Wireless sensor network cluster head selection method and system based on energy consumption rate prediction

Abstract

The invention discloses a wireless sensor network cluster head selection method and system based on energy consumption rate prediction, and the method comprises the steps: carrying out the operation with a wheel as a unit after the network is initialized and a chain topology is constructed; the node energy state is collected at the beginning of each round; virtually evaluating the energy consumption of the whole network when each survival node is used as a cluster head; predicting a next round of energy consumption rate of the node based on historical data; calculating the variance of the whole network energy consumption rate corresponding to each candidate cluster head; and selecting the node with the minimum variance to serve as a cluster head, and carrying out data transmission and energy updating. The method breaks through the limitation of static cluster head selection, remarkably prolongs the service life of the network, inhibits energy holes, and improves the stability of data transmission through the prospective energy trend prediction and variance minimization criterion.

Description

A Method and System for Cluster Head Selection in Wireless Sensor Networks Based on Energy Consumption Rate Prediction Technical Field

[0001] This invention relates to the field of wireless sensor network technology, and in particular to a method and system for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction. Background Technology

[0002] Wireless Sensor Networks (WSNs) consist of a large number of resource-constrained sensor nodes and are widely used in scenarios such as environmental monitoring, smart cities, industrial IoT, and disaster early warning. In most applications, sensor nodes are typically deployed in environments where batteries are difficult to replace or cannot be replenished. Therefore, how to extend the overall network uptime and improve data transmission stability under limited energy conditions has become a key technical issue in the research and application of wireless sensor networks.

[0003] In existing wireless sensor network routing technologies, chain-based data collection protocols have attracted widespread attention due to their ability to reduce node communication frequency and lower overall energy consumption. As shown in Figure 1, the existing PEGASIS (Power-Efficient Gathering in Sensor Information Systems) protocol constructs a single chain topology, allowing nodes to communicate only with their neighboring nodes. The cluster head node on the chain is responsible for transmitting the aggregated data to the sink node, which to some extent improves the energy consumption overhead caused by frequent cluster reconfiguration in traditional clustering protocols. However, existing PEGASIS and its improved schemes still have significant shortcomings in practical applications. Specifically, most existing schemes mainly rely on the node rotation order, remaining energy, or distance between the node and the sink node to make decisions during the cluster head selection process, which is a static or real-time judgment mechanism. Due to significant differences in the node's location in the network, link structure, and distance from the sink node, even with the same remaining energy, the energy consumption rate borne by different nodes when acting as cluster heads may still vary greatly. This difference often leads to some nodes rapidly depleting their energy in a short period of time, resulting in high-energy-consuming nodes and energy gaps, ultimately causing premature network failure.

[0004] Furthermore, most existing technologies only focus on the energy state of the current cycle, lacking the ability to predict future energy consumption trends of nodes, and thus failing to effectively control the long-term energy balance of the network from a global perspective. Especially in large-scale wireless sensor networks or scenarios where the aggregation node is located outside the network, traditional cluster head selection strategies struggle to balance energy balance and network stability, severely limiting further improvements in network lifetime.

[0005] Therefore, there is an urgent need for a cluster head selection method that can comprehensively consider node energy consumption behavior, predict future energy consumption trends, and achieve energy consumption balance from the perspective of the entire network, so as to overcome the above-mentioned defects in the existing technology. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction. By introducing the dynamic indicator of energy consumption rate, the future energy consumption trend of nodes is predicted and used as the basis for cluster head selection, thereby achieving long-term balance of node energy consumption in wireless sensor networks.

[0007] The technical solution adopted in this invention is:

[0008] A method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction includes the following steps:

[0009] Initialize the wireless sensor network: Randomly deploy two or more sensor nodes within the monitoring area to construct a chain-like topology wireless sensor network, and configure initial energy for each sensor node;

[0010] Round-based operation: Data transmission is carried out in rounds. At the beginning of each round, each node reports the remaining energy and historical energy consumption data.

[0011] Virtual evaluation of candidate cluster heads: Calculate the energy consumption of all sensor nodes in the network when each surviving sensor node is used as the virtual current round cluster head;

[0012] Energy consumption rate prediction: Based on the historical energy consumption data of sensor nodes, predict the energy consumption rate of each sensor node in the next round. , ;in, Δe represents the total energy consumption of the previous round, and Δe represents the energy consumption of the virtual cluster head.

[0013] Energy consumption variance calculation: Calculate the variance of the total network energy consumption rate corresponding to each candidate cluster head. ;

[0014] Cluster head selection: Selection that minimizes variance The smallest sensor node is used as the current wheel cluster head;

[0015] Data transmission and updates: The cluster head sensor nodes aggregate data and transmit it to the sink node, update the remaining energy of each sensor node, and execute the round operation after incrementing the round number.

[0016] Furthermore, each sensor node communicates only with its neighboring sensor nodes;

[0017] Furthermore, the chain topology is constructed using a greedy algorithm, and sensor node connections are based on the nearest neighbor principle.

[0018] Furthermore, each sensor node has the same initial energy and a fixed position.

[0019] Furthermore, the energy consumption of all sensor nodes in the network includes the energy consumption of cluster head nodes and non-cluster head nodes; the energy consumption of cluster head nodes includes data reception energy consumption, data fusion energy consumption and energy consumption for transmission to the aggregation node, while the energy consumption of non-cluster head nodes only includes the energy consumption for forwarding to the next hop node.

[0020] Furthermore, the energy consumption Δe of the virtual cluster head in the energy consumption rate prediction is calculated based on the radio model, which distinguishes between free space fading and multipath fading, and the transmission energy consumption is related to the square or fourth power of the distance.

[0021] In summary, the first Wheel with sensor nodes When acting as a cluster head, the energy consumption of all sensor nodes in the network is expressed as follows:

[0022] (1) When The energy consumption of the cluster head node is expressed as:

[0023] (12);

[0024] (2) When The energy consumption of the non-cluster head node is expressed as:

[0025] (13);

[0026] in, For sensor nodes In the The total energy consumption of the first round is zero, and it should be noted that the total energy consumption of the sensor nodes corresponding to the previous round is zero. It is the energy consumption for receiving data, a constant value that does not change with distance; For the first Wheel as sensor node of cluster head The energy consumed in sending data; For sensor nodes In the Total energy consumption of the wheel; Indicates the first Wheel sensor node When acting as a cluster head, the sensor node The energy consumed in sending data; Let N be the number of sensor nodes, where sensor node 1 is the cluster head and sensor node N is the cluster tail.

[0027] in, (11);

[0028] in, Represents sensor nodes The distance to the next hop node, where k represents the size of the data packet; This refers to the power consumption per bit of the electronic circuitry at both the transmitting and receiving ends. The transmit amplifier power consumption under the free-space fading model;

[0029] Furthermore, the energy consumption rate of each sensor node in the next round The calculation formula is:

[0030] ;

[0031] in, Let Δe be the total energy consumption of the previous round, and Δe be the energy consumption of the first round. In-wheel sensor nodes Energy consumption when using a virtual cluster head.

[0032] Furthermore, the variance of the total network energy consumption rate corresponding to each candidate cluster head. The calculation formula is:

[0033] ,

[0034] in, Let be the energy consumption rate of sensor node j when sensor node i acts as the cluster head.

[0035] Furthermore, for large-scale wireless sensor networks with N>50 sensors, the K-means clustering algorithm is used to divide the network into at least two partitions. The variance of each partition is then calculated, and a cluster head is selected for each partition. That is, when the number of sensors N>50, the number of partitions k≥2 to reduce computational complexity.

[0036] A cluster head selection system for wireless sensor networks based on energy consumption rate prediction includes:

[0037] At least two sensor nodes are randomly deployed within the monitoring area to form a wireless sensor. The sensor nodes have sensing, data fusion and wireless communication capabilities.

[0038] The aggregation node, located outside the wireless sensor network, is responsible for receiving data transmitted by the sensor nodes;

[0039] The processing module is used to execute the cluster head selection algorithm, including an energy state collection unit, a candidate cluster head evaluation unit, an energy consumption rate prediction unit, a variance calculation unit, and a cluster head selection unit.

[0040] The storage module is used to record historical energy data of the sensor nodes;

[0041] The system is configured to operate in a wheel-based manner, dynamically selecting cluster heads to achieve balanced energy consumption.

[0042] Furthermore, the energy state harvesting unit is used to collect the remaining energy data of the corresponding sensor node at the end of the previous round from each surviving sensor node at the beginning of each round. The data includes actual energy consumption data ΔE(r−1) and provides a basic input for subsequent energy consumption rate prediction; energy state data is transmitted wirelessly to the sink node or distributed and stored locally on the sensor node.

[0043] The candidate cluster head evaluation unit is used to calculate the energy consumption of all sensor nodes in the network in each round when each surviving sensor node is used as the virtual current round cluster head;

[0044] The energy consumption rate prediction unit is used to predict the energy consumption rate of each node in the next round based on historical energy data and candidate cluster head evaluation results. ;

[0045] The variance calculation unit is used to calculate the statistical variance of the energy consumption rate of all surviving sensor nodes in the network for each candidate cluster head node. ;

[0046] The cluster head selection unit is used to compare the variances of all candidate cluster head nodes. Select the sensor node that minimizes the variance as the current wheel cluster head;

[0047] Furthermore, the processing module is integrated into the aggregation node, which centrally executes the cluster head selection algorithm; or it is distributed among the sensor nodes, and cluster head selection is completed through inter-node collaboration.

[0048] The present invention, by adopting the above technical solution, has the following beneficial effects compared with the prior art:

[0049] 1. By analyzing the energy consumption of nodes in historical rounds, the potential energy consumption rate of nodes in subsequent rounds can be predicted, avoiding decision-making biases caused by selecting cluster heads solely based on current remaining energy. 2. With minimizing the dispersion of energy consumption rates across the entire network as the optimization objective, the node most conducive to maintaining overall network energy balance is selected as the cluster head in each round, suppressing the generation of high-energy-consuming nodes and energy gaps. 3. By balancing the energy consumption trends of each node, the failure time of the first node is significantly delayed, improving network stability and overall lifespan, and enhancing the reliability of long-term network operation. 4. This invention is applicable to wireless sensor networks with chain-based routing structures and can be combined with existing routing construction mechanisms, possessing good versatility and engineering implementation value.

[0050] This invention can effectively overcome the problems of uneven energy consumption, premature node failure, and lack of energy consumption trend prediction in existing wireless sensor network cluster head selection technology, and achieve efficient and stable operation of wireless sensor networks. Attached Figure Description

[0051] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments;

[0052] Figure 1 is a schematic diagram of the existing PEGASIS network model structure;

[0053] Figure 2 is a schematic diagram of the network model structure of P-PEGASIS of the present invention;

[0054] Figure 3 is a schematic diagram of the radio model structure;

[0055] Figure 4 is a schematic diagram showing the location of the communication destination in the node region according to the present invention;

[0056] Figure 5 is a schematic diagram comparing the survival time of the P-PEGASIS of the present invention with that of existing PEGASIS, RPC, and EB-PEGASIS-SCH.

[0057] Figure 6 is a schematic diagram comparing the remaining energy of the P-PEGASIS of the present invention with that of existing PEGASIS, RPC, and EB-PEGASIS-SCH.

[0058] Figure 7 is a schematic diagram comparing the throughput of the P-PEGASIS of the present invention with that of existing PEGASIS, RPC, and EB-PEGASIS. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0060] As shown in any one of Figures 2 to 7, this invention discloses a method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction, comprising the following steps:

[0061] Initialize the wireless sensor network: Randomly deploy two or more sensor nodes within the monitoring area to construct a chain-like topology wireless sensor network, and configure initial energy for each sensor node;

[0062] Round-based operation: Data transmission is carried out in rounds. At the beginning of each round, each node reports the remaining energy and historical energy consumption data.

[0063] Virtual evaluation of candidate cluster heads: Calculate the energy consumption of all sensor nodes in the network when each surviving sensor node is used as the virtual current round cluster head;

[0064] Energy consumption rate prediction: Based on the historical energy consumption data of sensor nodes, predict the energy consumption rate of each sensor node in the next round. , ;in, Δe represents the total energy consumption of the previous round, and Δe represents the energy consumption of the virtual cluster head.

[0065] Energy consumption variance calculation: Calculate the variance of the total network energy consumption rate corresponding to each candidate cluster head. ;

[0066] Cluster head selection: Selection that minimizes variance The smallest sensor node is used as the current wheel cluster head;

[0067] Data transmission and updates: The cluster head sensor nodes aggregate data and transmit it to the sink node, update the remaining energy of each sensor node, and execute the round operation after incrementing the round number.

[0068] Furthermore, each sensor node communicates only with its neighboring sensor nodes;

[0069] Furthermore, the chain topology is constructed using a greedy algorithm, and sensor node connections are based on the nearest neighbor principle.

[0070] Furthermore, each sensor node has the same initial energy and a fixed position.

[0071] Furthermore, the energy consumption of all sensor nodes in the network includes the energy consumption of cluster head nodes and non-cluster head nodes; the energy consumption of cluster head nodes includes data reception energy consumption, data fusion energy consumption and energy consumption for transmission to the aggregation node, while the energy consumption of non-cluster head nodes only includes the energy consumption for forwarding to the next hop node.

[0072] Furthermore, the energy consumption Δe of the virtual cluster head in the energy consumption rate prediction is calculated based on the radio model, which distinguishes between free space fading and multipath fading, and the transmission energy consumption is related to the square or fourth power of the distance.

[0073] In summary, the first Wheel with sensor nodes When acting as a cluster head, the energy consumption of all sensor nodes in the network is expressed as follows:

[0074] (1) When The energy consumption of the cluster head node is expressed as:

[0075] (12);

[0076] (2) When The energy consumption of the non-cluster head node is expressed as:

[0077] (13);

[0078] in, For sensor nodes In the The total energy consumption of the first round is zero, and it should be noted that the total energy consumption of the sensor nodes corresponding to the previous round is zero. It is the energy consumption for receiving data, a constant value that does not change with distance; For the first Wheel as sensor node of cluster head The energy consumed in sending data; For sensor nodes In the Total energy consumption of the wheel; Indicates the first Wheel sensor node When acting as a cluster head, the sensor node The energy consumed in sending data; Let N be the number of sensor nodes, where sensor node 1 is the cluster head and sensor node N is the cluster tail.

[0079] in, = ,and Let i be the distance from node i to the sink.

[0080] (11);

[0081] in, Represents sensor nodes The distance to the next hop node, where k represents the size of the data packet; This refers to the power consumption per bit of the electronic circuitry at both the transmitting and receiving ends. The transmit amplifier power consumption under the free-space fading model;

[0082] Furthermore, the energy consumption rate of each sensor node in the next round The calculation formula is:

[0083] ;

[0084] in, This is the total energy consumption of the previous round. Indicates the estimation of the first wheel node The energy consumed when node i is the cluster head includes the energy consumed by other nodes when node i receives energy from them. and the energy consumed in transmitting information to the sink ; .

[0085] Furthermore, the variance of the total network energy consumption rate corresponding to each candidate cluster head. The calculation formula is:

[0086] ,

[0087] in, Let be the energy consumption rate of sensor node j when sensor node i acts as the cluster head.

[0088] Furthermore, for large-scale wireless sensor networks with N>50 sensors, the K-means clustering algorithm is used to divide the network into at least two partitions. The variance of each partition is then calculated, and a cluster head is selected for each partition. That is, when the number of sensors N>50, the number of partitions k≥2 to reduce computational complexity.

[0089] A cluster head selection system for wireless sensor networks based on energy consumption rate prediction includes:

[0090] At least two sensor nodes are randomly deployed within the monitoring area to form a wireless sensor. The sensor nodes have sensing, data fusion and wireless communication capabilities.

[0091] The aggregation node, located outside the wireless sensor network, is responsible for receiving data transmitted by the sensor nodes;

[0092] The processing module is used to execute the cluster head selection algorithm, including an energy state collection unit, a candidate cluster head evaluation unit, an energy consumption rate prediction unit, a variance calculation unit, and a cluster head selection unit.

[0093] The storage module is used to record historical energy data of the sensor nodes;

[0094] The system is configured to operate in a wheel-based manner, dynamically selecting cluster heads to achieve balanced energy consumption.

[0095] Furthermore, the energy state harvesting unit is used to collect the remaining energy data of the corresponding sensor node at the end of the previous round from each surviving sensor node at the beginning of each round. The data includes actual energy consumption data ΔE(r−1) and provides a basic input for subsequent energy consumption rate prediction; energy state data is transmitted wirelessly to the sink node or distributed and stored locally on the sensor node.

[0096] The candidate cluster head evaluation unit is used to calculate the energy consumption of all sensor nodes in the network in each round when each surviving sensor node is used as the virtual current round cluster head;

[0097] The energy consumption rate prediction unit is used to predict the energy consumption rate of each node in the next round based on historical energy data and candidate cluster head evaluation results. ;

[0098] The variance calculation unit is used to calculate the statistical variance of the energy consumption rate of all surviving sensor nodes in the network for each candidate cluster head node. ;

[0099] The cluster head selection unit is used to compare the variances of all candidate cluster head nodes. Select the sensor node that minimizes the variance as the current wheel cluster head;

[0100] Furthermore, the processing module is integrated into the aggregation node, which centrally executes the cluster head selection algorithm; or it is distributed among the sensor nodes, and cluster head selection is completed through inter-node collaboration.

[0101] This invention is the first to introduce energy consumption rate into the PEGASIS cluster head selection mechanism, predicting the energy load a node may bear in the next round by using historical energy consumption information, thus overcoming the limitations of traditional static strategies that rely solely on remaining energy. This invention aims to minimize the variance of energy consumption rate, achieving long-term balance of node energy consumption from a global perspective and effectively suppressing the energy void problem. Through energy consumption trend equalization design, this invention significantly delays the occurrence of the first dead node and improves network stability and data throughput performance.

[0102] The specific principles of this invention will be explained in detail below:

[0103] To overcome the energy imbalance problem caused by improper cluster head selection in traditional PEGASIS, this invention proposes P-PEGASIS, which employs a cluster head selection machine based on balanced energy consumption rate. As shown in Figure 2, in P-PEGASIS, nodes no longer select cluster heads solely based on turn order or remaining energy, but instead introduce energy consumption rate as a key decision indicator. Based on the actual energy consumption of a node in past rounds, the potential energy consumption change if it becomes a cluster head in the next round is predicted, thereby assessing the node's impact on the overall network energy distribution. For ease of explanation, this invention conceptually classifies nodes into three categories based on their energy consumption rate: high-energy-consuming nodes: significantly high energy consumption rate (common in traditional PEGASIS, as shown by the red nodes in Figure 1); medium-energy-consuming nodes: energy consumption rate at a moderate level (as shown by the yellow nodes in Figure 1 or 2); and low-energy-consuming nodes: relatively low energy consumption rate (as shown by the blue nodes in Figure 1 or 2). In P-PEGASIS, the goal of cluster head selection is not low energy consumption for a single node, but rather to make the energy consumption rate distribution of all nodes as balanced as possible. Therefore, by minimizing the variance of the overall node energy consumption rate, P-PEGASIS selects the node that best maintains the energy balance of the entire network as the cluster head in each round.

[0104] As shown in Figure 2, the wireless sensor network of this invention no longer contains obvious high-energy-consuming nodes; the entire network consists only of medium- and low-energy-consuming nodes. This ensures that the energy consumption trends of each node are consistent, effectively delaying the occurrence of the first dead node and significantly reducing the probability of energy voids. P-PEGASIS uses the same energy transfer model as the PEGASIS protocol. As shown in Figure 3, the transmitter consumes energy to drive the amplifier circuit and transmit data, while the receiver consumes energy when receiving signals. If k-bit information is transmitted in a space of length d, the energy consumption for wireless transmission and reception are as follows:

[0105] (1);

[0106] (2);

[0107] in, This represents the energy dissipation per bit used to operate the electronic circuitry of the transmitter and receiver. and These represent the transmitter amplifier power consumption under free-space fading and multipath fading, respectively. It is the distance threshold between the two fading models. .

[0108] Before selecting the cluster head, this invention first calculates the number of cluster heads according to the definition of formula (4). wheel node The energy consumption rate matrix of all nodes when the cluster head is in operation is denoted as follows: .

[0109] (5);

[0110] In equation (5), Represents the node in the r-th round. When acting as a cluster head, the node The rate of energy consumption. According to equation (4):

[0111] (6);

[0112] In equation (6), Represents the node in the r-th round. When acting as a cluster head, the node The energy consumed is:

[0113] (7);

[0114] In equation (7), Indicates the first wheel node Total energy consumed Indicates the first wheel node When acting as a cluster head, the node The energy consumed in receiving data from neighboring nodes; Indicates the first wheel node When acting as a cluster head, the node The energy consumed in sending data. Of which,

[0115] (8);

[0116] In equation (8), k represents the size of the data packet. This is the energy consumption for receiving data, a constant that does not change with distance. Assume the head of the chain is numbered 1, and the tail of the chain is numbered N. Represents a node The position coefficient.

[0117] (1) When hour, (9);

[0118] (2) When hour, (10);

[0119] It can be represented as: (11);

[0120] In equation (11), Represents a node Distance to the next hop node.

[0121] In summary, the first Wheel, node When acting as a cluster head, the node The energy consumed can be expressed as:

[0122] (1) When hour, (12);

[0123] (2) When hour, (13);

[0124] Equations (12) and (13) respectively describe the first The energy consumed by a node when it is a cluster head and when it is not a cluster head, and the energy consumed by other nodes. Indicates the first wheel node The energy consumed to send data to the sink; Indicates the first Wheel at the node When acting as a cluster head, the node The energy required to send data to the next hop node. It can be seen that... and , , or Proportional and It is directly proportional. Therefore, also with , , or Proportional.

[0125] To ensure that the energy consumption of other nodes is as balanced as possible when a node acts as a cluster head, this invention uses variance to measure the overall energy balance among the nodes.

[0126] (14);

[0127] In equation (14), Indicates the first Wheel, node The variance of the energy consumption rate of all nodes when acting as cluster head. Calculate the energy consumption rate of all nodes within the network when acting as cluster head. The minimum value is found, which corresponds to the cluster head node. For networks with a small number of nodes, the minimum value can be directly calculated to select the corresponding cluster head node. For networks with a large number of nodes, the K-means clustering algorithm can be used to partition the network first, calculate the variance of the energy consumption rate within each cluster, and compare the variances of all clusters. The cluster with the smallest variance value is selected as the cluster head node.

[0128] The cluster head selection process for estimating energy consumption rate balancing is as follows:

[0129] The P-PEGASIS algorithm uses rounds as its basic operating unit. The overall process can be divided into an initialization phase, a candidate cluster head evaluation phase, a cluster head selection phase, and a data transmission and energy update phase. Its core idea is that in each round, by predicting the distribution of the network's energy consumption rate when different nodes act as cluster heads, the algorithm selects the node that best balances the energy consumption rate, thereby avoiding the generation of locally high-energy-consuming nodes.

[0130] During network initialization, all sensor nodes are randomly and uniformly deployed within an L×L square monitoring area. All nodes have the same initial energy 𝐸0 and remain stationary after deployment. The sink node is located at a fixed position outside the monitoring area and possesses unlimited energy and computing power.

[0131] Subsequently, the network employs a greedy algorithm from the PEGASIS protocol to construct a chain topology. Specifically, starting with the node closest to the sink, it progressively connects its nearest unconnected neighbors until all nodes are included in the same chain. Once established, the chain topology remains unchanged during network operation, with only the cluster head role of a node changing with each round.

[0132] At the start of round 3, energy status is collected, and each surviving node records and reports its remaining energy at the end of the previous round. And the energy actually consumed in the previous round. This information provides the foundational data for subsequent energy consumption rate prediction. It should be noted that, unlike traditional PEGASIS, P-PEGASIS does not immediately determine the current cluster head, but instead first proceeds to the evaluation process of candidate cluster heads.

[0133] During the candidate cluster head evaluation phase, the algorithm sequentially assumes that each surviving node could potentially become the cluster head in the current round, and evaluates its impact on the overall network energy consumption balance under this assumption. For a candidate node *a*, the algorithm assumes it will act as the cluster head in round *l*, while all other surviving nodes will act as non-cluster head nodes, forwarding data along the link. Under this assumption, the algorithm calculates the energy consumption of the two types of nodes based on the wireless energy consumption model: 1) Energy consumption of cluster head nodes: Candidate cluster head nodes need to complete three tasks: receiving data from neighboring nodes on the link, merging the received data, and transmitting the merged data to the sink over long distances. Therefore, their energy consumption is usually significantly higher than that of ordinary forwarding nodes. 2) Energy consumption of non-cluster head nodes: Non-cluster head nodes only need to send their own data or merged data to the next-hop node according to the chain topology, and receive data from neighboring nodes when necessary. Their energy consumption is mainly determined by short-range communication. Based on this, the energy consumption of all nodes in round *l* under the assumption that node *a* is the cluster head can be obtained. .

[0134] Energy Consumption Rate Prediction and Variance Calculation: After obtaining the energy consumption estimates for each node, P-PEGASIS further introduces the dynamic indicator of energy consumption rate. For each node 𝑗, its energy consumption rate when the candidate cluster head is 𝑖 is defined as the energy consumption level within a unit round, used to characterize the rate of energy decline of the node. Subsequently, the algorithm calculates the statistical variance of the energy consumption rates of all surviving nodes in the entire network when node 𝑖 is the cluster head. This variance value... This reflects the dispersion of node energy consumption rates: a larger variance indicates that there are nodes in the network with significantly higher or lower energy consumption, resulting in an uneven energy distribution; a smaller variance indicates that the energy consumption trends of each node are more consistent, and the network energy distribution is more uniform. By repeating the above process for all candidate nodes, the algorithm can obtain the variance of the energy consumption rate corresponding to each node when it is a cluster head.

[0135] Cluster head selection strategy: After evaluating candidate cluster heads, P-PEGASIS selects the node with the smallest energy consumption rate variance from all surviving nodes as the cluster head for the current round. This selection criterion is essentially a global energy balance optimization strategy; its goal is not to minimize the energy consumption of a single node, but to make the energy consumption rate distribution of the entire network as flat as possible. This mechanism effectively avoids the problem in traditional PEGASIS where distant nodes rapidly deplete their energy due to taking turns serving as cluster heads, thereby suppressing the generation of high-energy-consuming nodes and energy gaps.

[0136] Data transmission, energy update, and round progression: After the cluster head is determined, the network enters the actual data transmission phase. All non-cluster head nodes forward data hop-by-hop along the predetermined links. The cluster head node merges the received data and sends the merged data to the Sink. After data transmission is complete, the algorithm updates the remaining energy of each node according to the wireless power consumption model. If the remaining energy of a node is less than or equal to zero, the node is marked as dead and will not participate in cluster head evaluation and data transmission in subsequent rounds. Finally, the round counter is incremented, and the algorithm enters the next round until the network meets the termination condition (e.g., all nodes are dead or the maximum number of rounds is reached).

[0137] This invention investigates the network performance of the next-round energy-rate balanced cluster head selection algorithm (P-PEGASIS) from two aspects. First, the P-PEGASIS algorithm is compared with PEGASIS, RPC, and EB-PEGASIS-SCH algorithms to evaluate its performance in different scenarios. Second, the P-PEGASIS algorithm is combined with EB-PEGASIS; that is, based on the shortest chain structure obtained by EB-PEGASIS, the P-PEGASIS algorithm is introduced for cluster head selection, and the results are compared with those of the original EB-PEGASIS to verify the advantages of P-PEGASIS in cluster head selection. Throughout the simulation, performance evaluation metrics include network lifetime, overall energy consumption, and the number of packets successfully reaching the sink.

[0138] In the simulation scenario, 100 nodes are randomly distributed in a 100m × 100m area. As shown in Figure 4, the sink is located at coordinates (50, 200) 100 meters away from the node area. Specific simulation-related parameter settings are shown in Table 1.

[0139] Table 1 - Simulated Wireless Sensor Network Parameter Settings

[0140]

[0141] To comprehensively evaluate the performance of the P-PEGASIS algorithm (represented by SH, Selection Head in the figure), this invention compares and analyzes it with PEGASIS, RPC, and EB-PEGASIS-SCH. Evaluation metrics include network lifetime, node remaining energy, node energy consumption, and throughput. To ensure the generalizability of the results, 11 different random network scenarios were constructed, and the average performance of the four algorithms was compared.

[0142] Figure 5 shows the network lifetime performance of four algorithms, where lifetime is defined as the number of rounds in which the first dead node appears. The results show that the P-PEGASIS algorithm only encounters its first dead node in round 1381, significantly outperforming the other algorithms. In contrast, the first dead nodes of EB-PEGASIS-SCH, PEGASIS, and RPC occur in rounds 1090, 865, and 635, respectively. Therefore, the lifetime of the P-PEGASIS algorithm is 1.23 times that of EB-PEGASIS-SCH, 1.60 times that of PEGASIS, and 2.17 times that of RPC. Furthermore, the final number of rounds for P-PEGASIS, EB-PEGASIS-SCH, PEGASIS, and RPC are 3752, 4054, 3765, and 3315, respectively. These results validate the significant advantage of P-PEGASIS in extending network lifetime.

[0143] The above phenomenon arises from the differences in cluster head selection mechanisms among different algorithms. PEGASIS uses a sequential, round-robin approach to select cluster heads. While this ensures that each node serves as a cluster head an equal number of times, the inconsistent distances from nodes to the sink lead to varying energy consumption, with nodes farther away more prone to premature death. RPC, an improvement on PEGASIS, divides the network into several regions and determines the next hop based on the polar angle from the node to the sink and the distance between them. Cluster heads are typically chosen from nodes closer to the sink. When the sink is located at the network center, the region division is relatively balanced, and the energy consumption for transmission from the cluster head to the sink is low. However, in the scenario described in this invention, where the sink is located outside the network, the distance from the cluster head to the sink increases significantly, resulting in excessively high energy consumption for each transmission and causing cluster head nodes to die prematurely. EB-PEGASIS-SCH introduces a fairness coefficient based on remaining energy to determine the cluster head. While this improves energy distribution to some extent, it still does not consider the dynamic characteristics of energy consumption and its rate in the next round. In contrast, the P-PEGASIS algorithm considers both the energy state of the next round and the energy consumption rate in cluster head selection, thus achieving a more balanced energy distribution. Therefore, P-PEGASIS is superior to the three algorithms mentioned above in its cluster head selection mechanism, resulting in more uniform node energy consumption and a significantly delayed occurrence of the first dead node.

[0144] Figure 6 further compares the performance of PEGASIS, RPC, EB-PEGASIS-SCH, and P-PEGASIS in terms of remaining energy distribution. In the experiment, when the first dead node appeared, all nodes were sorted in descending order of remaining energy percentage. The results show that P-PEGASIS has the most balanced remaining energy distribution, with 94% of nodes having less than 63% remaining energy. In contrast, some nodes in PEGASIS and RPC have as much as 82% remaining energy, with RPC exhibiting the largest fluctuation range, between 19% and 82%. Even the relatively good EB-PEGASIS-SCH has 93% of its nodes with more than 70% remaining energy. This indicates that during operation, P-PEGASIS can more effectively balance the energy consumption of each node, thereby improving the overall network stability and lifespan.

[0145] Figure 7 illustrates the throughput performance of four algorithms: PEGASIS, RPC, EB-PEGASIS-SCH, and P-PEGASIS. The results show that P-PEGASIS exhibits the fastest throughput growth and maintains stability for the longest period, followed by EB-PEGASIS-SCH, while RPC performs the worst. Because SH achieves relatively balanced node energy consumption, its throughput remains at a high level from the start of data transmission until approximately 3500 rounds. In contrast, although EB-PEGASIS-SCH has a slightly longer overall transmission time, its throughput growth slows down after approximately 1000 rounds due to the early appearance of the first dead node. PEGASIS uses a greedy algorithm to build links to reduce transmission distance, but the energy consumption of long-distance cluster heads is excessive and easily depleted; when a cluster head fails, the network needs to reconstruct the link, resulting in a step-like decrease in throughput. As for RPC, its first dead node appears earliest and it has the fewest transmission rounds, thus experiencing a slowdown in throughput growth early on and having the lowest overall throughput level.

[0146] Simulation results demonstrate that the protocol method of this invention exhibits significant advantages in multiple performance metrics: it significantly delays the occurrence time of the first dead node, improves the overall network lifetime, and outperforms comparative algorithms such as PEGASIS, RPC, and EB-PEGASIS-SCH in terms of node remaining energy distribution and data throughput. These results validate the comprehensive improvement effect of P-PEGASIS in terms of energy balance, robustness, and network stability.

[0147] This invention, employing the above technical solution, offers the following advantages compared to existing technologies: 1. By analyzing the energy consumption of nodes in historical rounds, it predicts the potential energy consumption rate of nodes in subsequent rounds, avoiding decision-making biases caused by selecting cluster heads solely based on current remaining energy. 2. With minimizing the dispersion of energy consumption rates across the entire network as the optimization objective, it selects the node most conducive to maintaining overall network energy balance as the cluster head in each round, suppressing the generation of high-energy-consuming nodes and energy gaps. 3. By balancing the energy consumption trends of each node, it significantly delays the failure time of the first node, improving network stability and overall lifespan, and enhancing the reliability of long-term network operation. 4. This invention is applicable to wireless sensor networks with chain-based routing structures and can be combined with existing routing mechanisms, possessing good versatility and engineering implementation value.

[0148] This invention can effectively overcome the problems of uneven energy consumption, premature node failure, and lack of energy consumption trend prediction in existing wireless sensor network cluster head selection technology, and achieve efficient and stable operation of wireless sensor networks.

[0149] Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Without conflict, the embodiments and features in the embodiments of this application can be combined with each other. The components of the embodiments of this application described and illustrated herein can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

Claims

1. A method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction, characterized in that: Includes the following steps: Initialize the wireless sensor network: Randomly deploy two or more sensor nodes within the monitoring area to construct a chain-like topology wireless sensor network, and configure initial energy for each sensor node; Round-by-round operation: Data transmission is performed in rounds. At the beginning of each round, each node reports its remaining energy and historical energy consumption data; Virtual evaluation of candidate cluster heads: Calculate the energy consumption of all sensor nodes in the network when each surviving sensor node serves as the virtual current cluster head; Energy consumption rate prediction: Based on the historical energy consumption data of sensor nodes, predict the energy consumption rate of each sensor node in the next round. Energy consumption variance calculation: Calculate the variance of the total network energy consumption rate corresponding to each candidate cluster head. Cluster head selection: Selecting cluster heads that minimize variance The smallest sensor node is used as the current wheel cluster head; Data transmission and updates: The cluster head sensor nodes aggregate data and transmit it to the sink node, update the remaining energy of each sensor node, and execute the round operation after incrementing the round number.

2. The method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction according to claim 1, characterized in that: The chain topology is constructed using a greedy algorithm, with each sensor node having the same initial energy and a fixed position; sensor node connections are based on the nearest neighbor principle, with each sensor node communicating only with its neighboring sensor nodes.

3. The method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction according to claim 1, characterized in that: The energy consumption of all sensor nodes in the network includes the energy consumption of cluster head nodes and non-cluster head nodes; the energy consumption of cluster head nodes includes data reception energy consumption, data fusion energy consumption and transmission energy consumption to the aggregation node, while the energy consumption of non-cluster head nodes only includes the energy consumption of forwarding to the next hop node.

4. The method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction according to claim 1, characterized in that: No. Wheel with sensor nodes When acting as a cluster head, the energy consumption of all sensor nodes in the network is expressed as: (1) When The energy consumption of the cluster head node is expressed as: (12); (2) when The time is the energy consumption of the non-cluster head node, expressed as: (13); among them, For sensor nodes In the The total energy consumption of the first round is zero, and it should be noted that the total energy consumption of the sensor nodes corresponding to the previous round is zero. It is the energy consumption for receiving data, a constant value that does not change with distance; For the first Wheel as sensor node of cluster head The energy consumed in sending data; For sensor nodes In the Total energy consumption of the wheel; Indicates the first Wheel sensor node When acting as a cluster head, the sensor node The energy consumed in sending data; Let N be the number of sensor nodes, where sensor node 1 is the cluster head and sensor node N is the cluster tail; where, (11); among which, Represents sensor nodes The distance to the next hop node, where k represents the size of the data packet; This refers to the power consumption per bit of the electronic circuitry at both the transmitting and receiving ends. This represents the power consumption of the emitter amplifier under the free-space fading model.

5. The method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction according to claim 1, characterized in that: The energy consumption rate of each sensor node in the next round The calculation formula is: ;in, Let Δe be the total energy consumption of the previous round, and Δe be the energy consumption of the first round calculated based on the radio model. In-wheel sensor nodes Energy consumption when using a virtual cluster head.

6. The method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction according to claim 5, characterized in that: The variance of the total network energy consumption rate corresponding to each candidate cluster head The calculation formula is: ,in, Let be the energy consumption rate of sensor node j when sensor node i acts as the cluster head.

7. The method for selecting cluster heads in wireless sensor networks based on energy consumption rate prediction according to claim 1, characterized in that: For large-scale wireless sensor networks with N>50 sensors, the K-means clustering algorithm is used to divide the network into at least two partitions. The variance of each partition is calculated and a cluster head is selected for each partition.

8. A wireless sensor network cluster head selection system based on power consumption rate prediction, comprising the wireless sensor network cluster head selection method based on power consumption rate prediction according to any one of claims 1 to 7, characterized in that: The system includes: At least two sensor nodes are randomly deployed within the monitoring area to form a wireless sensor network. The sensor nodes have sensing, data fusion, and wireless communication capabilities. A convergence node, located outside the wireless sensor network, is responsible for receiving data transmitted by the sensor nodes. A processing module is used to execute the cluster head selection algorithm, including an energy state collection unit, a candidate cluster head evaluation unit, an energy consumption rate prediction unit, a variance calculation unit, and a cluster head selection unit. A storage module is used to record historical energy data of the sensor nodes. The system is configured to operate in rounds, dynamically selecting cluster heads to achieve energy consumption balance.

9. The wireless sensor network cluster head selection system based on energy consumption rate prediction according to claim 8, characterized in that: The energy state collection unit is used at the beginning of each round to collect the remaining energy data of the corresponding sensor node at the end of the previous round from each surviving sensor node. The data includes actual energy consumption data ΔE(r−1) and provides a basic input for predicting the energy consumption rate; energy state data is transmitted wirelessly to the aggregation node or distributed and stored locally on the sensor node. The candidate cluster head evaluation unit is used to calculate the energy consumption of all sensor nodes in the network in each round when each surviving sensor node is used as the virtual cluster head for the current round; the energy consumption rate prediction unit is used to predict the energy consumption rate of each node in the next round based on historical energy data and candidate cluster head evaluation results. ； The variance calculation unit is used to calculate the statistical variance of the energy consumption rate of all surviving sensor nodes in the network for each candidate cluster head node. The cluster head selection unit is used to compare the variances of all candidate cluster head nodes. The sensor node that minimizes the variance is selected as the current wheel cluster head.

10. The wireless sensor network cluster head selection system based on energy consumption rate prediction according to claim 8, characterized in that: The processing module is integrated into the aggregation node, which centrally executes the cluster head selection algorithm; or it is distributed in the sensor nodes, and the cluster head selection is completed through inter-node cooperation.

Images