Wireless Sensor Network (WSN) is composed of sensing nodes that wirelessly transmit data. Sensor nodes are spread in an environment in order to monitor the specific physical phenomena such as temperature, humidity, light, pressure. These nodes have many advantages compared with the wired sensor technologies: compact size, mobility feature, ease of installation and low power consumption [1].

Sensor nodes are able to sense the physical parameters, measure the data and gather information from the environment. A typical sensor node is generally composed of power unit, sensor unit, processing unit and communication unit. Sensing and processing unit are responsible for collecting the data and processing the signals acquired from sensors before transferring them to the network. The communication unit task is to transmit the data to the main base station, where it can be further analysed by the user. The power consumption of the supply unit is the main problem in sensor nodes technology because of the constrained energy resources [2]. Two main reasons for the distribution of the energy consumption mainly lead to the asymmetry of the energy consumption between nodes. One is the energy consumption from the communications between nodes and base station, and the other is from the Cluster Head (CH) for receiving data from other nodes [3]. Baggio [4] developed energy-efficient T-Medium Access Control (MAC) protocol in precision agriculture using WSN’s. Sensor nodes should work autonomously for a long period of time especially in harsh environment or areas with no available power infrastructure [5]. The limited sensor node energy with the dynamic network topology information has always been the important design concern in WSN’s. Clustering of the nodes by grouping the nodes into hierarchies is also an effective way to reduce communication distance and the amount of message [6].

WSN’s are used for intelligent monitoring of temperature, humidity, pressure, water level, remote health monitoring of patients, military applications and many other applications. Podgornik et al. [7] present the concept of a platform for analysis of the energy consumption and energy efficiency in urban and industrial areas.

Sensor nodes are usually battery powered wireless devices. The life of the sensor network is mainly influenced by the constrained power resources of the nodes. The small size of the nodes is the main factor for power resource limitations. For example, the total energy that can be stored in a small sensor node is in the range of 1 J [8]. For typical small systems like Wireless Integrated Network Sensors (WINS) [9], to provide better operation time of the batteries the average current consumption should be less than 30 μA. In this type of sensor nodes, the power units are charged from coin cells made from Lithium (Li) (the typical dimensions of the coins are: 2.5 cm in diameter and 1 cm in thickness) [9]. One possible solution to extend the duration of the batteries is by energy harvesting [10], which means extracting energy from the environment. One possible solution for energy conservation is by using photovoltaic cells. Self-powered WSN’s provide the possibility of very long sensor node life while their deployment would have the least impact on the existing infrastructure. With the integration of renewable energy, especially solar and wind power, it is possible to provide power supply for remote communities. The combination of the renewable energies, wind, solar and energy storage enable sustainable energy production for remote communities and the energy cost can be lower than diesel production as well [11]. Also, Ma et al. [12] made a final decision to take renewable energy, i.e., solar and wind energy or their combination, since it would be the lowest cost option when compared with remote diesel power generation. In addition, they concluded that renewable energy technology is now mature enough to provide utility quality power supply at a reasonable cost.

Therefore, the primary goal in wireless sensor technologies is focused on power conservation while in the traditional networks the main goal is to achieve high Quality of Service (QoS). Actually, the trade-off between energy consumption and QoS in terms of reliable data delivery is active research [13]. It is challenging to optimize the network performances that minimize the energy consumption and in the meanwhile guarantee QoS.

The minimization of the power node battery consumption is investigated in more detail in order to obtain lower power operation which extends the battery life of the nodes [14].

One of the greatest issues that WSN’s faced is energy constraint. It is challenging to prolong energy-power and network life of sensor nodes. Many research topics are focused on efficient energy consumption approaches such as data aggregation, topology construction and energy harvesting in WSN’s [15]. Even though there are different methodologies in related works, they have the same ultimate goal which is efficient energy consumption. Therefore, the existing problem of energy constrained resources in WSN’s and the possible ways to solve it is the main motive to start our work in this field.

The aim of this study is to use the stochastic method in order to optimize the energy consumption of the wireless sensor nodes which depends on the frequency of the transmitted data and the period of the transmission process.

The paper is organized as follows: in the second section a literature review is given. The third section describes the genetic algorithm implemented in WSN’s. The mathematical model and the measurement results for optimization of the energy consumption are presented in the fourth section. Conclusions are presented in the last section.

In recent years, a diversity of works is related to efficient energy consumption in sensor network technologies. Data aggregation is a technique to eliminate redundant data and reduce energy consumption. In their work, Heinzelman et al. [15] is focused on efficient energy consumption in delivering data from sensor nodes to the base station. Data aggregation protocols eliminate redundant sensor nodes’ data sent to the base station. Boonsongsrikul et al. [16] proposed a new topology construction for sensor network topologies in order to minimize bottleneck of data transmission and reception. Most of the works to prolong the life of the sensor nodes are focused on clustering algorithms. One of the most popular Low Energy Adaptive Clustering Hierarchy (LEACH) algorithms for extending the life of the nodes with relatively simple principle and low complexity was proposed by Heinzelman et al. [17].

Several improved algorithms based on LEACH were proposed. Multi-hop LEACH algorithm was proposed by Biradar et al. [18]. In this work, if the Base Station (BS) is out of the scope of the CH, other CH’s could work as relay nodes for it. Besides the decentralized cluster algorithms, some centralized algorithms were proposed [19], which configured the BS as the main optimizing controller to achieve the selection process of the CH’s. To optimize the energy consumption in addition to LEACH, Hybrid Energy Efficient Distributed (HEED) algorithm was also proposed [20]. The HEED clustering protocol adopted the distributed clustering process during which the node with more remaining energy had a higher probability to be the CH. Some authors select the CH’s based on the residual energy of the sensor nodes [21]. In this work they choose the sensor nodes with the highest residual energy as the CH nodes, mainly because the CH consumes more energy than the nCH nodes. In [22], the authors select the CH’s based on their proximity to the BS which is computed in the form of a distance metric. It is assumed that since the transmission distance plays important role in the energy dissipation model, the sensor nodes with the shortest distances to the BS will consume less energy when forwarding packets to the BS.

Some research works are focused on energy harvesting in WSN [23]. Energy harvesting in WSN means gathering energy from the environment or other energy sources and converting it to electrical energy of sensor nodes. Approaches relevant to energy harvesting can be categorized into two groups: based architectures and based protocols. In the based architectures, the energy harvesting is related to additional hardware such as solar cells or piezoelectrical materials that convert other sources into electrical energy. Energy storage devices such as recharged batteries or storage capacitors are required for energy-harvesting sensor nodes. When more electrical circuits are added in a sensor node, more energy consumption is required. To prolong its life, rather than using only electrical circuits to measure residual energy, a sensor node should be able to evaluate its remaining energy by learning, reasoning and information shared among multiple nodes. Zhang et al. [23] analysed the maximization of the life in a single cluster topology using the same energy model as LEACH. In the based protocols, an energy harvesting protocol refers to a communication protocol relevant to a sensor node being able to harness energy. Yoo et al. [24] proposed schemes that allow sensor nodes to adjust duty-cycle based on residual energy. Duty – cycle should be adjusted on residual energy in order to reduce the sleep latency and to balance energy consumption.

In this paper we discuss about the possibilities and challenges of our proposed stochastic optimization method-genetic algorithm for effective energy consumption in WSN. We use different scenarios for analysis of the energy consumption of sensor nodes depending on the frequency of the transmitted data and the period of transmission of all the active components. The duty cycles in sensor nodes consist of active and sleep mode. After the initialization of the sensor and transmitting the data to the base station, the sensor node transits to low power mode and stays in this mode until the next transmission period. The transition period between the different modes of operation should be with small duty-cycles in order to obtain maximum efficiency.

Effective energy consumption approaches have different methodologies and performance, but still there are some common aspects such as data delivery from sensor node to the base station, requirements of extra hardware, balancing topologies and scalability in WSN.

The parameters used for the analysis are mainly taken from the previous research in [25]. The duty cycles in active and sleep mode of the energy consuming components in the sensor nodes are measured for different scenarios. In this study the mathematical model for energy consumption optimization in wireless sensor networks is presented. The model is designed according to the different input values which depend on the frequency of the transmitted data and the period of the transmission. For the optimization process the genetic algorithm with population size from 20 to 100 is used.

For the optimization process the input values are chosen to be data packets with different transmission period and different data package length as the energy consumption mainly depends from these parameters.

The input parameters in the calculation of the energy consumption using the GA optimisation process are given in Table 2.

The optimization goal function is defined according to the following equation of the total energy A consumed for transmission of one complete information package:

The purpose of the optimization process is to achieve minimum consumed energy per information package. However, the original GA program is used to optimize the goal function and in the optimization process the auxiliary goal function (f_opt) is given by the following formula:

The genetic parameters to obtain optimal solution are calculated according to the following values: cross-over probability is 0.65, mutation probability 0.03, population size 16, maximal number of generations 30,000.

In Figures 1–4, the GA optimisation process is presented through the 30,000 generations. The main purpose to optimize the energy consumption is to minimize the parameters N_P, N_data, N_mac as well as optimization of the total transmission time per information package of the sensor node.

Figure 1.

Optimal goal function

Figure 2.

Optimization of the input variable 1 (information package MAC heading length)

Figure 3.

Optimization of the input variable 2 (data string length in the information package)

Figure 4.

Optimization of input variable 3 (number of data strings per information package of the sensor node in transmitting mode)

In Figures 5–8, the optimal output values for the total information package length, total transmission time per information package and the total energy consumption of the sensor node components are presented.

Figure 5.

Total information package length (transmitting mode)

Figure 6.

Total transmission time per information package of the sensor node (transmitting mode)

Figure 7.

Total energy consumption per information package of the microcontroller

Figure 8.

Total energy consumption per information package of the transmitter

The optimal value for the energy consumption of the sensor node, A = 780.0901 μJ is obtained for the following optimal input variables presented in Table 3.

From the Table 3 we can conclude that in the optimization process the length of the MAC heading information package is equivalent to the minimal value. The total information package data is equivalent to data string length in the information package. The optimal energy consumption of the sensor node per information package is presented on Figure 9.

Figure 9.

Total energy consumption per information package of the sensor node

In Table 4 some of the optimal results (total energy consumption of the sensor module, total energy consumption of the microcontroller, total energy consumption of the transmitter, total transmission time per information package of the sensor node and total information package length) are shown.

The purpose of the optimization process is to achieve minimum consumed energy of the sensor nodes per information package. However, the original GA program maximizes the goalfunction and it can be concluded that to obtain minimal energy consumption the total information package length of the data should be bigger and the frequency of the transmitted package information should be sent in long time intervals. The total transmission time per information package of the sensor node is 0.224 ms.

In this paper, to minimize the energy consumption of the wireless sensor nodes the stochastic method is used-genetic algorithm that is optimized depending on the frequency of the transmitted data and the period of the transmission process of the sensor nodes. From the results of the optimization process the optimal input parameters for minimal energy consumption of the sensor node are derived.

In the optimization process the energy consumption of the sensor nodes is calculated as a result of all the components consumers in the sensor node: sensor module, microcontroller and the radio transmitter. Minimization of the energy consumption of sensor nodes depending on the frequency of the transmitted data and the period of transmission of all the components are analysed depending on different duty cycles in sensor nodes (active and sleep mode).

By using the GA the minimum consumed energy per information package is calculated. It can be concluded that the frequency of the transmitted data packets should be sent in time intervals with bigger information packages in order to obtain minimal energy consumption.

For further investigation the number of optimization input variables could be increased. The applied methodology is universal and can be applied to other WSN optimization problems.