log of NAN for the implemented model at dis-
the proposed model is the hybridization of two conventional tance of 55 is 7.1%, and 7.1% superior to GWO, and WOA
algorithms FF and DA, so that its analysis is named as set 2. algorithms with more alive nodes. Similarly, the NAN with
From the experimental results, enhanced outcomes are found respect to set 2 in terms of number of rounds and distance is
to be attained by the implemented model over the other com- shown in Fig. 5c, d.
pared techniques. For set 1, from iteration 1 to 4, the proposed
model shows worst result when compared to PSO, GWO, and 5.4 Analysis on normalized network energy
WOA. But from iteration 4, the proposed model results best
performance over other conventional models. Iteration 6–9 The analysis on NNE for the presented FPU-DA model is
of the proposed model ensured 52.94%, 27.27%, 1.25%, and compared over other models is illustrated by Fig. 6. From
27.27% better than GA, PSO, GWO, and WOA. The cost Fig. 6, better NNE is found to be achieved by the proposed
function of the adopted model at 10th iteration is 56.25%, scheme over other schemes in terms set 1 and 2. From the
12.5%, 12.5%, and 93.75% better than GWO, WOA DA algo- Fig. 6a, the presented FPU-DA model at 2000th round is
rithms with minimal cost. In addition, for set 2 of iteration 50% and 50% better than the GWO, and WOA algorithms.
1 of Fig. 4b, the presented FPU-DA scheme is 73.33%, and Similarly, Fig. 6b shows better NNE when compared to FF
40% better than FF and DA algorithm, at 2nd iteration of and DA algorithm. Accordingly, from Fig. 6c, d, increased
the proposed model illustrates 33.33%, and 26.66% superior energy is found with respect to number of rounds. Therefore,
to FF and DA models. Further, 4th iteration of the proposed the improvement of the adopted FPU-DA technique for CHS
model is 33.33%, and 7.14% superior to FF and DA. Finally, in WSN has been substantiated successfully.
at iteration 8–10 the proposed model shows 12.5%, and 50%,
better than the FF and DA algorithms. Thus, it is known 5.5 Analysis on delay
that the adopted FPU-DA could offer better outcomes over
other compared schemes with minimum cost function. Fig- The analysis on delay for the suggested FPU-DA based CHS
ure 4b shows that the proposed model outperforms minimal model for varying rounds is given by Table 3. From the
cost functions when compared to conventional FF and DA attained outcomes, the delay of the presented model is very
algorithms. Thus, Fig. 4 proved that the proposed model con- much minimal than the other compared models. As per the
verges too fast when compared to conventional models. results in Table 3, at 1st round, the implemented FPU-DA
model is 16.93%, and 27.1% better than the FF, and DA
5.3 Analysis on number of alive nodes algorithms with less delay. In addition, Table 4 shows the
proposed FPU-DA scheme at 725th round is 10.91%, and
Figure 5 demonstrates the NAN and the log of NAN in terms 9.51% better than the FF, and DA algorithms. Thus, the supe-
of number of rounds and distance respectively. From the anal- riority of the presented FPU-DA model in terms of delay has
ysis of set 1, the NAN is found to be higher in the presented been verified in an effective manner.

123
 T. A. Alghamdi

Fig. 5 Analysis on number of

alive nodes and log of number
of alive nodes for proposed and
conventional schemes with
respect to a, b number of rounds
and distance for set 1, c,
d number of rounds and distance
for set 2

Fig. 6 Analysis on normalized

network energy and normalized
energy for proposed and
conventional schemes with
respect to a, c energy difference
and number of rounds for set 1,
b, d energy difference and
number of rounds for set

5.6 Analysis on risk probability sets. The probability of risk is much lower for the adopted
scheme over the traditional schemes. From the analysis of
Tables 5 and 6 give the risk probability analysis of the pro- Table 4, at 1500th round, the implemented FPU-DA scheme
posed FPU-DA based CHS model for varying rounds and

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

Table 3 Analysis on delay: proposed over conventional models with Table 6 Analysis on risk probability proposed and conventional models
respect to varying rounds of set 1 with respect to varying rounds of set 2
Round PSO [22] WOA [24] GWO [25] GA [27] FPU-DA Round DA [31] FF [32] FPU-DA

1 1.0571 1.1715 1.1546 0.759 0.84699 1 0.15338 0.34237 0.077825

100 0.83069 0.92534 0.79312 1.0597 1.1825 100 0.13819 0.4007 0.12692
225 1.0862 1.162 0.96268 0.98002 1.1019 225 0.13052 0.44212 0.10803
500 0.98002 1.1859 1.1065 1.0404 1.0442 500 0.10036 0.28249 0.090417
725 0.97315 1.0767 1.2677 0.98824 1.0012 725 0.042746 0.27396 0.018887
1000 1.0555 1.1088 0.98194 1000 1.0555 1000 0.096761 0.4007 0.072902
1225 1.0455 0.74423 0.81323 1225 1.0455 1225 0.082799 0.32746 0.067928
1500 1.0196 1.1733 0.72301 1500 1.0196 1500 0.14174 0.31127 0.077875
1726 0.95661 1.0196 0.81323 1726 0.95661 1726 0.15438 0.22321 0.14079
2000 0.99624 1.0252 0.90933 2000 0.99624 2001 0.074224 0.37026 0.077825
1000 0.15338 0.34237 0.077825

Table 4 Analysis on delay: proposed and conventional models with

respect to varying rounds of set 2 5.7 Algorithmic analysis
Round DA [31] FF [32] FPU-DA
The algorithmic analysis of the proposed FPU-DA based
1 1.162 1.0196 0.84699 CHS model is given by Fig. 7. To analyse the performance of
100 1.0484 0.92534 1.1825 the proposed model the fixed value of α  0.5, β  0.2 and
225 1.1369 0.99624 1.1019 γ  1] is varied by means of 5 sets by setting random values
500 1.0309 1.2537 1.0442 for α, β and γ . Here the outcomes are attained by carrying
725 1.1065 1.1238 1.0012 out five set of experiments for α, β and γ values of FF model
1000 1.1088 1.0555 0.98194 namely, [0.1 0.3 1], [0.2 0.4 1], [0.3 0.5 1], [0.4 0.6 1] and
1225 0.74423 1.0455 0.81323
[0.5 0.7 1] respectively.
1500 1.1733 1.0196 0.72301
1726 1.0196 0.95661 0.81323
5.8 Analysis on bench mark functions
2001 1.0252 0.99624 0.90933
The performance of the presented optimization model is
1000 1.162 1.0196 0.98194
tested with the help of multi-model bench mark functions
[35], results is provided in Table 7.

Table 5 Analysis on risk probability: proposed and conventional mod- 5.9 Alive node analysis
els with respect to varying rounds of set 1
Round PSO [22] WOA [24] GWO [25] GA [27] FPU-DA The performance comparison of proposed and conventional
models with respect to the number of alive nodes is shown
1 0.22719 0.084121 0.085493 0.30265 0.077825
in Fig. 8. Figure 8a represents the comparison based on
100 0.16067 0.085493 0.13052 0.15301 0.12692
set 1 models, Fig. 8b illustrates the comparison with set
225 0.26095 0.072902 0.10036 0.27354 0.10803
2 models and Fig. 8c portrays the comparison with other
500 0.1958 0.089094 0.10803 0.20702 0.090417
conventional clustering models. All these comparisons are
725 0.17687 0.13189 0.096761 0.089094 0.018887 made by varying the number of rounds. In Fig. 8a the pro-
1000 0.22719 0.12062 0.044069 0.22487 0.072902 posed model at round 2100 is 49% better than GA, PSO
1225 0.38053 0.11433 0.09539 0.22487 0.067928 and GWO subsequently, 13% better than WOA. Thus, set
1500 0.30867 0.11295 0.10173 0.27851 0.077875 1 based conventional models confirmed maximized lifes-
1726 0.17194 0.10798 0.10036 0.31525 0.14079 pan of the proposed model. Similarly, Fig. 8b depicts better
2000 0.19703 0.072902 0.1494 0.21692 0.077825 results over other models of set 2. Finally, Fig. 8c compared
proposed model with other clustering models. The other
models include K-means clustering model, FCM, DBSCAN,
and LEACH. From the figure the proposed model at round
is 23.45% and 31.05% better than the GWO and WOA algo- 2000 is 87.5%, 37.5%, 80%, and 11.11% better than FCM,
rithms. DBSCAN, and LEACH. And hence we can conclude that the

123
 T. A. Alghamdi

Fig. 7 Algorithmic analysis of the proposed model by varying the a α, b β and c γ values of FF model

Table 7 Analysis on delay:

proposed and conventional Bench mark functions PSO [22] WOA [24] GWO [25] GA [27] FPU-DA
models with respect to varying
Ackley 10.894 9.2738 10.036 6.3795 7.8856
rounds of set 2
Boh1 3.6 3.6 3.6 3.6 3.6
Colville 0 0 0 0 0
Easom − 0.99991 −1 −1 −1 −1
Hart3 − 0.30048 − 0.30048 − 0.30048 − 0.30048 − 0.30048
Mich − 41.215 − 70.828 − 51.674 − 139.47 − 45.814
Powell 3.90e+05 1.70e+05 2.96e+05 50376 2.21e+05
Powerfunc 0.073181 2.56e−07 0.000384 0.000806 0.095228
Rosen 8.18e+06 2.47e+06 6.90e+06 1.49e+05 8.82e+05
Schw 4.16e+05 4.16e+05 4.16e+05 4.16e+05 4.15e+05

Fig. 8 Analysis on alive nodes for proposed and conventional schemes with respect to a set 1, b set 2 and other models

surveying capability of the proposed model is better com- converges faster when compared to other models. The mean
pared to other models. of the convergence plot shows 0.11064, which is much less
value compared to other models.

5.10 Statistical analysis

5.11 Computational analysis
The statistical analysis of the alive nodes, normalized energy,
and convergence graph is illustrated in Table 8. In alive node The computational time analysis of the proposed and the
analysis, the mean of the proposed model shows 70 alive conventional model is illustrated in Table 9. The computa-
nodes after the completion of 2000 rounds. While, other tional time of the proposed model is 7.18%, 3.64%, 3.01%,
models show lower number of alive nodes. Similarly, the and 2.92% better than FF, GWO, WOA, and DA algorithms.
proposed model shows higher in normalized energy when From the table, we can conclude that the proposed model
compared to conventional models. Also, the proposed model attains best performance with less computational time.

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

Table 8 Statistical analysis of alive nodes, normalized energy and con- 8.09%, and 49.26% better than the FF, GWO, WOA and DA
vergence algorithms. Also, at 1500th round, the implemented FPU-DA
Models FF GWO WOA DA FPU-DA scheme was 74.98%, 23.45%, 31.05% and 45.06% better
than the FF, GWO, WOA and DA algorithms. From the
Alive nodes
experimental results, the implemented FPU-DA based CHS
Best 19 28 27 26 35
in WSN was proved to be best considered to other models.
Worst 100 100 100 100 100 Since more research works have been developed every
Mean 65.926 65.767 66.865 65.135 70.774 year, there will be a great future scope to propose advanced
Median 67 55 57 55 61 meta-heuristic algorithm to attain better coverage and con-
STD 31.941 30.763 30.183 31.982 24.449 nectivity performance in WSN. In addition, objective func-
Normalized energy tion considered here are distance and delay, and to make
Best 0.029041 0.049905 0.033991 0.049836 0.095484 improved performance, other relevant constraints can be con-
Worst 0.54958 0.54957 0.54957 0.54958 0.54958 sidered. Despite the adaptive algorithms find suitability on
Mean 0.19819 0.20634 0.20155 0.20467 0.23513 solving the context aware sensor deployment problem, the
Median 0.13213 0.13225 0.1343 0.13174 0.17645 algorithm has effect on the initialization and the number of
STD 0.16234 0.15389 0.15948 0.15561 0.13754 iterations. So, appropriate procedure for initialization and
Convergence graph number of iterations can lead to a trade-off in terms of achiev-
Best 0.12256 0.090712 0.088942 0.15737 0.081747 ing the objectives at reasonable computational cost. This can
Worst 0.20968 0.15181 0.13253 0.25871 0.15345 be well studied in the future.
Mean 0.17079 0.11719 0.12006 0.17081 0.11064
Median 0.18097 0.090712 0.12239 0.15737 0.087994
STD 0.038041 0.022609 0.014873 0.032586 0.034954 References
Table 9 Computational time 1. Pandey, O. J., & Hegde, R. M. (2018). Low-latency and energy-
Models Computational
balanced data transmission over cognitive small world WSN. IEEE
time
Transactions on Vehicular Technology, 67(8), 7719–7733.
FF 9756 2. Senouci, M. R., & Mellouk, A. (2019). A robust uncertainty-
aware cluster-based deployment approach for WSNs: Coverage,
GWO 9434 connectivity, and lifespan. Journal of Network and Computer
WOA 9377 Applications, 146, 102414.
DA 9368 3. Vieira, R. G., Cunha, A. M., Ruiz, L. B., & Camargo, A. P. (2018).
On the design of a long range WSN for precision irrigation. IEEE
FPU-DA 9102
Sensors Journal, 18(2), 773–780.
4. Hintsch, T., & Irnich, S. (2018). Large multiple neighborhood
search for the clustered vehicle-routing problem. European Journal
6 Conclusion of Operational Research, 270(1), 118–131.
5. Ahmad, A., Javaid, N., Khan, Z. A., Qasim, U., & Alghamdi, T.
A. (2014). Routing scheme to maximize lifetime and through-
This paper has presented a novel clustering model with put of wireless sensor networks. IEEE Sensors Journal, 14(10),
optimal CHS by considering four major criterions such as 3516–3532.
energy, delay, distance and security. The analysis such as con- 6. Alghamdi, T. A. (2016). Cluster based energy efficient routing pro-
tocol for wireless body area networks. Trends in Applied Sciences
vergence analysis, alive node analysis, normalized network
Research, 11(1), 12–16.
energy, delay analysis, risk probability analysis, algorithmic 7. Alghamdi, T. A. (2018). Secure and energy efficient path optimiza-
analysis and statistical test has been made. The evalua- tion technique in wireless sensor networks using DH method. IEEE
tion results proved that the proposed model attains superior Access, 6, 53576–53582.
8. Krishnan, M., Yun, S., & Jung, Y. M. (2019). Enhanced clustering
results compared to other models. From the Alive node and ACO-based multiple mobile sinks for efficiency improvement
analysis, the proposed model at 2000th round was 45.95%, of wireless sensor networks. Computer Networks, 160, 33–40.
18.92%, 24.32% and 24.32% better than the FF, GWO, WOA 9. Radhika, S., & Rangarajan, P. (2019). On improving the lifespan of
and DA algorithms. On considering the convergence anal- wireless sensor networks with fuzzy based clustering and machine
learning based data reduction. Applied Soft Computing, 89, 105610.
ysis, the NAN for the proposed model at 2000th round is 10. Jesudurai, S. A., & Senthilkumar, A. (2019). An improved energy
45.95%, 18.92%, 24.32%, and 24.32% better than the FF, efficient cluster head selection protocol using the double cluster
GWO, WOA and DA algorithms. In addition, the delay at heads and data fusion methods for IoT applications. Cognitive Sys-
1st round, the implemented FPU-DA model was 16.93%, tems Research, 57, 101–106.
11. Zhao, B., Ren, Y., Gao, D., Xu, L., & Zhang, Y. (2019). Energy
26.64%, 27.7%, and 27.1% better than the FF, GWO, WOA utilization efficiency evaluation model of refining unit based on
and DA algorithms. From the analysis, the presented model Contourlet neural network optimized by improved grey optimiza-
in terms of risk probability at 1st round was 77.27%, 8.97%, tion algorithm. Energy, 185, 1032–1044.

123
 T. A. Alghamdi

12. Chen, L., Yang, D., Zhang, D., Wang, C., & Nguyen, T.-M.-T. 26. Yahiaoui, S., Omar, M., Bouabdallah, A., Natalizio, E., & Challal,
(2018). Deep mobile traffic forecast and complementary base sta- Y. (2018). An energy efficient and QoS aware routing protocol for
tion clustering for C-RAN optimization. Journal of Network and wireless sensor and actuator networks. AEU - International Journal
Computer Applications, 121, 59–69. of Electronics and Communications, 83, 193–203.
13. Behera, T. M., Mohapatra, S. K., Samal, U. C., & Khan, M. S. 27. Tianshu, W., Gongxuan, Z., Xichen, Y., & Ahmadreza, V. (2018).
(2019). Hybrid heterogeneous routing scheme for improved net- Genetic algorithm for energy-efficient clustering and routing in
work performance in WSNs for animal tracking. Internet of Things, wireless sensor networks. Journal of Systems and Software, 146,
6, 100047. 196–214.
14. Yarinezhad, R., & Hashemi, S. N. (2019). Solving the load 28. Ennaciri, A., Erritali, M., & Bengourram, J. (2019). Load balancing
balanced clustering and routing problems in WSNs with an FPT- protocol (EESAA) to improve quality of service in wireless sensor
approximation algorithm and a grid structure. Pervasive and Mobile network. Procedia Computer Science, 151, 1140–1145.
Computing, 58, 101033. 29. Sarkar, T. A., & Murugan, S. (2019). Cluster head selection for
15. Saini, A., Kansal, A., & Randhawa, N. S. (2019). Minimization energy efficient and delay-less routing in wireless sensor network.
of energy consumption in WSN using hybrid WECRA approach. Wireless Networks, 25(1), 303–320.
Procedia Computer Science, 155, 803–808. 30. Liu, T., Li, Q., & Liang, P. (2012). An energy-balancing clustering
16. Wang, L., Lehman, V., Hoque, A. K. M. M., Zhang, B., Yu, Y., & approach for gradient-based routing in wireless sensor networks.
Zhang, L. (2018). A secure link state routing protocol for NDN. Computer Communications, 35(17), 2150–2161.
IEEE Access, 6, 10470–10482. 31. Jafari, M., & Chaleshtari, M. H. B. (2017). Using dragonfly
17. Thangaramya, K., Kulothungan, K., Logambigai, R., Selvi, M., algorithm for optimization of orthotropic infiniteplates with a
Ganapathy, S., & Kannanc, A. (2019). Energy aware cluster and quasi-triangular cut-out. European Journal of Mechanics A/Solids,
neuro-fuzzy based routing algorithm for wireless sensor networks 66, 1–14.
in IoT. Computer Networks, 151, 211–223. 32. Gandomi, A. H., Yang, X.-S., Talatahari, S., & Alavi, A. H. (2013).
18. Waqas, M., Malik, S. U. R., Akbar, S., Anjum, A., & Ahmad, N. Firefly algorithm with chaos. Communications in Nonlinear Sci-
(2019). Convergence time analysis of OSPF routing protocol using ence and Numerical Simulation, 18, 89–98.
social network metrics. Future Generation Computer Systems, 94, 33. Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf opti-
62–71. mizer. Advances in Engineering Software, 69, 46–61.
19. Ansari, A. R., & Cho, S. (2018). CHESS-PC: Cluster-HEad selec- 34. Mirjalili, S., & Lewis, A. (2016). The whale optimization algo-
tion scheme with power control for public safety networks. IEEE rithm. Advances in Engineering Software, 95, 51–67.
Access, 6, 51640–51646. 35. Boothalingam, R. (2018). Optimization using lion algorithm: A
20. Faheem, M., Butt, R. A., Raza, B., Ashraf, M. W., Ngadi, A., & biological inspiration from lion’s social behavior. Evolutionary
Gungorb, V. C. (2019). Energy efficient and reliable data gathering Intelligence, 11, 31–52.
using internet of software-defined mobile sinks for WSNs-based 36. Ahmad, A., Javaid, N., Qasim, U., Ishfaq, M., Khan, Z., &
smart grid applications. Computer Standards and Interfaces, 66, Alghamdi, T. (2014). RE-ATTEMPT: A new energy-efficient rout-
103341. ing protocol for wireless body area sensor networks. International
21. Toor, A. S., & Jain, A. K. (2019). Energy aware cluster based Journal of Distributed Sensor Networks, 10, 464010.
multi-hop energy efficient routing protocol using multiple mobile 37. Zhu, E., Zhang, Y., Wen, P., & Liu, F. (2019). Fast and stable clus-
nodes (MEACBM) in wireless sensor networks. AEU - Interna- tering analysis based on grid-mapping K-means algorithm and new
tional Journal of Electronics and Communications, 102, 41–53. clustering validity index. Neurocomputing, 363, 149–170.
22. Kaur, S., & Mahajan, R. (2018). Hybrid meta-heuristic 38. Wang, Q., Guo, S., Hu, J., & Yang, Y. (2018). Spectral partitioning
optimization-based energy efficient protocol for wireless sensor and fuzzy C-means based clustering algorithm for big data wire-
networks. Egyptian Informatics Journal, 19(3), 145–150. less sensor networks. Journal on Wireless Communications and
23. Mohamed, R. E., Ghanem, W. R., Khalil, A. T., Elhoseny, M., Saj- Networking, 2018, 54.
jad, M., & Mohamed, M. A. (2018). Energy efficient collaborative 39. Jafari, H., Nazari, M., & Shamshirband, S. (2018). Optimization
proactive routing protocol for wireless sensor network. Computer of energy consumption in wireless sensor networks using density-
Networks, 142, 154–167. based clustering algorithm. International Journal of Computers
24. Sharawi, M. & Emary, E. (2017). Impact of grey wolf optimization and Applications, 2018, 1–10.
on WSN cluster formation and lifetime expansion. In 2017 Ninth 40. Moorthi, M., & Thiagarajan, R. (2020). Energy consumption and
international conference on advanced computational intelligence network connectivity based on Novel-LEACH-POS protocol net-
(ICACI), Doha, pp. 157–162. works. Computer Communications, 149, 90–98.
25. Jadhav, A. R. & Shankar, T. (2017). Whale optimization based
energy-efficient cluster head selection algorithm for wireless sensor
networks. Neural and Evolutionary Computing. arXiv:1711.0938
Publisher’s Note Springer Nature remains neutral with regard to juris-
9.
dictional claims in published maps and institutional affiliations.

123
Energy efficient protocol in wireless sensor network: optimized cluster head selection model

Turki Ali Alghamdi received the

bachelor’s degree in computer
science from King Abdulaziz
University, Jeddah, Saudi Arabia,
the master’s degree in Distributed
Systems and Networks from the
University of Hertfordshire, Hat-
field, United Kingdom, in 2006
and the Ph.D. degree from the
University of Bradford, United
Kingdom, in 2010. He is an
Associate Professor in Computer
Science Department, at Univer-
sity of Umm Al-Qura (UQU),
Makkah, Saudi Arabia and he
is the Founding Director of UQU Smart Campus Center (SCC). He
has more than 15 years of research and development, academia and
project management experience in IT. He holds CDCDP and CDCMP
certificates. He is passionate about developing the translational and
collaborative interface between industry and academia. His research
interests focusing on Wireless Sensor Networks, Energy and QoS
Aware Routing Protocols, Network Security, IoT and Smart Cities.
He serves as peer reviewer for several international Web of Science
Indexed journals and organizer for International Conferences and
workshops.

123