A heuristic deep feature system for energy management in wireless sensor network

The Wireless-Sensor-Network (WSN) has been employed in all digital applications for several purposes like sensing, storing, and sharing information. However, managing energy consumption is more critical because of the movable environment. Several existing models have addressed these energy management issues. Still, those models lack in optimizing the energy usage of the WSN during the collision environment. This has motivated to find the best solution for energy optimization with an intelligent model. So, the present research article aims to develop the novel Buffalo-based Deep Belief Energy Management Framework (BDBEMF) for the WSN application. Initially, the required number of sensor nodes was created then the BDBEMF was designed to monitor the high consumption nodes. In addition, the Low-energy adaptive-clustering-hierarchy protocol has been considered for the communication process. Consequently, the Cluster Head has been selected based on less energy utilization and high-density hubs. The data rate of each node has been measured, and the high leaded data has been shared to work fewer nodes to balance the energy. Finally, the amount of alive and dead nodes was validated with few communication metrics. The presented model has gained maximum throughput and less energy consumption.


Introduction
The WSN comprises many tiny, low-cost hubs with sensing elements, energy, and memory constraints [1]. Multiple issues arise when learning each component in this specific type of network [2]. The recent advancements in WSN applications have facilitated the widespread adoption of low-power [3], low-cost, and multi-operational sensors with smaller diameters and short communication ranges. Cheap, intelligent sensors, networked via wireless networks and widely distributed [4], provide unprecedented monitoring and management of homes, the environment, and cities [5]. Additionally, network sensors are used in a broad range of defense applications, enabling unique innovations for observation, surveillance, and various tactical uses [6]. The ability of WSNs to self-localize can be a desirable characteristic [7]. Without knowledge about the data position in environmental monitoring applications such as agriculture, bus-fire management, and monitoring water quality, measurement info is worthless [8].
The structure of WSN in the user environment is defined in Fig. 1. Here, the 20 sensor nodes are present with one base station. Also, all nodes are in wireless and movable status. Additionally, different positions estimate various purposes and applications, including inventory management, transportation, road traffic analysis, health monitoring, intrusion detection, and surveillance. Moreover, the advancements in communication technologies [9] and sensing have employed many low-power and low-cost sensors that have been established [10,11]. Numerous small battery-powered sensor hubs are dispersed around a physical region in a WSN [12]. Furthermore, each sensing facility in the sensing environment has captured the data, such as temperature, radiation, vibration [13], and other environmental parameters [14]. Numerous energy management techniques have been presented in the past, in that more methods are based on the data assumption [15], data collecting, and processing function [16]. Hence, it consumes substantially less energy than transmission [17]. But, in today's world, WSN is a practical application for the communication process in all digital fields [18]. So, the arranged sensor hubs in the WSN environment must accomplish multiple processes like target finding observation and sharing the data [19]. These various processes have resulted in high energy consumption than the previous WSN [20].
Several works, like the Data Dissemination model [21], blockchain with key management process [22], etc., were designed to end these energy management issues, but those approaches resulted in other cases. Hence, the present article has aimed to develop a novel optimized deep feature strategy to manage energy utilization in the WSN environment. Moreover, Low-Energy-Adaptive-Clustering-Hierarchy (LEACH) is an advanced protocol network in the WSN environment with a clustering model. In addition, the main reason for the high energy consumption in the WSN is the sensing capacity of each node. Hence, several models were introduced to end these energy consumption issues, but those approaches have met the different issues based on prediction and energy optimization. The WSN has afforded the most acceptable communication range and sensing output. However, it is high in resource and computational costs. These issues have motivated this present study to design the energy management system's monitoring features.
Hence, the novelty of this work lies in introducing the African buffalo and deep belief model in the WSN for the energy management system. Here, the monitoring strategy has been processed based on buffalo fitness. It helps to keep the WSN in optimal status. Moreover, the designed, optimized energy utilization system is planned to be implemented with the LEACH protocol. The LEACH is the advanced protocol effectively utilized in the WSN application. Incorporating the energy management scheme has enriched LEACH's performance. The key strategy of the proposed model is as follows.In the primary phase, the WSN environment developed with the desired number of nodes.
• In the primary phase, the WSN environment is developed with the desired number of nodes. • Consequently, the LEACH protocol has been developed to execute the communication process. • Moreover, the novel BDBEMF has been designed with the required energy constraint modules. • If the collision occurs during the transmission process, half the data is shared with the other hubs that have tended to optimize the energy usage between the sensor nodes. • Finally, the function of the proposed design is validated by calculating and comparing the key metrics with other models in energy consumption, throughput, delay, packet drop, and packet transmission.
The planned research design is organized as follows; recent related literature is described in Sect. 2, the problem in Leach protocol is described in the third section, the solution for the discussed issues is elaborated in the fourth section, and the outcome of the proposed solution is described in the fifth section, and the research argument is concluded in Sect. 6.

Related works
A few recent works associated with this research work are described below.
To control the energy usage in the WSN communication, Kuthadi et al. [21] have designed the Data Dissemination model, which has incorporated the optimal energy constraints. In addition, for better communication, an adaptive protocol has been developed. Here, the data was shared from one point to multiple sources, and then the utilized energy resources were measured. Hence, the reduced energy consumption and wide data broadcasting rate have been recorded. But, it is not qualified for security.
Securing communication is an essential task in WSN. Hence, the blockchain with a critical management process has been framed by Jia et al. [22] in the WSN to manage energy usage and secure communication. This medium supports extensive cloud data and has gained the most delicate personal score. However, if the data size has crossed the average limit during the transmission, extra energy resources were required to proceed with the communication process.
Osamy et al. [23] have designed a clustering strategy that relies on optimal head node selection for the clustering process. Hence, the created head node has collected the sources of their group node together from the base station and forwarded it to the particular hub. Simultaneously, the start of the cluster users has been collected by the cluster head and then sent to the target through the base station. However, if the cluster node gets injected, it degrades communication.
A protocol like genetic and LEACH has been developed for the WSN environment by Bhola et al. [24]. Here to minimize the energy resources, the fitness process of the genetic model is utilized to identify the best route to share the data. Hence, if the shortest path has been selected correctly, it reduces the execution and communication process. As a result, the usage of energy has been diminished. But, while preceding the communication through the shortest path, packet drop will happen if any node is disabled.
On the other hand, clustering algorithms were implemented by Israel Agbehadji et al. [25] for the energy optimization objective in the WSN environment. Moreover, this clustering algorithm has been designed based on heuristic models. Here, dual goals have been taken into consideration: maximizing the node lifetime and optimizing the energy usage. It has afforded the finest optimized energy usage results. But, it has taken more time and high design complexity to execute the process than the other models.
In short, the recent approaches have met different problems while providing a better-optimized energy consumption outcome, as detailed in Table 1. So, the present work has introduced novel optimized deep networks to enhance LEACH performance. Here, the LEACH protocol has updated the additional parameters by minimizing high energy consumption and data overhead monitoring features with a load balancing scheme. It has tended to result in high data transmission with less energy consumption. Also, to maintain the optimal status load balancing scheme has been introduced. Thus it can reduce the collision that occurred in the WSN, which has reduced the communication delay and maximized the throughput range. In addition, the better performance of the proposed model has reduced the computational cost.

System model and problem statement
The WSN has a lot of facilities and advancements to use in digital applications; considering those advantages; there are many issues such as pack drop, poor communication range, high energy consumption, less lifetime node, and so on. Considering all those issues, the high energy consumption is the primary cause of the network performance degradation. Moreover, it is described that the wide range of energy utilizing hubs in the sensor environment might degrade the entire WSN communication system by reducing its lifetime. This reason has motivated this research to manage energy consumption in the WSN environment.
The reason for considering LEACH in this present research work is high energy consumption. The problems that have occurred in conventional LEACH are described in Fig. 2. Usually, the LEACH model in WSN requires more energy to execute because the dead nodes have been increased while the iteration counts were increased. Hence, in the WSN, if the dead nodes are expanded, the other alive nodes need more energy for better data transmission. Considering these demerits, a novel optimized deep neural mechanism has been implemented to reduce the dead nodes by neglecting the high energy utilization nodes in earlier stages.

Proposed BDBEMF in WSN
The present work aims to design the novel Buffalo-based Deep Belief Energy Management Framework (BDBEMF) that has been planned and implemented for the WSN system. Hence, the robustness of the proposed model has been evaluated in the LEACH protocol. Finally, the key metrics have been measured and compared with other models; the proposed architecture is described in Fig. 3. Here, the optimal energy value is fixed in the buffalo fitness model. The iteration is processed during the execution until it reaches the fixed optimal energy constraints.
Hence, a load balancing strategy was developed to reach optimal energy utilization status. This can avoid congestion during the data transmission process. Finally, the improvement score of the designed framework is validated by making a comparative analysis with other associated works.

Node formation
The routing protocol that is considered in this research work is LEACH. Usually, the LEACH contains a clustering scheme to manage energy utilization for communication functions; hence, the LEACH model is taken. However, data overloading might disturb the WSN system and has tended to come with more energy. Therefore, a novel monitoring mechanism and load balancing model have been implemented in this research. Moreover, the planned technique is named BDBEMF, which functions with the principle of buffalo optimization [28] and the deep belief neural model [29]. The creation of a number of sensor nodes is processed by Eq. (1). The node initialization variable is determined as f ðs n Þ, and the sensor hubs are determined as s n .
f ðs n Þ ¼ s n f1; 2; 3; 4. . .nÞ ð 1Þ To avoid packet flow rate, the data has been transmitted through the active hubs. In WSN, there are nodes with sleep status or fewer data ranges, which will be disabled at any time. So, if the data is not sent through the active hubs, packet drop occurs. Hence, to avoid this issue, the active status nodes have been predicted before initiating the datasharing process.
Here, b is the monitoring variable, which is taken from the buffalo algorithm, s ni is the source sensor node and s nj is the receiver node. Moreover, the status activation evaluation variable is described as T s , the iteration of sensor nodes generation is represented as s n ðt ¼ 1Þ. Node formation in the MATLAB environment is detailed in Fig. 4. Here, 200 hubs were generated with several cluster nodes.

Cluster head selection
The cluster head was selected by analyzing the density of the node and energy consumption. Hence, the eligible cluster head is elected by Eq. (3). Here, 0.5 is the fixed range attained from the buffalo model, the first learning parameter fitness [28]. Furthermore, n d is node density and n e is the energy utilization percentage of each node. Moreover, the cluster-head is denoted as C h and the E c which are the energy utilization variables. Here, the measure of energy utilization is validated by Eq. (3).
If the measured energy usage rate is less than or equal to 1, it is elected as a cluster head equated in Eq. (4). But, if it is not matched, then another node has to be estimated. Hence, the process is iterated continuously till the condition is met.

Load balancing
In addition, the balancing module has been processed by Eq. (5), the variable L b is the load balancing parameter, n d is the node density, and R d denotes the data rate. To find the collision occurrences, the maximum data rate is identified, and then the data rate of each node is subtracted from the maximum data rate. From this calculation, if the results  are more significant than 0.6, then it indicates the occurrence of a collision. Here, the data sharing parameter is determined as D s . Here, 0.6 is taken from the buffalo's second learning parameter fitness [28].
If the load balancing analysis has met the condition, then half the data is shared with the freehubs in the developed WSN. Here, the beta is the monitoring parameter in the buffalo model, that is, head buffalo, which is utilized to find the safe location. The data migration or sharing has been executed using Eq. (6).
Consequently, freehubs have to be identified to reduce the congestion rate. Hence, the free status node is estimated by Eq. (7). If any node in the connected WSN has less than 0.5 Mbps data rate, that hub is said to be freehub. Moreover, half of the data from the high-loaded node is shared with the freehubs. Hence, the loads in the WSN system were balanced, and the energy utilization rate is in the optimal status.
The designed mathematical formulation is described in the form of pseudo-code, defined in the algorithm 1. The working procedure of the novel BDBEMF is given in Fig. 5. Here, the energy utilization rate has been optimized by enabling the load balancing model and clustering approach.

Results and discussion
The planned design is checked in the MATLAB environment running on the Windows 10 platform. The required sensor nodes have been developed, and the novel BDBEMF has been modeled with several features, such as load balancing, energy optimization, congestion control, and high data delivery. The execution parameters and their specification is detailed in Table 2.

Case study
To check the improvement score of the proposed technique in the leaching environment, normal LEACH and BDBEMF-based LEACH has been validated. Also, the parameters are validated in dual phases before and after applying the buffalo fitness.
There are a total of 200 nodes. In those nodes, few hubs are CH, some are less energy utilization nodes, and some are full energy-consuming nodes. These node frames have been generated before applying the chimp optimization model. Node design with the elected CH. is described in Fig. 6.
The amount of alive and dead hubs was calculated in dual phases before incorporating the buffalo fitness and after including the buffalo fitness solutions. Moreover, in B.O, the number of dead nodes was increased, indicated by the yellow triangles presented in Fig. 7, and the graphical representation of the same is in Fig. 8. In addition, the dead nodes, which have consumed more energy, are described as yellow triangles with a plus symbol. Also, the green nodes notify the live nodes; the green circle with red color plus indicates the widest energy utilization node, and the only green circle indicates the less energy consumption nodes.
Moreover, the CH. is represented as a blue plus symbol. Before applying the buffalo fitness, the dead nodes are in maximum counts, leading to high energy wastage outcomes.
Hence, to design the energy-optimized WSN, a novel BDBEMF has been developed in the present work. The performance of the BDBEMF after applying the buffalo fitness is explained in Fig. 9. Measuring the energy wastage based on iteration count is much more important to check the robustness of the designed model in the LEACH concept. Hence, the metrics of energy dissipation are validated in dual cases before and after applying the buffalo fitness. The presented novel BDBEMF model has scored the reduced energy dissipation rate in double instances. This desirable outcome is gained because of the monitoring model process in the initial layer. So, in the starting phase, the high energy utilization hubs were removed, resulting in fewer dead nodes.
Hence, the energy dissipation score has been calculated in nJ, which is in the negligible state; it has verified the robustness of the presented model. Also, considering the before Optimization (B.O.), after applying the buffalo optimization has helped to earn the finest outcome for 200 iterations, the recorded energy dissipation score is 0.05 nJ. In addition, before using the buffalo algorithm process, the recorded energy dissipation score was 0.15 nJ. These statistics are analyzed in Fig. 10.
The communication delay has been validated based on seconds; after incorporating the buffalo fitness, the communication delay has been considerably reduced, which is described in Fig. 11. Here, the red line indicates after optimization, and the green line represents B.O; the recorded delay measure is detailed in Fig. 11. The designed WSN nodes are movable, so the fluctuation is recorded during the data transmission process. The overall outcome for the 200 rounds iteration is tabulated in Table 3.

Comparative analysis
To measure the improvement score of the presented model, some recent related works have been considered, such as LEACH, Genetic optimization Leach (GOLEACH) [26], clustering LEACH [27], and EERP [27].

Energy consumption
This comparative analysis has been performed to calculate the energy consumption rate depreciation. Hence, the energy usage is validated by Eq. (8).
Here, E t is the recorded total energy consumption for the data transmission, E r is the consumed energy during data receiving process, and the waiting state energy utilization rate is determined as E w Hence, by taking the energy utilization rate of each process, the total energy utilization rate was measured.
The parameter energy utilization has been measured in dual phases before and after applying the buffalo fitness. Hence, after 200 rounds, the recorded energy usage is 0.07 J for optimized LEACH and 0.09 without using the buffalo functions, as shown in Fig. 12. There are several sensor hubs in the WSN, some are in free status, and some are in work status. Moreover, the free status node has  consumed some optimal energy that is considered energy dissipation or energy wastage. Hence, the parameter energy dissipation rate has been measured to know the unwanted energy consumption. Also, to find the energy usage rate of the work status hubs, the parameter energy consumption has been measured.

Throughput ratio and communication delay
By measuring the time utilization of each process, the average communication delay was recorded. Here, T t denotes the total time for packet transmission, Here, T t denoted the entire time for the communication broadcasting process, T r represents the time utilization for the data receiving process T w , and determines the time for the waiting process. Hence, the delay is measured by Eq. (9).
The communication range of the WSN medium has been determined by Throughput validation. Moreover, the parameter throughput is measured by initializing the data sharing functions. In addition, the throughput metric was measured using Eq. (10).
The model that gained the most acceptable range of throughput scores will have the finest data transmission rate. Hence, the presented model has earned the throughput measure 10100 Kbps after applying the buffalo fitness and 4037 Kbps recorded before incorporating the Buffalo fitness. Also, for the BDBEMF (A.O.) recorded delay is 1.84 s, and for BDBEMF (B.O.), the gained delay is 7.3 s. These statistics are structured in Fig. 13.
The alive node count has been validated to measure the successive score of the designed model.
Here, the presented novel BDBEMF has recorded the full live node as 166 after completing 200 rounds. Before applying the buffalo fitness, the registered alive nodes are 34 for 200 games. In addition, the existing scheme C-LEACH has gained the actual node count of 24 for 200 rounds, and the model EERP has attained 60 actual nodes. Considering these approaches, the proposed model contains more alive nodes after complete iteration rounds, around 166. The comparison assessment is exposed in Fig. 14.
Before applying the optimization, high energy consumption has been recorded, indicating more dead nodes. In WSN, the increase of dead nodes has degraded the network performance by resulting in high energy consumption. For this reason, the optimal function has been incorporated into this LEACH protocol. It has verified the need for the optimal model in the WSN environment.
Based on the percentage of the live node, the lifetime of WSN has been determined. Here, after applying the optimization, the proposed model has recorded 400 s as network lifetime; before applying the buffalo model, only  300 s had been recorded as network lifetime. Moreover, the comparison of network lifetime is defined in Fig. 15. For the energy management application, estimating the residual energy is important to check the proposed model's optimal status and optimization performance. Hence, residual energy has been measured for two scenarios: residual network energy and nodes residual energy. After applying the buffalo algorithm, a suitable outcome has been gained, which is 27 J of nodes residual energy and 50 J network residual energy. These statistics are described in Fig. 16.

Discussion
The novel BDBEMF has earned the finest outcome in all performance assessments that have verified the presented model's robustness. The key merit of this designed model is dead node reduction. The proposed model has considerably reduced the dead nodes and maximized the actual node's rate. This procedure has helped to minimize the packet flow rate and diminished energy usage. Hence, the packet drop measures are figured in Fig. 17.
Moreover, the overall outcome of the implemented approach is tabulated in the table. Thus, the performance results have proved the robustness score of the presented model, and it is suitable for WSN applications to optimize energy utilization. In addition, all performance metrics have revealed the finest performance for A.O. because before applying the buffalo algorithm, the cluster head selection and load balancing have been performed in a usual manner. But after performing the optimal buffalo function, the process has been iterated repeatedly to reach the optimal status. When the fixed optimal status has been reached, the process gets stopped. This is the reason for gaining the finest outcome after applying the optimal function. The overall finest outcome of the designed scheme is tabulated in Table 4. Significantly, the recorded energy dissipation rate is meager, 0.05 nJ, in a negligible state. So, it has more residual energy, which is about 8 J, after completing 200 rounds. The optimal data overhead rate has been fixed as 0.6 Mbps; during the data transmission, if the fixed data rate has been crossed, then the migration process has been initiated, and half the data of the particular hub has been shared to another less data rate hub. This way, the collision has been handled by the proposed model. When a collision happens, the optimal energy process is activated. Hence, the buffalo algorithm has been chosen to provide continuous monitoring features at an optimal state.      To check the reliability of the designed model, one of the standard datasets has been considered, and the performance has been measured. Those details are described in Table 5. The proposed model has recorded the finest outcome in all metrics. It verifies the need for the proposed model in the WSN application. Also, the proposed model has recorded a minimum delay rate of 2 s, and the time taken for the execution is 3 s. It reduces the computational cost.

Conclusion
The novel BDBEMF has been designed for the WSN application to manage the energy utilization rate. In addition, the LEACH protocol is considered to enable the communication process. Finally, the planned model is tested in the MATLAB platform. The efficiency rate has been validated regarding dead and alive nodes, throughput, packet drop, energy consumption, and wastage. Compared to other models, the novel BDBEMF has earned a high throughput score of 10100Kbps. It has improved the throughput rate up to 30%. Moreover, the reduced delay score yielded by the designed model is 1.84 s. Compared to other models, it minimized the delay rate by 3 s. In addition, the recorded energy consumption score by the developed model is 0.07 J; compared to normal LEACH, it has reduced the energy usage rate by 2%. Also, the recorded network residual energy is 50 J; it has maximized the residual energy by 8%. The lifetime of the designed WSN environment is 400 s; compared to other models, 50% of the lifetime has been increased. Hence, the presented model is appropriate for the WSN application for the energy management process. However, the security features are not implemented in this proposed model. In the future, incorporating the security features with this proposed model will give a better network performance outcome by affording the highest privacy range.