Tunnel dual-energy clustering routing method based on multi-objective optimization

Abstract

A tunnel dual-energy clustering routing method based on multi-objective optimization comprises the steps that a three-dimensional space model of a single-tube tunnel is constructed and used for representing the geometric structure of the tunnel; the method comprises the following steps of: providing a multi-target clustering routing (TCR) algorithm and a dual energy dynamic supply (DETCR) algorithm of the multi-target clustering routing (TCR) algorithm, and providing a multi-target clustering routing (DETCR) algorithm of the multi-target clustering routing (TCR) algorithm and a dual energy dynamic supply (DETCR) algorithm of the multi-target clustering routing (DETCR) algorithm of the multi-target clustering routing (DETCR) algorithm of the multi-target clustering routing (TCR) algorithm of the multi-target clustering routing (TCR) algorithm; on the basis of a tribe competition and member cooperation algorithm (CTCM), a multi-target tribe competition and member cooperation algorithm (MOCTCM) is provided, the residual energy, the communication distance and the energy consumption of nodes, the election frequency of relay nodes (RN) and other factors are comprehensively considered, the optimal cluster head (Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, Optimal Cluster Head, Optimal Relay Node, optima-RN is selected, so that a decision basis with higher adaptability and global optimization capability is provided. According to the method, the problems of non-uniform node energy consumption and limited energy supply in a dynamic and energy-limited tunnel environment are effectively solved, the life cycle of the network is remarkably prolonged, and the energy efficiency level of the network is improved.

Description

Technical Field

[0001] The present invention belongs to the field of Internet of Things and relates to a tunnel dual-energy clustering routing method based on multi-objective optimization. Background Art

[0002] Clustering routing algorithms in wireless sensor networks (WSNs) play a crucial role in improving network efficiency, extending network lifetime, and optimizing resource utilization. Existing research can be broadly categorized into two types, based on their distributed or centralized nature. Early representative algorithms, such as LEACH (Low-energy Adaptive Clustering Hierarchy) proposed by Heinzelman et al., employ a distributed architecture in which each node autonomously selects a communication channel (CH) based on a certain probability, aiming to reduce energy consumption and improve network lifetime. Although LEACH boasts high energy efficiency and low topology management overhead, it lacks global information and cannot effectively address the energy imbalance in CH selection. To address this issue, Heinzelman et al. further proposed the centralized routing algorithm LEACH-C (LEACH-centralized). This algorithm utilizes global energy information obtained by the base station and factors such as node distance to optimize CH selection, thereby achieving better energy balance and extending network lifetime.

[0003] Building on the LEACH algorithm, a growing number of improved algorithms have emerged. In particular, the introduction of intelligent optimization algorithms has made multi-objective optimization a key approach to improving cluster routing efficiency. For example, genetic algorithms, particle swarm optimization algorithms, butterfly optimization algorithms, and honey badger algorithms are used to balance objectives such as energy, distance, and load. However, these algorithms often simplify multiple objectives into a single objective through weighted summation, resulting in highly subjective weighting and ignoring conflicts between objectives. In recent years, multi-objective optimization algorithms have directly addressed the Pareto frontier, providing a more optimal solution set for clustering and routing selection. The DMaOWOA (distributed many-objective clustering using whale optimization algorithm) algorithm proposed by Kotary et al., through reference-point-based leader selection and whale optimization techniques, can provide superior clustering results in WSNs. It demonstrates superior performance compared to traditional distributed clustering algorithms on both synthetic and real datasets. However, its computational complexity increases with increasing network and data size. Zhang et al. proposed the MOALO-FCM (multi-objective antlion with fuzzy clustering algorithm) algorithm, combining fuzzy clustering with multi-objective antlion optimization to improve RN energy balance and effectively enhance the network lifetime, energy balance, and optimization stability of WSNs in both 2D and 3D environments. However, this approach still suffers from drawbacks such as high time complexity and low experimental scenario complexity. Singh et al. proposed a strategy combining multi-objective optimization with edge intelligent adaptation, using the Grey Wolf Optimizer (GWO) and bird-edge computation to generate Pareto optimal solutions to optimize QoS (Quality of Service) management for WSN-IoT applications, balancing competing objectives such as energy and latency, and improving resource utilization and scalability. Yuan et al. proposed a multi-objective routing (MOR) protocol based on NSGA-II to optimize energy consumption, end-to-end latency, link quality, and congestion control in underwater WSNs to simultaneously meet the diverse needs of underwater IoT (Internet of Things) applications. Simulation results validated its feasibility and the necessity of multi-objective optimization. Sun et al. proposed a secure routing protocol for WSNs based on multi-objective ant colony optimization. This protocol uses node residual energy and path trust as optimization targets, combines an improved DS (Dempster–Shafer) evidence theory with conflict preprocessing for node trust assessment, and optimizes multi-objective routing using a Pareto optimal solution mechanism. Simulation results show that this method can effectively counter black hole attacks. However, this method only considers limited objectives and constraints and does not address aspects such as network reliability, failure probability, and lifecycle optimization.The above studies have made significant progress in optimizing multiple objectives of WSN and IoT applications, but they mostly focus on optimization problems in two-dimensional or static environments, ignoring the complex constraints in three-dimensional space and the impact of dynamic energy supply on network performance.

[0004] In three-dimensional environments, the spatial layout, communication range, and topology of network nodes become more complex, placing higher demands on network coverage, connectivity, and energy efficiency. Furthermore, dynamic energy supply, especially in scenarios using renewable energy sources such as solar energy and vibration energy, poses additional challenges to network energy management due to the impact of environmental changes on energy supply. To address this issue, recent research has gradually focused on the impact of constraints and dynamic energy supply in three-dimensional environments on the performance of WSNs, and has proposed several corresponding optimization strategies. Yang et al. proposed a BWSN model based on a three-dimensional bridge space model (SM-SSB) and introduced a multi-objective CRITIC-TOPSIS clustering (CTC) routing algorithm to select the Optima-CH and Optima-RN. This algorithm comprehensively considers three-dimensional spatial distance, node density, and residual energy, significantly improving energy efficiency. Results show that the CTC algorithm outperforms traditional methods in three-dimensional environments. Zheng et al. proposed a multi-hop routing protocol based on the Secretary Bird Optimization Algorithm (SBOA) to optimize network lifetime and energy allocation in three-dimensional bridge wireless sensor networks (BWSNs). This protocol combines a fuzzy C-means algorithm for clustering and utilizes SBOA to find the optimal path between the CH and BS, while also improving energy efficiency through a re-clustering mechanism. Experimental results show that this protocol outperforms existing methods in terms of network lifetime, energy efficiency, and energy consumption balance. However, neither this method nor the CTC algorithm considers the energy supply issue. Zhang et al. designed an ultra-wide full-band gap phononic crystal (PnC) structure to improve the performance of piezoelectric energy harvesting (PEH) to power wireless sensors. In the same year, they also studied the impact of the size of incomplete line defects on the PnC-based PEH system to optimize energy positioning and collection performance. Results showed that when the defect is located in the fourth layer of the supercell, the system performance is optimal, with an output voltage of 22.54V and a power of 12.78mW, providing an effective energy harvesting solution for self-powered wireless sensors. Although the above studies have made important progress in optimizing PEH structures to improve energy harvesting performance, they mainly focus on material and structural design and do not combine the harvested energy with the actual application scenarios of wireless sensor nodes.To further promote the application of energy harvesting technology in WSNs, Lu et al. proposed a dynamic solar energy supply routing protocol (DSFOI-MMA) for field observation instrument networks (FOIN) based on a multi-objective mayfly optimization algorithm (MMA). This protocol ensures energy supply to low-energy nodes by designing a dynamic solar energy harvesting model and dynamically selects Optima-CHs using MMA. Simulation results show that DSFOI-MMA exhibits excellent performance in terms of network lifetime and number of CHs, enhancing the automation of monitoring data. However, this method is still based on a two-dimensional environment and does not consider topological constraints and uneven energy distribution in three-dimensional space. In practical applications, sensor node deployment in three-dimensional environments is more complex, and energy supply is significantly affected by spatial location. Therefore, further optimization of routing strategies is needed to adapt to the challenges of three-dimensional scenarios.

[0005] In recent years, with the research and application of tunnel structure health monitoring systems (TSHMS), intelligent tunnel management has gradually become possible. By deploying a variety of sensors (such as strain sensors, crack sensors, and temperature and humidity sensors), TSHMS collects real-time data on the tunnel structure's status, providing a scientific basis for tunnel safety assessments and early warnings. Since the 1990s, TSHMS has been widely used in major tunnel projects both domestically and internationally, such as the Xiamen Undersea Tunnel, the Hong Kong-Zhuhai-Macao Bridge Undersea Tunnel, the Nanjing Yangtze River Tunnel, and the English Channel Tunnel, becoming a crucial tool for ensuring the long-term safe operation of tunnels.

[0006] At the same time, the rapid development of IoT technology and wireless sensor networks (WSNs) has provided a new technical path for the deployment of TSHMS. Through TWSNs, wireless collection and transmission of tunnel structural data can be achieved, significantly reducing wiring costs, complexity, and maintenance difficulties. For example, traditional tunnel monitoring systems typically require a large number of cables to connect sensors, which is not only complex to install but also susceptible to environmental interference, resulting in data distortion. TWSNs, on the other hand, offer advantages such as flexible deployment, low cost, and strong scalability, effectively overcoming the limitations of traditional wired systems. However, the application of TWSNs in tunnel environments also faces unique challenges, such as limited node deployment space, a large number of node types, and the need for manual node deployment in three-dimensional space. Therefore, when researching TWSN technology, in addition to considering the general characteristics of WSNs (such as energy constraints and short transmission distances), it is also necessary to optimize the design based on the specific characteristics of the tunnel environment.

[0007] At present, the research on tunnel wireless sensor network is mostly based on linear or planar model to carry out node deployment and routing algorithm design, but in the process of tunnel structure characteristic research, the safety and stability of the tunnel are jointly influenced by multiple factors such as surrounding rock pressure, groundwater seepage, lining structure, construction joint, crack distribution and geological conditions. Therefore, the traditional linear or planar model is difficult to fully reflect these complex influencing factors, thus leading to the inability to meet the demand for all-around and multi-dimensional monitoring of tunnel structure. In China's tunnel engineering, mountain tunnel and urban subway tunnel account for the main part, among which mountain tunnel is mostly in complex geological environment, and urban subway tunnel is faced with the interference of dense traffic load and surrounding construction activities. It is of great significance to promote the construction of intelligent tunnel to deeply explore and analyze the tunnel structure characteristics, construct a three-dimensional wireless sensor monitoring network conforming to the three-dimensional spatial distribution characteristics of the tunnel, and optimize the design of routing algorithm. In addition, due to the limited energy of TWSN node, the core of TWSN routing algorithm design is to ensure the efficient interconnection and intercommunication of network nodes and adapt to the tunnel monitoring demand, and at the same time, how to effectively alleviate and avoid the 'hot spot' problem and 'energy hole' phenomenon caused by uneven node energy consumption, so as to prolong the network life cycle and improve the reliability of monitoring data. SUMMARY

[0008] The application provides a tunnel dual-energy clustering routing method based on multi-objective optimization, which can effectively solve the problems of uneven node energy consumption and limited energy supply in dynamic and energy-limited tunnel environment, significantly prolong the network life cycle and improve the energy efficiency level.

[0009] The technical scheme adopted by the application is:

[0010] A tunnel dual-energy clustering routing method based on multi-objective optimization specifically includes: constructing a three-dimensional spatial model of a single-tube tunnel to characterize the geometric structure of the tunnel; secondly, proposing a multi-objective clustering routing algorithm Tunnel Clustering Routing (TCR) and a dual-energy dynamic supply algorithm Dual Energy-based Tunnel Clustering Routing (DETCR); based on the Competition of Tribes and Cooperation of Members algorithm (CTCM), a Multi-Objective Competition of Tribes and Cooperation of Members (MOCTCM) algorithm is designed, which comprehensively considers the remaining energy, communication distance, energy consumption and relay node (RN) election frequency of the node to provide a decision basis for adaptability and global optimization ability for the selection of the optimal cluster head (Optimal Cluster Head) and the optimal relay node (Optimal Relay Node); the TCR algorithm uses the Pareto solution set generated by MOCTCM to select the optimal relay node through the Technique for Order Preference by Similarity to Ideal The TOPSIS method was used to select Optima-CH, and the coefficient of variation method was used to dynamically weight each objective function to determine Optima-RN. On this basis, DETCR integrated solar light tracking with a vehicle frequency-driven vibration energy harvesting mechanism to achieve efficient energy replenishment for low-energy nodes, thereby improving the network's life cycle and operational stability.

[0011] Further:

[0012] The specific implementation steps of this method are as follows:

[0013] Step 1: Create a 3D spatial model of a single-tube tunnel:

[0014] The constructed three-dimensional space model single tube tunnel (3D-SM-STT) is abstracted into a rectangular parallelepiped with dimensions L*W*H, and a semi-cylindrical cavity with a radius of R. Photovoltaic panels are deployed at the tunnel entrance and exit areas, and electromagnetic wave penetration power supply or magnetic induction coupling energy transfer non-contact power supply technology effectively achieves stable power supply for high-power sensors embedded in the structure.

[0015] Step 2: Node deployment: During the tunnel construction phase, sensor nodes are deployed in the 3D spatial model. During node deployment, low-discrepancy sequences are used to initialize the node positions based on the tunnel structure.

[0016] Step 3: Network initialization. This phase is performed only once after node deployment to ensure network operation. This phase includes network partitioning and the establishment of a dual-energy supply model based on variable-time interval light tracking and vehicle frequency drive. The dual-energy function model includes a solar energy supply model (SESM) and a vibratory energy model (VEM).

[0017] Network partitioning adopts a partitioning strategy based on geographical division;

[0018] The solar energy model SESM adopts an energy collection method by adjusting the angle of the photovoltaic panel by tracking the sun's altitude angle;

[0019] The vibration energy model (VEM) is based on the vehicle noise frequency characteristics and its electrical energy response, simulating the vehicle frequency under multiple factors to achieve efficient vibration energy collection;

[0020] Both models introduce an energy allocation balancing mechanism to improve energy utilization efficiency and extend the network lifecycle;

[0021] Step 4: Setup: The base station (BS) first elects cluster heads (CHs) and RNs based on information obtained during network initialization. It then constructs a cluster structure with the selected cluster heads as the core, dividing the nodes into clusters to form a hierarchical network topology. Furthermore, the BS performs inter-cluster routing optimization, rationally configuring multi-hop forwarding paths between cluster heads to improve data transmission efficiency and achieve a balanced distribution of network energy consumption.

[0022] Step 5: In the steady-state phase, ordinary nodes collect data and transmit data to the CH according to the Time Division Multiple Address (TDMA) time slots allocated by the BS. They enter sleep mode during non-transmission periods to reduce data conflicts and energy consumption.

[0023] Going further:

[0024] In step 4, the BS performs cluster structure construction, CH selection and inter-cluster routing optimization based on the TCR algorithm and the DETCR algorithm:

[0025] The TCR algorithm mainly includes three parts: first, decoding: decoding provides the basis for the subsequent selection of CH and RN; second, MOCTCM is used to select Optima-CH, and Optima-RN is determined based on the RN objective function; when selecting Optima-CH, the optimal solution is determined by combining the MOCTCM algorithm with various objective functions: residual energy f1, average distance between nodes in the partition f2, and distance from node to RN f3; and when selecting Optima-RN, the objective function value of each node is calculated through various objective functions: square of distance between RN and CH f4, energy consumption f5, and RN count control f6, and the coefficient of variation method is used to dynamically weight each objective function to determine the optimal solution; third, the optimal cluster and optimal inter-cluster routing are formed based on Optima-CH and Optima-RN; when Optima-CH and Optima-RN are determined, the BS broadcasts the message, and the ordinary node Ordinary Based on the received information, Node ON determines the cluster structure using the distance factor. The BS then allocates time slots to each node to ensure conflict-free data transmission. Nodes within the cluster send data to the CH via a single hop, while the CH transmits the data to the BS via multiple hops. The DETCR algorithm incorporates the SESM and VEM energy supply models on top of the TCR algorithm, significantly improving the network lifecycle. The objective functions described above are:

[0026] 1) Residual energy f1

[0027] After network initialization, the BS can calculate the residual energy of each node. Compared with a single residual energy value, the combined global average energy and distribution information can dynamically adjust the weight of node energy in CH selection to adapt to energy-uneven network scenarios. The objective function is as follows:

[0028]

[0029] Among them E i is the residual energy of node i, unit: J, E avg is the average energy of all nodes in the partition, unit: J, E max , E min are the maximum and minimum values ​​of the residual energy in the partition, unit: J;

[0030] 2) Average distance f2 between nodes in a partition

[0031] The average distance between nodes within a partition is one of the key factors affecting data transmission within a cluster. This paper comprehensively considers distance, distance distribution uniformity, and inverse distance compensation, and the objective function is as follows:

[0032]

[0033] wherein D ij represents the distance between node i and node j, unit: m, the average distance between node i and other nodes in the partition unit: m, m represents the total number of nodes in the partition;

[0034] 3) the distance f3 of the node to the RN

[0035] The distance of the node to the RN is one of the key factors affecting the inter-cluster data transmission; since the RN is not determined in the CH selection stage, and its selection range is located in the next partition, the distance of the node to the center of the next partition is defined as the distance of the node to the RN when constructing the objective function; at the same time, by combining the intra-cluster distance to dynamically adjust the weight, the node in the sparse area is avoided to become the CH, so as to improve the efficiency of multi-hop communication, and the objective function is as follows: wherein D next (i) is the distance of node i to the RN, unit: m, D max is the maximum distance between nodes in the partition, unit: m;

[0036]

[0037] 4) the square of the distance f4 of the RN to the CH

[0038] The selection of the RN plays a key role in optimizing the data transmission efficiency and network energy balance; since the node energy consumption is closely related to the transmission distance, and the transmission energy consumption is in a power relationship with the distance; therefore, the square of the distance of the RN to the CH node is taken as the optimization target, and the objective function is as follows:

[0039]

[0040] 5) energy consumption f5

[0041] Energy consumption balance is the core mechanism to ensure the sustainability and stability of the network; the objective function is constructed by combining the ratio of the residual energy E i of the node, unit: J, and the transmission energy consumption E tx (i), unit: J, and the variance thereof:

[0042]

[0043] 6) RN count control f6

[0044] Controlling the node to be frequently selected as the RN can prevent the node from failing due to excessive energy consumption; the present application sets an initial value L con (i) = 0.1 for each node, and is incremented in each subsequent round, and when the value of the node is closer to 1, the probability of being selected as the RN is greater; once the node is selected as the RN, the value is restored to the initial value in the next round, and the objective function is as follows:

[0045]

[0046] The MOCTCM algorithm introduces three key steps into CTCM to enhance its ability to solve multi-objective problems, specifically:

[0047] First, the calculation and archiving of non-dominated solutions: During the iteration process, each tribe generates a corresponding solution set based on the multi-objective fitness function. After completing a round of iteration, the fitness values ​​of all tribes are compared to screen out non-dominated solutions and store them in the non-dominated solution set to maintain the superiority and diversity of the solutions.

[0048] Secondly, the local optimal solution is updated based on the dominance relationship: the new fitness value generated after the tribe iteration is compared with the local optimal solution before the iteration. If the new solution dominates the original local optimal solution, it is replaced; if not, a random function is introduced to update it with a certain probability to balance global exploration and local development capabilities;

[0049] Finally, the global optimal solution is updated based on TOPSIS: After each round of iteration, TOPSIS sorting is applied to the non-dominated solution set, and the global optimal solution, namely Optima-CH, is selected based on the proximity to the ideal solution to guide the next round of iteration to improve the optimization convergence and solution quality.

[0050] In the solar model SESM described in step 3, the photovoltaic panel is always perpendicular to the sunlight, and the corresponding energy TE da The output looks like this:

[0051] TE da =H0S pv =E e

[0052] Among them, S pv Indicates the area of ​​the photovoltaic panel, unit: m 2 , H0 represents the intensity of sunlight directly hitting the photovoltaic panel, unit: Lux, E e , represents the energy under vertical illumination, unit: J;

[0053] In the vibration energy supply model (VEM), the energy captured by the sensor when a vehicle passes by is as follows. U represents voltage (V); I represents current (A); and t represents time (s). To evaluate the total energy capture, it is necessary to understand the characteristics of the traffic flow in the tunnel. A Poisson distribution is used to construct a basic traffic flow model, and Gaussian noise and an accident attenuation factor are introduced to enhance the characterization of randomness and suddenness. The instantaneous traffic flow Q(t) at time t is as follows:

[0054] W=Pt=UIt

[0055] Q(t)=max(P(t)+G(t)-A(t),0)

[0056] Among them, P(t) is a random variable P(t)~Poisson(λ(t)) that obeys the Poisson distribution parameter λ(t), λ(t) represents the average traffic flow at a certain moment, which is affected by the peak period λ peak , unit: veh / min and trough period λ offpeak , unit: veh / min, the influence is as follows; G(t) is a model with mean 0 and variance σ 2 Normal distribution G(t)~N(0,σ 2 ), represents the flow attenuation term caused by the emergency, ζ is the accident impact factor, which represents the flow reduction ratio during the accident; θ is the accident attenuation rate, which represents the recovery speed; t a Indicates the time when the accident occurred;

[0057]

[0058] The data collection method described in step 5 is as follows: the ordinary node ON first collects data and transmits data to the CH according to the TDMA time slot allocated by the BS, and enters sleep mode during non-transmission periods to reduce data conflicts and energy consumption. After receiving the data from the ON, the CH node first fuses the data to reduce redundant information and improve transmission efficiency. Then, the CH transmits the data to the BS through multi-hop transmission using the CSMA / CA protocol through optimal inter-cluster routing to ensure reliable data transmission.

[0059] Based on the structural characteristics and monitoring requirements of the "single-hole tunnel", this paper constructs a 3D-SM-STT model, and combines the MOCTCM algorithm to propose a cluster routing algorithm DETCR for dynamic dual-energy supply of TWSN. This algorithm focuses on solving the multi-objective optimization problem of Optima-CH and Optima-RN selection in a three-dimensional wireless sensor monitoring network, as well as the problem of how to supply solar energy, vibration energy and energy balance, thereby realizing the autonomous construction of efficient paths and stable data transmission. By introducing a dual-energy replenishment mechanism and a dynamic routing strategy, the DETCR algorithm significantly improves network energy efficiency, extends the network life cycle, and provides a high-efficiency and energy-saving solution for WSNs in complex tunnel environments. The details are as follows:

[0060] (1) A three-dimensional tunnel spatial model (3D-SM-STT) with the characteristics of TWSN node deployment is constructed. Compared with the traditional two-dimensional plane model, this model can more realistically reflect the spatial distribution characteristics of the tunnel structure and provide a more reasonable simulation environment for verifying the performance of the cluster routing algorithm in TWSN.

[0061] (2) A dual energy supply model based on variable time interval illumination tracking and vehicle frequency drive is proposed. The solar energy model optimizes the photovoltaic panel angle to achieve vertical incidence to improve collection efficiency. The vibration energy model simulates vehicle frequency under multiple factors based on the vehicle noise frequency characteristics and its electrical energy response to achieve efficient vibration energy collection. Both models introduce an energy distribution balancing mechanism to improve energy utilization efficiency and extend the network lifecycle.

[0062] (3) Based on the existing CTCM algorithm, MOCTCM is proposed. This algorithm integrates the dual-energy dynamic power supply mechanism with the tunnel space model to achieve joint optimization of multiple optimization objectives, including node residual energy, distance, energy consumption, and the number of times a node is selected as a RN. Within this multi-objective optimization framework, the base station comprehensively evaluates the performance indicators of each node and dynamically selects the optimal CH and optimal RN in each partition to further improve the overall network performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a schematic diagram of a single-hole tunnel space model of the present invention;

[0064] Figure 2 This is a working principle diagram of the TWSN cluster routing algorithm of the present invention;

[0065] Figure 3 Schematic diagram of decoding of MOCTCM of the present invention;

[0066] Figure 4 is a flow chart of the present invention;

[0067] Figure 5 The network life cycle of the present invention;

[0068] Figure 6 This is the energy supply diagram of the present invention;

[0069] Figure 7 This is the residual energy diagram of the present invention. DETAILED DESCRIPTION

[0070] The present invention will be further described in detail below with reference to the accompanying drawings and experiments.

[0071] 1. Single-hole tunnel spatial model

[0072] In TSHMS, wireless sensor nodes can be preset at key structural locations such as the lining layer, tunnel wall or inside the tunnel wall during the tunnel construction phase according to monitoring requirements to achieve long-term monitoring of the tunnel. Figure 1As shown, the 3D-SM-STT model constructed by the application abstracts its structure as a cuboid of L*W*H, and has a semi-cylindrical cavity with a radius of R (equivalent to removing a semi-cylinder with a radius of R in the cuboid of L*W*H), to realistically simulate the spatial form and layout constraints of the tunnel. In addition, photovoltaic panel assemblies are arranged at the entrance and exit areas of the tunnel, and through non-contact energy supply technologies such as electromagnetic wave penetration power supply or magnetic induction coupling energy transmission, stable power supply can be effectively realized for high-power sensors embedded in the structure. The 3D-SM-STT model not only reflects the structural complexity and spatial limitations of the tunnel environment, but also provides a unified and practically feasible modeling basis for subsequent node layout optimization, energy efficiency scheduling, and routing strategy design.

[0073] The working principle diagram of the TWSN clustering routing algorithm based on the 3D-SM-STT model is shown in Figure 2 As shown, the data transmission is optimized through hierarchical structure. To enhance the illustration clarity of the regional division, the same type of ONs distributed in Area A and Area B are marked as "ON in area A" and "ON in Area B", respectively, although their functions are consistent. This representation method is only used to distinguish the spatial area to which the nodes belong. The sensor nodes deployed at specific positions of the tunnel select CH and RN according to the preset mechanism, form a cluster structure and build a multi-hop transmission path. The ONs in the cluster transmit monitoring data to the CH, and the CH transmits the data to the BS through the RNs step by step, to realize periodic data acquisition and efficient transmission.

[0074] II. DETCR algorithm design

[0075] The DETCR algorithm consists of four phases: node deployment, network initialization, setup, and steady-state. During the node deployment phase, nodes are pre-placed in a three-dimensional tunnel model during tunnel construction. During the network initialization phase, the network is first divided into regions, and a dynamic solar energy supply model and a vibration energy supply model are established. During this phase, the base station (BS) broadcasts an inquiry message (Inquiry-REQ) to all nodes using the CSMAMAC protocol. The message includes BS location information, network size, and instructions for activating the initial communication channel, laying the foundation for subsequent cluster construction. During the setup phase, a set of candidate cluster heads (CCHs) is first constructed. Based on this, the MOCTCM algorithm is used to dynamically select the optimal cluster head (CH). After CH election, the optimal RN is determined based on the RN election objective function and the coefficient of variation method, completing the cluster structure. Finally, inter-cluster routing is optimized based on the selected CHs and RNs, establishing the optimal cross-cluster path to achieve efficient data transmission and balanced network energy consumption. During this phase, the BS is responsible for cluster formation, CH node identification, CH-to-BS path optimization, TDMA time slot allocation, and control information broadcasting. The broadcast control information includes ON information, channel identifiers, inter-cluster routing tables, and time slot allocation schemes. During the steady-state phase, the network collects data and transmits data within and between clusters. Nodes transmit data based on the routing table and TDMA time slots, maintaining communication only during designated active periods and entering a low-power sleep mode the rest of the time. This effectively reduces energy consumption and further improves network energy efficiency.

[0076] 1) Residual energy (f1)

[0077] Residual energy is one of the most important factors in the CH election process. It not only influences the selection of a CH but also determines whether a node can support multiple rounds of data aggregation and forwarding tasks. Electing nodes with higher residual energy can improve network stability and extend the network lifecycle. After network initialization, the BS can calculate the residual energy of each node. Compared to a single residual energy value, combining global average energy and distribution information can dynamically adjust the weight of node energy in CH selection to adapt to network scenarios with uneven energy distribution. The objective function is shown in Equation (1).

[0078]

[0079] 2) Average distance between nodes in a partition (f2)

[0080] The average distance between nodes in a partition is one of the key factors affecting data transmission within a cluster. The present invention comprehensively considers distance, distance distribution uniformity and inverse distance compensation, and the objective function is shown in formula (2).

[0081]

[0082] 3) Node-to-RN distance (f3)

[0083] The node-to-RN distance is one of the key factors affecting inter-cluster data transmission. Since the RN is not determined in the CH selection stage and its selection range is in the next partition, the present application defines the node-to-RN distance as the distance from the node to the center of the next partition when constructing the objective function. At the same time, by combining the intra-cluster distance to dynamically adjust the weight, the node in the sparse area is avoided to become the CH, thereby improving the efficiency of multi-hop communication. The objective function is shown in equation (3).

[0084]

[0085] The objective function of the above-defined CH election, such as the residual energy f1, the average distance between nodes in the partition f2, and the node-to-RN distance f3, can be expressed as equation (4):

[0086]

[0087] where t represents time, M represents the target number, X = [X1, X1,..., X N ]∈Ω, X represents the decision vector, Ω represents the decision space, and F(X, t) represents the minimized objective function at time t;

[0088] 4) RN-to-CH distance square (f4)

[0089] The selection of RN plays a key role in optimizing data transmission efficiency and network energy balance. Since the node energy consumption is closely related to the transmission distance, and the transmission energy consumption is in a power relationship with the distance. Therefore, the square of the distance from the RN to the CH node is taken as the optimization target, and the objective function is shown in equation (5).

[0090]

[0091] 5) Energy consumption (f5)

[0092] Energy consumption balance is the core mechanism to ensure the sustainability and stability of the network. The present application constructs the objective function as equation (6) by combining the ratio of the node residual energy E i and the transmission energy consumption E tx (i) and the variance thereof.

[0093]

[0094] 6) RN count control (f6)

[0095] Controlling the node to be frequently selected as RN can prevent the node from failing due to excessive energy consumption. The present application sets an initial value L con(i) = 0.1 and increases in each subsequent round. The closer the value of a node is to 1, the greater the probability of being selected as an RN. Once a node is selected as an RN, the value is restored to the initial value in the next round. The objective function is shown in Equation (7).

[0096]

[0097] The objective function of RN election defined above, such as the square of the distance between RN and CH f4, energy consumption f5, and RN count control f6, can be expressed as formula (8):

[0098] minF(X,t)=ω1*f4(X,t)+ω2*f5(X,t)+ω3*f6(X,t) (8)

[0099] Among them, ω1, ω2, and ω3 represent the objective function weights respectively.

[0100] (a) Network initialization phase

[0101] Network initialization is essential for normal network operation. In the DETCR protocol, network initialization is performed only once after node deployment to ensure normal network operation. This phase includes network partitioning, establishing a solar energy supply model (SESM), and establishing a vibratory energy model (VEM).

[0102] ① Network partition

[0103] In TWSNs, network partitioning is a key technology designed to improve network energy efficiency, balance node loads, and optimize data transmission paths. A rational partitioning strategy not only reduces energy consumption between nodes but also improves data transmission stability, thereby extending the network lifecycle. To address the unique constraints of tunnel environments, this paper adopts a geographically based partitioning strategy to ensure uniform distribution of communication channels (CHs), optimize inter-cluster communication, and improve overall network energy efficiency. After determining the partition distance, the number of Optima-CHs is calculated to further enhance the network's structural rationality and operational efficiency.

[0104] ② Establishment of solar energy model

[0105] As a stable renewable energy source, solar energy can continuously supply energy to sensor nodes, with low maintenance costs and good applicability. However, due to factors such as day and night and climate, energy management needs to be optimized to improve utilization efficiency. This paper combines the geographical characteristics of TWSN to improve the solar power supply model. Considering that the light intensity and time have a sinusoidal relationship, a model is designed to track the solar altitude angle to dynamically adjust the angle of the photovoltaic panel to achieve vertical light incidence and maximize energy collection. Energy E under vertical illuminatione Output as shown in formula (9).

[0106] E e = S pv H0 (9)

[0107] Wherein, S pv represents the area of the photovoltaic panel, H0 represents the light intensity of the sunlight directly on the photovoltaic panel, as shown in formula (10).

[0108]

[0109] In the formula, T α is the sunshine time, and the supplementary phase angle t sr represents the sunrise time of the day, and the light intensity function amplitude A is shown in formula (11).

[0110]

[0111] Wherein day represents a day in a year, the winter solstice is defined as the first day, α+o represents the latitude value relative to the Tropic of Cancer, the Tropic of Cancer is the maximum value 1, the value range of δ is [1, 2], δ=1 represents a leap year, and δ=2 represents a common year.

[0112] When the sunlight deflection angle β, the effective area S epv = S pv cosβ, so the energy collected by the solar energy supply model is shown in formula (12).

[0113]

[0114] However, the present application proposes an energy collection method for adjusting the angle of the photovoltaic panel by tracking the solar altitude angle, so that the photovoltaic panel is always perpendicular to the sunlight, and the corresponding energy TE da Output as shown in formula (13). In the energy distribution strategy, a remaining energy proportion driven distribution mechanism is adopted, and nodes with lower energy are preferentially supplemented to reduce the energy variance of the nodes and improve the network energy balance.

[0115] TE da = H0S pv = E e (13)

[0116] ③Vibration energy model establishment

[0117] Piezoelectric vibration energy harvesters have become an important solution for harvesting ambient vibration energy due to their simple structure, weather resistance, and ease of miniaturization. They leverage the mechanical-to-electrical energy conversion properties of piezoelectric materials to convert ambient vibrations into electrical energy. Existing research has made some progress in alleviating energy constraints in WSNs. Various energy harvesting models have been proposed, focusing on structural design, modeling, and parameter optimization. However, most of these models are complex and computationally expensive. Considering that the steady noise generated by vehicle traffic in tunnels can excite structural vibrations, this paper utilizes this as an energy source to further improve energy harvesting efficiency. Specifically, the frequency of vehicle noise in tunnels is concentrated between 100–1200 Hz, exhibiting a bimodal characteristic. Train noise has a higher frequency, ranging from 20–5000 Hz, with a primary concentration above 800 Hz. Related research has verified the energy conversion capability of sensors within this frequency range. This paper combines existing methods to construct an energy harvesting model based on noise-induced vibrations and applies it to clustered routing optimization in TWSNs to enhance the network's self-powered capabilities and overall performance.

[0118] The energy that the sensor can obtain when a vehicle passes by is shown in formula (14). In this formula, U represents voltage, I represents current, and t represents time. To evaluate the total energy acquisition, it is necessary to understand the characteristics of the traffic flow in the tunnel. Traffic flow usually has day and night fluctuations and high and low peak characteristics, and is affected by weather, holidays and emergencies. To more accurately simulate actual traffic flow, the present invention uses Poisson distribution to construct a basic traffic model and introduces Gaussian noise and accident attenuation factors to enhance the characterization of randomness and suddenness. The instantaneous traffic flow Q(t) at time t is shown in formula (15).

[0119]

[0120] Among them, P(t) is a random variable P(t)~Poisson(λ(t)) that obeys the Poisson distribution parameter λ(t), λ(t) represents the average traffic flow at a certain moment, which is affected by the peak period λ peak and trough period λ offpeak The influence of is shown in formula (16). G(t) is a model with mean 0 and variance σ 2 Normal distribution G(t)~N(0,σ 2 ), represents the flow attenuation term caused by the emergency, ζ is the accident impact factor, and represents the flow reduction ratio during the accident. is the accident attenuation rate, indicating the recovery speed. a This method can more realistically reflect the traffic flow characteristics in the tunnel and improve the applicability of traffic flow simulation.

[0121]

[0122] (b) Setup phase

[0123] After network initialization is complete, the protocol enters a periodic operation mode, with each round consisting of a setup phase and a steady-state phase. During the setup phase, the base station (BS) uses information obtained during the network initialization phase to perform key tasks such as cluster structure construction, channel selection, and inter-cluster routing optimization to ensure efficient data transmission and balanced network energy distribution. The setup phase of the DETCR algorithm primarily consists of three parts: decoding, selecting the optimal channel using MOCTCM and determining the optimal RN based on the RN objective function, and forming the optimal cluster and optimal inter-cluster routing based on the optimal channel and RN. The DETCR protocol measures the runtime of a TWSN using rounds. To better integrate the two energy supply models, SESM and VEM, with the TWSN, the step size of SESM and VEM is set to 0.5 hours, corresponding to one round in the TWSN.

[0124] ① Coding

[0125] Figure 3 The decoding of sensor node selection CH and RN in TWSN is given. TWSN has M partitions, including N sensor nodes, which communicate with BS through CH and RN. There are B CHs and K RNs in the solution space of TWSN. Each partition in TWSN is represented by P i Indicates that, first, the nodes whose residual energy in the partition is higher than the average residual energy are selected as CCH, and the CCH set is CC i Indicates that item C i Indicates CH, R i Denotes RN. The jth partition P of TWSN j Contains H i nodes, HC i CH nodes, HR i RNs, of which Then select the HC with the best fitness from CCH according to the objective function j CH, and finally select HR from non-CH nodes according to the RN objective function j The selected CH nodes and RNs have better robustness than other nodes.

[0126] ②CTCM

[0127] The CTCM algorithm, proposed by Chen et al., draws inspiration from the behavioral mechanisms of ancient tribes, such as competition for resources and collaboration among members. By simulating the dynamic competition between tribes and the cooperation between members, the algorithm constructs an optimization framework that balances global exploration and local exploitation. CTCM demonstrates outstanding convergence speed and stability, and effectively avoids local optima.

[0128] Step 1 Initialization

[0129] Assume that the population of a primitive society is p, the number of tribes is n, and each tribe has m members. Then the initial position X of the entire society is as shown in Equation (17), where d represents the dimension of the solution space, and lb and ub represent the lower and upper bounds of the search space, respectively.

[0130]

[0131] Each person has a velocity vector that determines where the person will go next. The total velocity matrix V of primitive society is shown in Equation (18).

[0132]

[0133] Step 2 Member Cooperation

[0134] In a tribe, the chief is usually responsible for managing and planning the tribe's future development. Most members follow the chief's plans and instructions, but each member also has his or her own personal ideas, causing their loyalty to fluctuate over time. Over time, this loyalty exhibits chaotic behavior, and this chaotic nature becomes more pronounced as the number of tribes increases. In this case, the concept of the sinusoidal chaos map shown in Equation (19) is used to characterize the changes in member loyalty.

[0135]

[0136] Each person will communicate with the tribe's chief to get instructions on what he or she should do next. He or she will also combine his or her own experience to provide an additional source of information for the next action. The speed update is shown in Equation (20).

[0137]

[0138] in, represents the speed of the mth member of the nth tribe at time t+1, represents the best fitness position found by the member in the entire period, represents the position at time t, It represents the best location found by the tribe during the entire period. c1 and c2 represent the parameters for tribe members to follow their own experience and obey the instructions of the chief, respectively. and It is the chaotic loyalty of each member.

[0139] Step 3: Tribal Competition

[0140] Random conflicts will occur between tribes, when the weaker tribe will flee, while the stronger tribe will not be affected. The speed update is as shown in Equation (21).

[0141]

[0142] in, is the optimal position found by the competitor, indicating the optimal fitness of the nth tribe, is the competitor's optimal fitness, represents the tribal escape coefficient, c3 is a chaotic random factor, The retreat speed is revealed. At the same time, the positions of tribe members are updated as shown in formula (22). If the updated positions of tribe members Beyond the feasible region [X min ,X max ], the velocity rebounds in a mirror image as shown in formula (23), and the position information is corrected.

[0143]

[0144] ③MOCTCM

[0145] In order to achieve multi-objective optimization, the present invention introduces three key steps into CTCM and expands it into MOCTCM to enhance its ability to solve multi-objective problems.

[0146] First, non-dominated solutions are calculated and archived. During the iteration process, each tribe generates a corresponding solution set based on the multi-objective fitness function. After completing a round of iteration, the fitness values ​​of all tribes are compared to screen out non-dominated solutions and store them in the non-dominated solution set to maintain the superiority and diversity of the solutions.

[0147] Secondly, the local optimal solution is updated based on the dominance relationship. The new fitness value generated after the tribe iteration is compared with the local optimal solution before the iteration. If the new solution dominates the original local optimal solution, it is replaced. If not, a random function is introduced to update it with a certain probability to balance global exploration and local development capabilities.

[0148] Finally, the global optimal solution is updated based on TOPSIS. After each round of iteration, TOPSIS sorting is applied to the set of non-dominated solutions. The global optimal solution, namely Optima-CH, is selected based on its proximity to the ideal solution. This is used to guide the next round of iteration to improve optimization convergence and solution quality.

[0149] After the Optima-CH election is complete, the fitness values ​​of non-CH nodes in the area are calculated. The coefficient of variation method is used to dynamically weight the overall fitness to determine the Optima-RN. This optimization strategy ensures the stability of the data transmission path and balances the network load.

[0150] After the CH and RN are determined, the BS supplies energy to the entire network based on the SESM and VEM and optimizes the energy allocation strategy to improve energy efficiency. To this end, this invention improves on the traditional uniform allocation approach by taking into account the remaining energy status of nodes and performing differentiated energy allocation according to a certain ratio. This minimizes energy variance between nodes and mitigates network imbalance failures. This strategy not only enhances the network's adaptability but also provides a more stable energy foundation for subsequent iterations, further optimizing inter-cluster communication performance.

[0151] ④Optimal cluster and optimal routing

[0152] After the BS selects the Optima-CH using the MOCTCM algorithm and determines the Optima-RN based on multiple RN objective functions, it broadcasts the Optima-CH and Optima-RN information to the entire network via control messages. Upon receiving this information, the ON selects the closest CH within its partition based on its distance to each CH, adding it to the cluster and completing cluster construction. The BS then allocates time slots to the ON and CH using the TDMA mechanism and broadcasts this information to the ON, CH, and RN to ensure orderly data transmission and avoid transmission conflicts. During data transmission, cluster nodes send data to the CH via a single hop, while the CH transmits the data to the BS via multiple hops, enabling efficient information aggregation and transmission.

[0153] (c) Steady-state stage

[0154] Once the optimal cluster structure and optimal routing are established, the network enters the steady-state phase. In this phase, ONs collect data and transmit data to CHs according to the TDMA time slots allocated by the BS. They enter sleep mode during non-transmission periods to reduce data conflicts and energy consumption.

[0155] After receiving data from the ON, the CH node first performs data fusion processing to reduce redundant information and improve transmission efficiency. Subsequently, the CH transmits the data to the BS through the optimal inter-cluster routing using the CSMA / CA protocol over multiple hops to ensure reliable data transmission.

[0156] The steady-state phase continues until the end of this round, after which the network re-enters the setup phase for the next round to adapt to topology changes and dynamic adjustments to node energy states.

[0157] Figure 4 Flowchart of the present invention.

[0158] 3. Experimental Verification of the Present Invention

[0159] This paper uses MATLAB R2020b for simulation experiments. In order to verify the adaptability of the algorithm under networks of different scales, two experimental scenarios are designed. Their parameters are shown in Table I.

[0160] 1) Network life cycle

[0161] The network lifecycle is a key metric for measuring TWSN performance, comprehensively reflecting the network's operational stability and sustainability under resource-constrained conditions. A longer lifecycle typically indicates that the algorithm employed performs well in energy consumption control and load balancing. This helps improve network stability, ensure the continuity of data transmission and the integrity of data collection, while also reducing maintenance costs and enhancing the application value and scalability of the TSHMS system in real-world deployments. Figure 5 The network lifecycle of five algorithms in two scenarios is demonstrated.

[0162] The results in the figure show that the EMOGJO, ESR-HAC, and CTCM algorithms generally have short lifecycles, primarily due to deficiencies in the channel selection and optimization strategy design. For example, in EMOGJO, the number of channels is determined by a random strategy, resulting in significant fluctuations in the network structure; ESR-HAC's objective function is relatively simple and lacks effective integration of key indicators; and CTCM lacks a reasonable multi-objective optimization strategy, resulting in uneven weight distribution of indicators in the objective function and difficulty in achieving effective coordination between energy and communication performance. These issues limit the algorithms' performance in terms of energy-balanced scheduling and network stability.

[0163] Table I Experimental scenario parameter settings

[0164]

[0165] The network parameters of 3D TWSN, MOCTCM parameters, SESM and VEM model parameters are shown in Table II.

[0166] Table II Experimental parameters

[0167]

[0168] In comparison, the TCR and DETCR algorithms perform better in terms of lifecycle indicators. TCR incorporates the spatial structural characteristics of the 3D-SM-STT model in the clustering process and improves the rationality of the cluster structure through regional division. When selecting CH and RN, the algorithm comprehensively considers multi-dimensional parameters such as the distance between nodes, remaining energy, and the number of historical selections, and uses a multi-objective optimization strategy to make dynamic decisions on node roles, further optimizing the cluster structure and energy utilization efficiency, thereby effectively delaying node failure and significantly improving the network lifecycle. Based on the basic framework of TCR, DETCR further introduces solar energy and vibration energy harvesting mechanisms to provide external energy supplements for low-energy nodes, effectively alleviating the problem of early failure caused by differences in node energy consumption, and further enhancing the continuity and stability of network operation. Figure 6The schematic diagram of the energy supply mechanism of solar energy and vibration energy. The time step of energy supply is 0.5h. Since the solar energy supply is affected by sunrise and sunset time, solar elevation angle, etc., the daily energy supply is different and the energy supply presents periodic changes, Figure 6 (a) shows the total amount of solar energy supply throughout the life cycle. The vibration energy supply is affected by the daily traffic volume, Figure 6 (b) shows the total amount of vibration energy supply throughout the life cycle.

[0169] 2) Energy consumption

[0170] In order to evaluate the performance of each algorithm in terms of node energy consumption, statistical analysis of the residual energy of nodes in the network was conducted. Figure 7 The residual energy of the network in scenario 1 and scenario 2 for the five algorithms is shown. The results show that the energy consumption of TCR algorithm in scenario 1 is better than EMOGJO and ESR-HAC, and similar to CTCM; in scenario 2, it is overall superior to the three comparison algorithms. In addition, the energy consumption performance of DETCR algorithm throughout the life cycle is always significantly better than other algorithms. The reason why the proposed algorithm has better energy consumption performance is that: first, the Optima-CH number is calculated according to the TWSN model; second, combined with network partition, the CH and RN in each partition are reasonably allocated, thus effectively avoiding the energy consumption caused by long-distance communication. CTCM performs similarly to TCR in scenario 1 because it also uses the partition strategy and Optima-CH number setting proposed in this paper. However, due to the unreasonable weight allocation in the objective function, although the residual energy of the nodes is high, there is still a problem of early failure of some nodes.

Claims

1. A tunnel dual-energy clustering routing method based on multi-objective optimization, characterized in that: Specifically include: Construct a three-dimensional spatial model of a single-tube tunnel to represent the tunnel's geometric structure; Secondly, a multi-objective clustering routing algorithm Tunnel Clustering Routing (TCR) and its dual energy dynamic supply algorithm Dual Energy-based Tunnel Clustering Routing (DETCR) are proposed; based on the tribe competition and member cooperation algorithm Competition of Tribes and Cooperation of Members, CTCM, a multi-objective tribe competition and member cooperation algorithm Multi-Objective Competition of Tribes and Cooperation of Members, MOCTCM is designed, which comprehensively considers the remaining energy, communication distance, energy consumption and relay node RelayNode, RN election frequency factors to provide a decision-making basis with adaptability and global optimization ability for the selection of optimal cluster head Optimal Cluster Head, Optima-CH and optimal relay node Optima-RN; TCR algorithm uses the Pareto solution set generated by MOCTCM and uses the Technique for Order Preference by Similarity to Ideal The TOPSIS method was used to select Optima-CH, and the coefficient of variation method was used to dynamically weight each objective function to determine Optima-RN. On this basis, DETCR integrated solar light tracking with a vehicle frequency-driven vibration energy harvesting mechanism to achieve efficient energy replenishment for low-energy nodes, thereby improving the network's life cycle and operational stability.

2. The tunnel dual-energy clustering routing method based on multi-objective optimization according to claim 1 is characterized in that: The specific implementation steps of this method are as follows: Step 1: Create a 3D spatial model of a single-tube tunnel: The constructed three-dimensional space model single tube tunnel (3D-SM-STT) is abstracted into a rectangular parallelepiped with dimensions L*W*H, and a semi-cylindrical cavity with a radius of R. Photovoltaic panels are deployed at the tunnel entrance and exit areas, and electromagnetic wave penetration power supply or magnetic induction coupling energy transfer non-contact power supply technology effectively achieves stable power supply for high-power sensors embedded in the structure. Step 2: Node deployment: During the tunnel construction phase, sensor nodes are deployed in the 3D spatial model. During node deployment, low-discrepancy sequences are used to initialize the node positions based on the tunnel structure. Step 3: Network initialization. This phase is performed only once after node deployment to ensure network operation. This phase includes network partitioning and the establishment of a dual-energy supply model based on variable-time interval light tracking and vehicle frequency drive. The dual-energy function model includes a solar energy supply model (SESM) and a vibratory energy model (VEM). Network partitioning adopts a partitioning strategy based on geographical division; The solar energy model SESM adopts an energy collection method by adjusting the angle of the photovoltaic panel by tracking the sun's altitude angle; The vibration energy model (VEM) is based on the vehicle noise frequency characteristics and its electrical energy response, simulating the vehicle frequency under multiple factors to achieve efficient vibration energy collection; Both models introduce an energy allocation balancing mechanism to improve energy utilization efficiency and extend the network lifecycle; Step 4: Setup: The base station (BS) first elects cluster heads (CHs) and RNs based on information obtained during network initialization. It then constructs a cluster structure with the selected cluster heads as the core, dividing the nodes into clusters to form a hierarchical network topology. Furthermore, the BS performs inter-cluster routing optimization, rationally configuring multi-hop forwarding paths between cluster heads to improve data transmission efficiency and achieve a balanced distribution of network energy consumption. Step 5: In the steady-state phase, ordinary nodes collect data and transmit data to the CH according to the Time Division Multiple Address (TDMA) time slots assigned by the BS, and enter sleep mode during non-transmission periods to reduce data conflicts and energy consumption.

3. The tunnel dual-energy clustering routing method based on multi-objective optimization according to claim 1 is characterized in that: In step 4, the BS performs cluster structure construction, CH selection and inter-cluster routing optimization based on the TCR algorithm and the DETCR algorithm: The TCR algorithm mainly includes three parts: first, decoding: decoding provides the basis for the subsequent selection of CH and RN; second, MOCTCM is used to select Optima-CH, and Optima-RN is determined based on the RN objective function; when selecting Optima-CH, the optimal solution is determined by combining the MOCTCM algorithm with various objective functions: residual energy f1, average distance between nodes in the partition f2, and distance from node to RN f3; and when selecting Optima-RN, the objective function value of each node is calculated through various objective functions: square of distance between RN and CH f4, energy consumption f5, and RN count control f6, and the coefficient of variation method is used to dynamically weight each objective function to determine the optimal solution; third, the optimal cluster and optimal inter-cluster routing are formed based on Optima-CH and Optima-RN; when Optima-CH and Optima-RN are determined, the BS broadcasts the message, and the ordinary node Ordinary Based on the received information, Node ON determines the cluster structure using the distance factor. The BS then allocates time slots to each node to ensure conflict-free data transmission. Nodes within the cluster send data to the CH via a single hop, while the CH transmits the data to the BS via multiple hops. The DETCR algorithm incorporates the SESM and VEM energy supply models on top of the TCR algorithm, significantly improving the network lifecycle. The objective functions described above are: 1) Residual energy f1 After network initialization, the BS can calculate the residual energy of each node. Compared with a single residual energy value, the combined global average energy and distribution information can dynamically adjust the weight of node energy in CH selection to adapt to energy-uneven network scenarios. The objective function is as follows: Among them E i is the residual energy of node i, unit: J, E avg is the average energy of all nodes in the partition, unit: J, E max , E min are the maximum and minimum values ​​of the residual energy in the partition, unit: J; 2) The average distance between nodes in the partition f2 The average distance between nodes within a partition is one of the key factors affecting data transmission within a cluster. This paper comprehensively considers distance, distance distribution uniformity, and inverse distance compensation, and the objective function is as follows: Among them D ij Represents the distance between node i and node j, unit: m, the average distance between node i and other nodes in the partition Unit: m, m represents the number of summary points in the partition; 3) Distance from node to RN f3 The distance from the node to the RN is one of the key factors affecting inter-cluster data transmission. Given that the RN has not yet been determined in the CH selection stage and its selection range is located in the next partition, the present invention defines the distance from the node to the center of the next partition as the distance from the node to the RN when constructing the objective function. At the same time, by dynamically adjusting the weight based on the intra-cluster distance, the nodes in the sparse area are prevented from becoming CHs, thereby improving the efficiency of multi-hop communication. The objective function is as follows: next (i) is the distance from node i to RN, unit: m, D max is the maximum distance between nodes in the partition, unit: m; 4) The square of the distance between RN and CH f4 The selection of RN plays a key role in optimizing data transmission efficiency and network energy balance. Since node energy consumption is closely related to transmission distance, and transmission energy consumption is in a power-order relationship with distance, the square of the distance from RN to CH node is used as the optimization target. The objective function is as follows: 5) Energy consumption f5 Energy consumption balance is the core mechanism to ensure the sustainability and stability of the network; the present invention combines the remaining energy E of the nodes i , unit: J and transmission energy consumption E tx (i), unit: J, ratio and variance, construct the objective function: 6)RN count control f6 The control node is frequently selected as RN to prevent the node from failing due to excessive energy consumption; the present invention sets an initial value L for each node. con (i) = 0.1 and increases in each subsequent round. The closer the value of a node is to 1, the greater the probability of being selected as an RN. Once a node is selected as an RN, the value returns to the initial value in the next round. The objective function is as follows:

4. The tunnel dual-energy clustering routing method based on multi-objective optimization according to claim 3 is characterized in that: The MOCTCM algorithm introduces three key steps into CTCM to enhance its ability to solve multi-objective problems, specifically: First, the calculation and archiving of non-dominated solutions: During the iteration process, each tribe generates a corresponding solution set based on the multi-objective fitness function. After completing a round of iteration, the fitness values ​​of all tribes are compared to screen out non-dominated solutions and store them in the non-dominated solution set to maintain the superiority and diversity of the solutions. Secondly, the local optimal solution is updated based on the dominance relationship: the new fitness value generated after the tribe iteration is compared with the local optimal solution before the iteration. If the new solution dominates the original local optimal solution, it is replaced; if not, a random function is introduced to update it with a certain probability to balance global exploration and local development capabilities; Finally, the global optimal solution is updated based on TOPSIS: After each round of iteration, TOPSIS sorting is applied to the non-dominated solution set, and the global optimal solution, namely Optima-CH, is selected based on the proximity to the ideal solution to guide the next round of iteration to improve the optimization convergence and solution quality.

5. The tunnel dual-energy clustering routing method based on multi-objective optimization according to claim 1 is characterized in that: In the solar model SESM described in step 3, the photovoltaic panel is always perpendicular to the sunlight, and the corresponding energy TE da The output looks like this: THE da =H0S pv =E e Among them, S pv Indicates the area of ​​the photovoltaic panel, unit: m 2 , H0 represents the intensity of sunlight directly hitting the photovoltaic panel, unit: Lux, E e , represents the energy under vertical illumination, unit: J; In the vibration energy supply model (VEM), the energy captured by the sensor when a vehicle passes by is as follows. U represents voltage (V); I represents current (A); and t represents time (s). To evaluate the total energy capture, it is necessary to understand the characteristics of the traffic flow in the tunnel. A Poisson distribution is used to construct a basic traffic flow model, and Gaussian noise and an accident attenuation factor are introduced to enhance the characterization of randomness and suddenness. The instantaneous traffic flow Q(t) at time t is as follows: W=Pt=UIt Q(t)=max(P(t)+G(t)-A(t),0) Among them, P(t) is a random variable P(t)~Poisson(λ(t)) that obeys the Poisson distribution parameter λ(t), λ(t) represents the average traffic flow at a certain moment, which is affected by the peak period λ peak , unit: veh / min and trough period λ offpeak , unit: veh / min, the influence is as follows; G(t) is a model with mean 0 and variance σ 2 Normal distribution G(t)~N(0,σ 2 ), represents the flow attenuation term caused by the emergency, ζ is the accident impact factor, which represents the flow reduction ratio during the accident; is the accident attenuation rate, indicating the recovery speed; t a Indicates the time when the accident occurred; 6. The tunnel dual-energy clustering routing method based on multi-objective optimization according to claim 1 is characterized in that: The data collection method described in step 5 is as follows: the ordinary node ON first collects data and transmits data to the CH according to the TDMA time slot allocated by the BS. It enters sleep mode during non-transmission periods to reduce data conflicts and energy consumption. After receiving the data from the ON, the CH node first performs data fusion processing to reduce redundant information and improve transmission efficiency. Subsequently, CH transmits the data to BS through multi-hop transmission using CSMA / CA protocol through optimal inter-cluster routing to ensure reliable data transmission.

Images