A wireless networking method for a marine ecological monitoring terminal

Abstract

The application belongs to the technical field of sea area monitoring wireless networking, and discloses a kind of sea area ecological monitoring terminal wireless networking method, and the whole network space spectrum database is constructed by node holographic calibration and sea space-time joint spectrum surveying and mapping, and three types of sea area regions of nearshore, offshore and far sea are dynamically divided in combination with environmental factors such as tide, wind wave, etc., each region executes differentiated clustering and topology rules, and is matched with sea area special routing prediction factor, adapts to multipath fading characteristics caused by seawater reflection, at the same time, far sea area relies on satellite backhaul to avoid network island risk in cooperation with topology, and heterogeneous network realizes dynamic cooperative access, not simple backup, so that networking architecture is deeply combined with sea time-varying environment, and the communication demand and transmission stability of different sea areas are considered;By constructing three-dimensional joint routing cost function of service, channel and energy, the dynamic weight is set according to the service priority of emergency alarm, routine monitoring and log reporting, to ensure low delay transmission of high priority service.

Description

Technical Field

[0001] This invention belongs to the field of wireless networking technology for marine monitoring, specifically a wireless networking method for marine ecological monitoring terminals. Background Technology

[0002] Marine ecological monitoring terminals are widely deployed in various sea areas, including nearshore, offshore, and open sea, for real-time monitoring of water quality parameters, hydrological and meteorological data, marine life, and emergency responses to oil spills and red tides. Their deployment scenarios are characterized by wide distribution, harsh environments, limited node power supply, drastic time-varying wireless channels, and significant differences in service priorities. Currently, there are three main wireless networking schemes for marine monitoring terminals, all of which have obvious core defects:

[0003] First, cellular mobile communication networking covers nearshore waters through shore-based 4G / 5G networks, but there is no effective coverage in offshore areas, and the terminal power consumption is high, making it unsuitable for long-term deployment nodes in the open sea powered by batteries.

[0004] Second, short-range wireless self-organizing networks, built on terrestrial protocols such as LoRa and ZigBee, are not adapted to the complex environmental characteristics of the sea. Affected by seawater reflection, tides, waves, and maritime communication interference, multipath fading is severe. Traditional fixed routing and clustering modes can easily lead to near-shore network congestion and the formation of network islands in the far sea. Fixed sleep mechanisms cannot meet the requirements of low power consumption and emergency low latency. The passive fault handling mechanism of this networking method can easily cause the loss of critical monitoring data.

[0005] Third, although satellite communication networking can achieve full coverage of the open sea, it has high communication costs, high terminal power consumption, and high transmission latency, which cannot support the routine networking of large-scale monitoring terminals and high-frequency data transmission.

[0006] Existing technologies have not achieved a deep integration of the time-varying marine environment with the network architecture. The static segmentation and clustering mode cannot adapt to dynamic channel changes, and the heterogeneous network integration is only a simple backup redundancy without dynamic collaborative access capabilities. It is difficult to cope with the problems of insufficient network stability and coverage caused by the complex time-varying marine environment. Summary of the Invention

[0007] The purpose of this invention is to provide a wireless networking method for marine ecological monitoring terminals to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a wireless networking method for marine ecological monitoring terminals, comprising the following specific steps:

[0009] Preferably, the initialization phase is as follows:

[0010] After all monitoring terminals to be networked are powered on, a four-dimensional holographic calibration of the nodes is completed. The inherent attributes include the node's geographical location, power supply type, monitoring service type, and hardware communication capabilities; the status attributes include the initial remaining energy, the current location's tidal and meteorological real-time data, and the node's real-time load; the channel environment prediction attribute is based on the node's geographical location's historical tidal, wind, wave, and maritime interference data to generate a channel time-varying law model, obtaining the link availability probability under different time periods and meteorological conditions; the heterogeneous network access attribute includes the parameters of all communication standards supported by the node, including cellular, ad hoc, and satellite communication, as well as the access priority.

[0011] Heterogeneous networks are prioritized for access based on the principle of self-organizing networks first, cellular supplementation, and satellite as a backup. Dynamic triggering rules are also set based on channel quality and sea area: Nodes in near-shore high-coverage areas prioritize access to the self-organizing network; when the self-organizing network link health is <0.6, they automatically switch to cellular communication. Nodes in near-shore relay areas use the self-organizing network as the core, with cellular communication used for redundancy in chain-like inter-cluster links; when the self-organizing network link is interrupted, the cellular communication backup link is immediately activated. Nodes in sparsely populated offshore areas prioritize using the self-organizing network for intra-cluster communication; inter-cluster and shore-based backhaul prioritize satellite communication; cellular communication is only attempted when the satellite link fails and the node is within 50km of the shore. All heterogeneous network switching is triggered uniformly by the cluster head, and channel synchronization is completed before switching to ensure no data loss.

[0012] After calibration, each node initiates joint spatiotemporal spectrum mapping and interference prediction in the sea area. It scans and detects the time domain occupancy rate, frequency domain interference intensity, and average signal-to-noise ratio of the channel within the available frequency bands for maritime communication. It identifies the characteristics of sea-specific interference sources such as ship AIS and maritime radar. Combined with the node's geographical location and historical maritime activity data, it generates a spectrum interference prediction map and a local spatiotemporal spectrum database.

[0013] Each node synchronizes its own calibration information and local spectrum data through the control channel, and the shore-based edge gateway aggregates and constructs the initial spatiotemporal spectrum database for the entire network.

[0014] Preferably, the dynamic clustering construction stage is as follows:

[0015] Based on the spatiotemporal spectrum database of the entire network and the holographic calibration data of nodes, the shore-based edge gateway executes a dynamic partitioning and heterogeneous clustering mechanism driven by the time-varying environment. According to the real-time channel quality, node distribution density and service carrying requirements, it dynamically divides three types of areas: nearshore high coverage area, near-shore relay area, and offshore sparse area. The partition boundaries are adjusted in real time according to the changes of tides, wind and waves and ship interference. Each partition executes differentiated clustering rules.

[0016] The nearshore high coverage area adopts a hybrid topology of intra-cluster star and inter-cluster mesh. The basic cluster size ranges from 15 to 25 nodes. During periods of high interference, the cluster size is automatically reduced. The cluster head election adopts a dynamic weight adaptive algorithm. The weight factors include the node's remaining energy, the number of hops to the shore-based gateway, the average signal-to-noise ratio of the channel, and the service carrying capacity. The weight of each factor is dynamically adjusted according to the environmental conditions. Each cluster elects one primary cluster head and a dynamically numbered number of backup cluster heads.

[0017] The near-shore relay area adopts a chain-like clustering and cross-cluster redundancy topology. The chain-like cluster structure is constructed layer by layer along the gradient of offshore distance and the available frequency band of the sea channel. Two cross-system redundant relay links are reserved between adjacent clusters, corresponding to the ad hoc network and cellular communication systems respectively. The cluster head is selected with priority given to nodes with moderate offshore distance, small channel time-varying fluctuations, and strong heterogeneous access capabilities. At the same time, a gateway node is set up for every 3 clusters along the chain path to connect to the shore-based cellular network to achieve link backup and load distribution.

[0018] Load balancing at the gateway node in the near-shore relay area is triggered and allocated based on the data transmission load rate within the cluster. When the load rate of the main transmission link of the chain cluster exceeds 70%, the gateway node immediately triggers the load balancing mechanism, switching the second- and third-level service data exceeding the load threshold to the cellular communication link for transmission, while the first-level service data is always retained for transmission on the primary self-organizing network link. The amount of data balancing for a single chain cluster does not exceed 50% of the total data transmission volume within the cluster to avoid cellular link congestion. The gateway node monitors the load status of the primary self-organizing network link and the cellular link in real time. When the load rate of the main link falls below 50%, it stops balancing and restores data transmission to the main link, ensuring no data duplication or loss during the balancing switch.

[0019] In the sparsely populated offshore area, a redundant relay clustering and satellite backhaul collaborative topology is adopted. The cluster size is dynamically controlled between 3 and 8 nodes according to the node distribution. Each cluster elects one master cluster head and two or more backup cluster heads. Adjacent clusters are bridged through multi-hop relay nodes. The cluster head node integrates a satellite communication module and serves as a backhaul gateway between the offshore area and the shore-based gateway. At the same time, a backup satellite communication link is reserved between adjacent clusters.

[0020] After clustering is completed, the master cluster head broadcasts configuration information, access priority and spectrum allocation rules to the nodes in the cluster, completes topology construction and network-wide synchronization, and synchronizes the clustering results to the shore-based gateway and cloud platform.

[0021] Preferably, the cross-domain routing decision stage is as follows:

[0022] Each cluster initiates route discovery, constructs end-to-end routing paths, and adopts a cross-layer and cross-domain dynamic routing mechanism driven by both channel prediction and service awareness. It collects link parameters and service priorities at each layer, and constructs a three-dimensional joint routing cost function encompassing service, channel, and energy. This three-dimensional joint routing cost function is built by setting dynamically differentiated weights according to service priority. For high-priority emergency alarm services, transmission delay and link redundancy are the core weights, accounting for no less than 60% of the total weight; channel prediction accounts for 30%; and energy consumption accounts for 10%. For medium-priority routine monitoring services, node energy consumption and link stability are the core weights; energy consumption accounts for 50%; channel prediction accounts for 30%; and service consumption accounts for 20%. For low-priority log reporting services, spectrum utilization and network load balancing are the core weights, each accounting for 40%, and energy consumption accounts for 20%.

[0023] The cost function sets a marine-specific multipath fading margin factor and a link availability prediction factor. Based on the spatiotemporal spectrum database, the risk of link interruption is predicted. For links with severe multipath fading and high risk of interruption, the cost value is automatically increased based on the three-dimensional joint routing cost function. At the same time, the path with high line-of-sight transmission probability between nodes is prioritized to adapt to the multipath fading characteristics caused by seawater reflection in the marine area.

[0024] Line-of-sight transmission paths between nodes are quantitatively determined based on three indicators: marine geographical environment, node deployment height, and distance from the shore. In nearshore and offshore areas, a node deployment height ≥ 3m and no islands, reefs, or other obstructions between two nodes are considered line-of-sight transmission paths. In offshore areas, a straight-line distance between two nodes ≤ 5km and a node deployment height ≥ 2m are considered line-of-sight transmission paths. The shore-based gateway pre-generates a map of line-of-sight transmission paths between all network nodes based on the marine geographic information database and the deployment height data of the nodes' holographic calibration. The map is updated in real time as the node positions change. When making routing decisions, line-of-sight paths are selected first from the map. If no line-of-sight path is available, the path with the lowest multipath fading coefficient among the non-line-of-sight paths is selected.

[0025] The path with the lowest cost calculated by the three-dimensional joint routing cost function is selected as the primary transmission path. At the same time, two paths with a node overlap of no more than 30% with the primary path and using different communication standards are selected as backup paths. Based on the channel prediction results, when the risk of primary path link interruption exceeds the warning threshold, low-priority services are switched to backup paths in advance. Finally, the primary and backup routing information is synchronized to all nodes of the path to complete the update and maintenance of the routing table.

[0026] Preferably, the resource scheduling phase is as follows:

[0027] Distributed time-frequency resource scheduling, coordinated by shore-based gateways and implemented by each cluster head in conjunction with spatiotemporal spectrum reuse and service priority, divides monitoring services into three priority levels: high priority for emergency alarms (Level 1), medium priority for routine monitoring (Level 2), and low priority for log reporting (Level 3). Differentiated scheduling is performed based on the network-wide spatiotemporal spectrum database. Channel allocation adopts a spatiotemporal joint dynamic reuse mechanism, dividing the available maritime frequency bands into multiple time-frequency resource blocks. Based on the spectrum interference prediction map, channel spatial reuse and temporal reuse are achieved in different sea areas and at different times.

[0028] Priority services are allocated dedicated low-interference channels. Priority services are allocated available channels with minimal interference in the current time domain and future transmission periods. Priority services are allocated idle non-dedicated channels. Simultaneously, based on ship AIS trajectory prediction, the idle windows of authorized maritime channels are predicted in advance to achieve temporary multiplexing of authorized channel idle periods. When authorized user signals are detected, collision-free avoidance is immediately implemented. Time slot scheduling adopts a TDMA adaptive mechanism, reserving dedicated time slots and emergency resource pools for priority services. Secondary services are allocated adaptive length time slots, and tertiary services are allocated competitive time slots. Low-power nodes in the open sea use time slot aggregation and optimal channel scheduling based on the network-wide spatiotemporal spectrum database. Interference coordination is coordinated by shore-based gateways. In densely populated nearshore areas, channel orthogonality or time slot staggering is used, while in open sea areas, spectrum space multiplexing is used. Finally, the cluster head broadcasts the time-frequency resource scheduling results to nodes within the cluster, completing the coordinated configuration of network-wide resources.

[0029] Preferably, the hibernation control phase is as follows:

[0030] All network nodes implement a three-dimensional collaborative adaptive sleep and wake-up mechanism based on channel, service, and energy. Differentiated sleep rules are executed according to topology roles to achieve intra-cluster and inter-cluster sleep timing coordination. Ordinary nodes within a cluster only wake up in the allocated transmission and control time slots. The sleep period of ordinary nodes within a cluster is adjusted based on service priority, node remaining energy, and channel prediction results. When the channel quality is predicted to be consistently poor based on the node's channel environment prediction attributes, the sleep period is automatically extended. When the channel quality is predicted to be good, the sleep period is automatically shortened to complete data transmission in a concentrated manner.

[0031] For offshore nodes powered by photovoltaic power, the sleep cycle is dynamically adjusted in conjunction with the light intensity prediction data. During the day when there is sufficient sunlight, the sleep cycle is shortened and the reporting frequency is increased, while at night the sleep cycle is extended and the power consumption is reduced. The wake-up windows of the backup cluster head and the main cluster head are staggered by 50%, and adjacent cluster heads are arranged in a staggered manner. The main cluster head is in a shallow sleep state under normal conditions and listens to the control channel. During high-risk periods, the listening frequency is increased. The offshore relay node adopts a relay wake-up method to reduce the duration of continuous listening.

[0032] In an emergency, when the marine ecological parameters monitored by the node approach the warning threshold, it switches to shallow hibernation in advance. After triggering an alarm, it switches to high-frequency wake-up. After the emergency is lifted, it returns to normal. Finally, the cluster head broadcasts the hibernation and wake-up sequence information of the entire network to ensure that the node wake-up window is aligned.

[0033] Preferably, the self-healing reconstruction stage is as follows:

[0034] The entire network nodes monitor link health in real time, and a distributed collaborative self-healing reconstruction mechanism is constructed for digital twin pre-simulation. The shore-based edge gateway collects the node status, link quality, environmental data, and service transmission data of the entire network in real time. A digital twin mirror of the marine network is built in the cloud, which maps the topology status, link health, node energy consumption, and spectrum occupancy status of the entire network in real time. At the same time, based on the prediction data of tides, weather, and ship trajectories, the network status changes are simulated in the digital twin system, and the risks of link interruption, node failure, network islands, and network congestion are predicted in advance, and a hierarchical self-healing reconstruction plan is generated in advance.

[0035] A three-level early warning and fault classification mechanism is set up. Based on the results of digital twin simulation and real-time monitoring data, when the link health is lower than the first-level early warning threshold, the routing cost function weight is adjusted in advance, the path selection and resource allocation are optimized, when it is lower than the second-level early warning threshold, the backup route is activated in advance, the transmission resources are reserved, and the node sleep cycle is adjusted, and when it is lower than the third-level early warning threshold, the topology pre-reconstruction is performed in advance.

[0036] In the early warning state, nodes autonomously execute self-healing plans; in the fault state, cross-cluster distributed reconstruction occurs; when ordinary nodes fail, cluster heads adjust resources and routes; when the primary cluster head fails, backup cluster heads take over; when relay links fail, paths are rebuilt; and remote islands access the backbone network through multi-cluster collaboration. After topology reconstruction is completed, nodes report reconstruction information and update the cloud-based digital twin image and the entire network information.

[0037] Preferably, the optimization management phase is as follows:

[0038] The shore-based gateway collects data from the entire network in real time and uploads it to the cloud. It performs cloud-edge-device collaborative closed-loop optimization driven by federated learning. Based on the collected data, the cloud platform analyzes the channel variation patterns, node energy consumption patterns, and service transmission characteristics in different sea areas, seasons, and environments. It identifies performance bottlenecks such as network congestion areas, high-interference channels, and high-fault-risk nodes, and uses a federated learning framework to build a distributed network optimization model.

[0039] The shore-based edge gateway trains the distributed optimization model locally based on network data of the local sea area, and only uploads the model parameters to the cloud. The cloud aggregates the model parameters of all edge nodes to generate a global optimization model, which is then distributed to each edge node and all network nodes. Based on the global model, the core network parameters are iteratively optimized, and the channel, interference, and route prediction models are iteratively optimized simultaneously to form a sea area-specific network knowledge base.

[0040] The cloud-based digital twin system generates a node operation and maintenance early warning list, simulates node replacement and addition scenarios, and automatically completes initialization after the new node is powered on, synchronizing the network configuration of the entire network of wireless networking of marine ecological monitoring terminals.

[0041] The beneficial effects of this invention are as follows:

[0042] 1. This invention constructs a network-wide spatiotemporal spectrum database through node holographic calibration and joint spatiotemporal spectrum mapping of the sea area. It dynamically divides the sea area into three categories: nearshore, offshore, and offshore, taking into account environmental factors such as tides and waves. Differentiated clustering and topology rules are implemented in each area, and a sea area-specific routing prediction factor is used to adapt to the multipath fading characteristics caused by seawater reflection. At the same time, the offshore area relies on satellite backhaul to coordinate topology and avoid the risk of network islands. The heterogeneous network achieves dynamic collaborative access rather than simple backup, so that the network architecture is deeply integrated with the time-varying environment of the sea area, taking into account the communication needs and transmission stability of different sea areas.

[0043] 2. This invention constructs a three-dimensional joint routing cost function for services, channels, and energy, and sets dynamic weights according to the service priorities of emergency alarms, routine monitoring, and log reporting to ensure low-latency transmission of high-priority services. At the same time, it adopts a spatiotemporal joint distributed time-frequency resource scheduling to allocate dedicated channels and time slots for services of different priorities. Combined with a three-dimensional collaborative sleep mechanism for channels, services, and energy, photovoltaic nodes dynamically adjust their sleep cycles based on sunlight. In emergency situations, nodes are woken up at high frequency, which not only avoids the loss of critical monitoring data but also reduces the ineffective power consumption of nodes, adapting to the characteristics of power supply constraints for offshore nodes.

[0044] 3. This invention uses a cloud-based digital twin mirror to simulate network state changes, sets up a three-level early warning and fault classification mechanism, and realizes pre-fault reconstruction and distributed cross-cluster reconstruction after faults, transforming from passive handling to proactive prediction and avoidance. At the same time, it adopts federated learning to drive cloud-edge-device collaborative optimization. The shore-based gateway trains the model locally and only uploads the parameters. The cloud aggregates and generates a global model and distributes iteratively to optimize the core network parameters and prediction model. It can also generate a node operation and maintenance early warning list. New nodes can be added and used immediately, which reduces bandwidth pressure and realizes network performance optimization throughout the entire life cycle, improving operation and maintenance efficiency. Attached Figure Description

[0045] Figure 1 This is a flowchart of the wireless networking process for the marine ecological monitoring terminal of the present invention;

[0046] Figure 2 This is a flowchart of the network initialization configuration process for this invention;

[0047] Figure 3 This is a flowchart of the fault self-healing process of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] like Figures 1 to 3 As shown, this embodiment of the invention provides a wireless networking method for marine ecological monitoring terminals, including the following specific steps:

[0050] The initialization phase is specifically as follows:

[0051] After all monitoring terminals to be networked are powered on, a four-dimensional holographic calibration of the nodes is completed. The inherent attributes include the node's geographical location, power supply type, monitoring service type, and hardware communication capabilities; the status attributes include the initial remaining energy, real-time tidal / meteorological data of the current location, and the node's real-time load; the channel environment prediction attribute is based on the historical tidal, wind, wave, and maritime interference data of the node's geographical location to generate a time-varying channel law model and obtain the link availability probability under different time periods and different meteorological conditions; the heterogeneous network access attribute includes the parameters of all communication standards supported by the node, such as cellular, ad hoc network, and satellite communication, as well as the access priority.

[0052] The channel time-varying law model is a time-series prediction model. The core modules include a marine environment feature extraction layer, a time-series prediction layer, and a probability output layer. The feature extraction layer takes the node's geographical location and historical tide / wave / maritime interference data as input and outputs core features such as channel attenuation and signal-to-noise ratio changes. The time-series prediction layer adopts an LSTM network structure with 64 hidden layer units, a time step of 24 hours, 100 training iterations, and a learning rate of 0.001. The probability output layer outputs the link availability probability under different time periods and meteorological conditions based on the sigmoid function. The model training uses local historical monitoring data of the nodes as the training set, and divides the training, validation, and test sets in a 7:2:1 ratio. Model training is completed when the validation set loss value is less than 0.05.

[0053] The node holographic calibration process involves two steps: automatic verification and shore-based review. First, the terminal locally completes the collection and self-verification of attribute data for each dimension. The data collection error for inherent attributes and status attributes must be ≤5%, the output deviation of the channel environment prediction attribute model must be ≤10%, and the heterogeneous network access attribute must complete the connectivity test of all communication parameters. After the local verification is passed, the terminal uploads the calibration data to the shore-based edge gateway through the control channel. The gateway reviews the uniqueness and rationality of the calibration data of all nodes in the network. After the review is passed, the node holographic calibration is completed. Terminals that fail the review trigger the recalibration process, with a maximum of 3 recalibrations. Terminal nodes that still fail the recalibration after 3 recalibrations are marked as faulty nodes, an operation and maintenance warning is generated, and network access is denied. The calibration process will be re-executed after manual investigation.

[0054] After calibration, each node initiates joint spatiotemporal spectrum mapping and interference prediction in the sea area. It scans and detects the time domain occupancy rate, frequency domain interference intensity, and average signal-to-noise ratio of the channel within the available frequency bands for maritime communication. It identifies the characteristics of sea-specific interference sources such as ship AIS and maritime radar. Combined with the node's geographical location and historical maritime activity data, it generates a spectrum interference prediction map for the next 24 hours and a local spatiotemporal spectrum database.

[0055] Each node synchronizes its own calibration information and local spectrum data through the control channel, and the shore-based edge gateway aggregates and constructs the initial spatiotemporal spectrum database for the entire network.

[0056] The spectrum interference prediction model is a spatiotemporal fusion prediction model, consisting of a spatial feature module, a temporal prediction module, and a spectrum generation module. The spatial feature module takes into account the latitude and longitude of nodes and the distribution characteristics of sea area interference sources. The temporal prediction module takes into account the channel temporal occupancy rate and historical data of frequency domain interference intensity. The model training batch size is 32, the regularization coefficient is 0.0001, and the Adam optimizer is used. The spectrum generation module fuses the spatiotemporal features and outputs a spectrum interference prediction spectrum with a spatial granularity of 1km and a temporal granularity of 1h. The spectrum contains the mapping relationship between interference intensity, interference duration, and interference source type for each frequency band.

[0057] The dynamic clustering construction stage is specifically as follows:

[0058] Based on the spatiotemporal spectrum database of the entire network and the holographic calibration data of nodes, the shore-based edge gateway executes a dynamic partitioning and heterogeneous clustering mechanism driven by the time-varying environment. According to the real-time channel quality, node distribution density and service carrying requirements, it dynamically divides three types of areas: nearshore high coverage area, near-shore relay area, and offshore sparse area. The partition boundaries are adjusted in real time according to the changes of tides, wind and waves and ship interference. Each partition executes differentiated clustering rules.

[0059] The three types of sea areas are dynamically determined based on three quantitative indicators: distance from shore, node distribution density, and average signal-to-noise ratio (SNR). Nearshore high-coverage areas: distance from shore ≤ 10km, node density ≥ 5 / km², average SNR ≥ 20dB; Nearshore relay areas: distance from shore 10~50km, node density 1~5 / km², average SNR 10~20dB; Farshore sparse areas: distance from shore > 50km, node density < 1 / km², average SNR < 10dB. The shore-based edge gateway collects time-varying data on node density and average SNR every 30 minutes. When the changes in these indicators exceed ±20% of the above thresholds, real-time iterative adjustments to the zone boundaries are triggered. Distance from shore, as a geographical benchmark, is only re-determined when the node deployment location changes.

[0060] The nearshore high coverage area adopts a hybrid topology of intra-cluster star and inter-cluster mesh. The cluster size ranges from 15 to 25 nodes. During periods of high interference, the cluster size is automatically reduced to reduce transmission conflicts. The cluster head election adopts a dynamic weight adaptive algorithm. The weight factors include the node's remaining energy, the number of hops to the shore-based gateway, the average signal-to-noise ratio of the channel, and the service carrying capacity. The weight of each factor is dynamically adjusted according to the environmental conditions. Each cluster elects one primary cluster head and a dynamically numbered number of backup cluster heads.

[0061] The dynamic weight adaptive algorithm is based on a multi-factor weight calculation model. The model input consists of four features: node remaining energy, hop count to shore-based gateway, average signal-to-noise ratio (SNR) of the channel, and service carrying capacity. These features are normalized to the [0, 1] interval. The model sets an environment adaptation coefficient. Under high interference, the SNR weight coefficient is increased to 0.35; under low energy, the node remaining energy weight coefficient is increased to 0.4; and under low load, the service carrying capacity weight coefficient is decreased to 0.15. The model outputs a comprehensive score for node cluster head election. The node with the highest score is the primary cluster head, and the nodes with the highest scores (2-4) are backup cluster heads. When the number of nodes in a cluster is less than 10, only one backup cluster head is retained. When the number of nodes in a cluster is 10-15, two backup cluster heads are retained. When the number of nodes in a cluster is 15 or more, three to four backup cluster heads are retained, and the algorithm is dynamically adjusted according to the cluster size.

[0062] The near-shore relay area adopts a chain-like clustering and cross-cluster redundancy topology. The chain-like cluster structure is constructed layer by layer along the gradient of offshore distance and the available frequency bands of the sea channel. Two cross-system redundant relay links are reserved between adjacent clusters, corresponding to ad hoc networks and cellular communication systems, respectively. Cluster heads are selected with nodes that are moderately offshore, have low channel time-varying fluctuations, and strong heterogeneous access capabilities. At the same time, a gateway node is set up for every 3 clusters along the chain path to connect to the shore-based cellular network to achieve link backup and load distribution. The cluster size of the near-shore relay area is dynamically adjusted according to the offshore distance and channel quality, with a baseline range of 8 to 15 nodes. The cluster size of the near-shore relay area is dynamically adjusted with the channel interference intensity. During high interference periods (channel signal-to-noise ratio <12dB), the cluster size is reduced to 8 to 10 nodes, while during low interference periods, it is maintained at 10 to 15 nodes.

[0063] The cross-cluster redundant relay links in the near-shore relay area implement daily maintenance rules of timed detection, fault marking, and immediate repair. This is completed collaboratively by the cluster heads of adjacent clusters. Every 15 minutes, connectivity and transmission quality are checked on two cross-standard links. The detection indicators include link signal-to-noise ratio, packet loss rate, and transmission latency. After the detection, if the link health score is ≥0.7, it is considered to be in a normal state. If the health score is between 0.5 and 0.7, it is marked as a low-risk fault and the link resource configuration is immediately adjusted and optimized. If the health score is <0.5, it is marked as a high-risk fault and the link repair mechanism is immediately activated. The cluster head renegotiates with the relay nodes to rebuild the link. The detection results and fault handling status are uploaded to the shore-based gateway and the cloud digital twin mirror in real time to ensure that the redundant links are always available.

[0064] In the sparsely populated offshore area, a redundant relay clustering and satellite backhaul collaborative topology is adopted. The cluster size is dynamically controlled between 3 and 8 nodes according to the node distribution. Each cluster elects one master cluster head and two or more backup cluster heads. Adjacent clusters are bridged through multi-hop relay nodes. The maximum number of hops for multi-hop relays between clusters in the sparsely populated offshore area does not exceed 3 hops to ensure controllable link transmission latency. The cluster head node integrates a satellite communication module and serves as a backhaul gateway between the offshore area and the shore-based gateway. At the same time, a backup satellite communication link is reserved between adjacent clusters to avoid the risk of network silos in the offshore area.

[0065] After clustering is completed, the master cluster head broadcasts configuration information, access priority and spectrum allocation rules to the nodes in the cluster, completes topology construction and network-wide synchronization, and synchronizes the clustering results to the shore-based gateway and cloud platform.

[0066] The satellite backhaul collaborative topology is implemented according to the process of primary link optimization and backup link standby. The satellite communication module of the main cluster head is the main backhaul channel within the cluster, and the maritime dedicated satellite frequency band is used first to complete the data backhaul. The data of all nodes in the cluster is first aggregated to the main cluster head, and then the main cluster head performs data aggregation and backhauls to the shore-based gateway via satellite. The satellite communication module of the backup cluster head is in a low-power standby state. When the satellite link health of the main cluster head is <0.5, the backup cluster head is immediately triggered to take over the backhaul task. The satellite communication backup link of the adjacent cluster is the cross-cluster backhaul channel. When the satellite link of a certain cluster fails completely, the data of that cluster is transmitted to the adjacent cluster head through multi-hop relay nodes and uses its satellite link to complete the backhaul. The satellite link switching time is ≤500ms.

[0067] All cluster heads across the entire sea area employ a hierarchical aggregation rule based on business priority to process data within their clusters. Level 1 emergency alarm data is not aggregated; upon receipt by the cluster head, it is immediately transmitted to the shore-based / satellite backhaul channel to ensure transmission latency. Level 2 routine monitoring data is aggregated in batches using similar indicators, with a time granularity of 1 minute. This integrates and packages monitoring data of the same type from nodes within the cluster before transmission, reducing the number of transmission frames. Level 3 log reporting data is aggregated in large batches, with a time granularity of 30 minutes. This compresses and aggregates log data from all nodes within the cluster before transmission, reducing transmission power consumption. All aggregated data is marked with the data source node identifier and collection timestamp to ensure data traceability. The data packet loss rate after aggregation and compression must be ≤0.1%.

[0068] The cross-domain routing decision-making stage is specifically as follows:

[0069] Each cluster initiates route discovery, constructs end-to-end routing paths, and adopts a cross-layer and cross-domain dynamic routing mechanism driven by both channel prediction and service awareness. It collects link parameters and service priorities at each layer, and constructs a three-dimensional joint routing cost function encompassing service, channel, and energy. This three-dimensional joint routing cost function is built by setting dynamically differentiated weights according to service priority. For high-priority emergency alarm services, transmission delay and link redundancy are the core weights, accounting for no less than 60% of the total weight; channel prediction accounts for 30%; and energy consumption accounts for 10%. For medium-priority routine monitoring services, node energy consumption and link stability are the core weights; energy consumption accounts for 50%; channel prediction accounts for 30%; and service consumption accounts for 20%. For low-priority log reporting services, spectrum utilization and network load balancing are the core weights, each accounting for 40%, and energy consumption accounts for 20%.

[0070] The cost function sets a marine-specific multipath fading margin factor and a link availability prediction factor. Based on the spatiotemporal spectrum database, it predicts the link interruption risk within the next hour. For links with severe multipath fading and high interruption risk, the cost value is automatically increased based on the three-dimensional joint routing cost function, reducing the probability of being selected. At the same time, it prioritizes paths with high line-of-sight transmission probability between nodes, adapting to the multipath fading characteristics caused by seawater reflection in the marine area.

[0071] The link availability duration prediction factor is generated by the routing prediction model, which is a regression prediction model specifically adapted for marine areas. The core layers include a feature input layer, a factor calculation layer, and a result output layer. The input layer consists of link multipath fading coefficient, channel signal-to-noise ratio, and marine environmental disturbance data from the spatiotemporal spectrum database. The factor calculation layer sets a seawater reflection attenuation coefficient correction value of 0.85 and calculates the link availability duration based on a linear regression algorithm. The model is trained using measured link data from the marine area, and training is completed when the goodness of fit R² ≥ 0.9. The output link availability duration prediction factor and multipath fading margin factor are incorporated into the routing cost function with a 1:1 weight.

[0072] The path with the lowest cost calculated by the three-dimensional joint routing cost function is selected as the primary transmission path. At the same time, two paths with no more than 30% overlap with the primary path nodes and using different communication standards are selected as backup paths. Based on the channel prediction results, when the risk of primary path link interruption exceeds the warning threshold, low-priority services are switched to backup paths in advance to achieve seamless link switching. Finally, the primary and backup routing information is synchronized to all nodes of the path to complete the update and maintenance of the routing table.

[0073] The resource scheduling phase is specifically as follows:

[0074] Distributed time-frequency resource scheduling, coordinated by shore-based gateways and implemented by each cluster head in conjunction with spatiotemporal spectrum reuse and service priority, divides monitoring services into three levels according to priority: high priority for emergency alarms, medium priority for routine monitoring, and low priority for log reporting. Differentiated scheduling is performed based on the network-wide spatiotemporal spectrum database. Channel allocation adopts a spatiotemporal joint dynamic reuse mechanism, dividing the available maritime frequency bands into multiple time-frequency resource blocks. Based on the spectrum interference prediction map, channel spatial reuse and temporal reuse are achieved in different sea areas and at different times.

[0075] The available maritime frequency bands are divided into standardized time-frequency resource blocks with a frequency domain granularity of 1MHz and a time domain granularity of 10ms. Each resource block contains a unique frequency domain number and a time domain number. The shore-based gateway marks all resource blocks with interference levels, classifying them into three categories: low interference (interference intensity < -80dBm), medium interference (interference intensity -80~-60dBm), and high interference (interference intensity > -60dBm). The interference level of the resource block is updated every 10 minutes based on the spectrum interference prediction map. High interference resource blocks do not participate in the channel allocation for primary and secondary priority services. When there are insufficient low / medium interference resource blocks, the shore-based gateway coordinates cross-regional resource scheduling, temporarily allocating low interference resource blocks for primary services, while secondary services use time slot staggered multiplexing of medium interference resource blocks.

[0076] The three-tiered monitoring services are quantitatively defined based on the type of marine ecological monitoring indicators and the degree of data anomalies. The first tier includes emergency alarm services: data exceeding standards such as water quality pH < 6.5 or > 8.5, water temperature sudden change of ≥ 5℃ within 1 hour, detection of oil spill / red tide signals, dissolved oxygen < 3 mg / L, and network alarm data such as terminal failure and link interruption. This type of data carries a dedicated first-tier priority identifier and is transmitted immediately after terminal collection. The second tier includes routine monitoring services: real-time monitoring data of water quality indicators such as pH, water temperature, and dissolved oxygen within normal ranges, and routinely collected data such as marine meteorological and hydrological current data, carrying a second-tier priority identifier and reported according to a preset cycle. The third tier includes log reporting services: non-real-time data such as terminal operating parameters, network communication logs, and data collection records, carrying a third-tier priority identifier and reported using a batch aggregation method.

[0077] First-priority services are allocated dedicated low-interference channels to avoid transmission conflicts. Second-priority services are allocated available channels with the least interference in the current time domain and future transmission periods. Third-priority services are allocated idle non-dedicated channels. At the same time, based on ship AIS trajectory prediction, the idle window of the authorized maritime channel is predicted in advance to realize the temporary multiplexing of the idle period of the authorized channel. When the authorized user signal is detected, it immediately avoids conflict.

[0078] The authorized maritime channel temporary reuse process involves four steps: prediction, application, reuse, and avoidance. First, the shore-based gateway predicts the available free window of the authorized channel 30 minutes in advance based on the ship's AIS trajectory, marking the free period and channel number. Second, the cluster head submits an authorized channel reuse application to the shore-based gateway according to the three-level service transmission requirements of its cluster. After the gateway approves the application, it allocates the corresponding free window. Then, the node completes the three-level service data transmission within the designated free window, continuously monitoring the authorized user signal of the authorized channel during the transmission process. Finally, when the authorized user signal strength is detected to be ≥-70dBm, a collision-free avoidance is immediately triggered, data transmission is stopped, and the node switches to a preset free non-dedicated channel. The avoidance response time is ≤10ms. Data that has not been transmitted is buffered and will continue to be transmitted after the next free window or channel switch.

[0079] Time slot scheduling adopts a TDMA adaptive mechanism, reserving dedicated time slots and emergency resource pools for Tier 1 services, allocating adaptive length time slots for Tier 2 services, and allocating contention-based time slots for Tier 3 services. Low-power nodes in the open sea use time slot aggregation and channel-optimal scheduling based on the network-wide spatiotemporal spectrum database. Interference coordination is coordinated by shore-based gateways. In densely populated nearshore areas, channel orthogonality or time slot staggering is used, while in the open sea areas, spectrum space reuse is used to improve spectrum utilization. Finally, the cluster head broadcasts the time-frequency resource scheduling results to nodes within the cluster, completing the coordinated configuration of network-wide resources.

[0080] The emergency resource pool is configured with 15% of the available maritime frequency band time-frequency resource blocks, selected from low-interference resource blocks. It includes dedicated frequency domain resource blocks and spare time slot resources. The resource pool does not participate in the resource allocation for Tier 2 and Tier 3 services and is always in a reserved standby state. The emergency resource pool adopts on-demand scheduling and dynamic replenishment rules. When Tier 1 service data is detected, the cluster head immediately schedules resources in the resource pool to complete the transmission. After the transmission is completed, the reserved state is restored. If a new Tier 1 service is triggered when resources in the resource pool are occupied, the shore-based gateway immediately temporarily allocates resources from low-interference resource blocks in the surrounding area to ensure that there are no resources waiting for Tier 1 services.

[0081] The time slot aggregation of the offshore low-power nodes is based on five adjacent basic time domain time slots (50ms) as an aggregation unit, integrating scattered small-capacity time slots into continuous large-capacity transmission time slots. Aggregated time slots are only allocated to the second and third-level services of the offshore nodes. Time slot aggregation is uniformly planned by the cluster head and differentiated according to the node's remaining energy and data buffer size. Nodes with remaining energy > 60% and data buffer size > 100KB are given priority for aggregation time slot allocation. The channel matching of the aggregation time slots is the optimal low-interference channel predicted in the spatiotemporal spectrum database of the entire network, and the aggregation time slots are replanned every 30 minutes according to the channel status to ensure transmission efficiency.

[0082] The specific details of the hibernation control phase are as follows:

[0083] All network nodes implement a three-dimensional collaborative adaptive sleep and wake-up mechanism that coordinates channel, service, and energy. Differentiated sleep rules are executed according to topology roles to achieve intra-cluster and inter-cluster sleep timing coordination. Ordinary nodes within a cluster are only woken up in the allocated transmission and control time slots, and enter deep sleep for the rest of the time. The sleep period of ordinary nodes within a cluster is adjusted based on service priority, node remaining energy, and channel prediction results. If the channel quality is predicted to be poor for the next 4 hours based on the node's channel environment prediction attributes, the sleep period is automatically extended to reduce invalid wake-ups and retransmissions. If the channel quality is predicted to be good for the next 1 hour, the sleep period is automatically shortened to concentrate on completing data transmission.

[0084] The sleep period for ordinary nodes within the cluster is set to five basic levels: 10s, 30s, 1min, 5min, and 30min. The level is dynamically adjusted according to three-dimensional indicators: when the remaining energy of the node is >80%, the sleep period is set to a low level (10s / 30s); when the remaining energy is 50%~80%, the sleep period is set to a medium level (1min / 5min); when the remaining energy is <50%, the sleep period is set to a high level (30min). If the channel quality is predicted to be good and there are Level 1 / Level 2 services, the sleep period is reduced by 1 to 2 levels based on the above. If the channel quality is predicted to remain poor and only Level 3 services exist, increase the level by 1-2 levels based on the above. For offshore nodes powered by photovoltaics, decrease the level by 1 level when the daytime illuminance is >20000lx and increase the level by 1 level when the nighttime illuminance is <5000lx. Keep the current dormant level unchanged when the illuminance is 5000-20000lx. Define the day / night criteria according to the local sunrise and sunset times or the illuminance ≥10000lx as daytime and <10000lx as nighttime.

[0085] For offshore nodes powered by photovoltaic power, the sleep cycle is dynamically adjusted in conjunction with the light intensity prediction data. During the day when there is sufficient sunlight, the sleep cycle is shortened and the reporting frequency is increased, while at night the sleep cycle is extended and the power consumption is reduced. The wake-up windows of the backup cluster head and the main cluster head are staggered by 50%, and adjacent cluster heads are arranged in a staggered manner. The main cluster head is in a shallow sleep state under normal conditions and listens to the control channel. During high-risk periods, the listening frequency is increased. The offshore relay node adopts a relay wake-up method to reduce the duration of continuous listening.

[0086] In an emergency, when the marine ecological parameters monitored by the node approach the warning threshold, it switches to shallow hibernation in advance. After triggering an alarm, it switches to high-frequency wake-up. After the emergency is lifted, it returns to normal. Finally, the cluster head broadcasts the hibernation and wake-up sequence information of the entire network to ensure that the node wake-up window is aligned and to avoid data loss.

[0087] In emergency situations, high-frequency wake-up adopts a quantization rule of 1-second wake-up interval and 50ms listening time. Nodes wake up once every second and continuously listen to the control channel and data transmission channel to ensure real-time collection and transmission of emergency data. The termination of the emergency state is determined uniformly by the cluster head. When the ecological indicators monitored by the node return to the normal range and the network alarm fault is resolved, the cluster head sends an emergency termination command. After receiving the command, the node gradually restores to the original dormancy cycle within 1-3 minutes according to the remaining energy status to avoid sudden power consumption surges. During high-frequency wake-up, photovoltaic power supply nodes prioritize the use of energy storage power, while non-photovoltaic nodes activate power consumption current limiting to ensure the operation of only the monitoring and communication core modules.

[0088] The self-healing reconstruction stage is specifically as follows:

[0089] The entire network nodes monitor link health in real time, and a distributed collaborative self-healing reconstruction mechanism based on digital twin simulation is constructed to achieve proactive prediction and avoidance. The shore-based edge gateway collects the node status, link quality, environmental data, and service transmission data of the entire network in real time, and constructs a digital twin mirror of the marine network in the cloud. It maps the topology status, link health, node energy consumption, and spectrum occupancy status of the entire network in real time at a 1:1 scale. At the same time, based on the prediction data of tides, weather, and ship trajectories, the network status changes in the next 12 hours are simulated in the digital twin system, and the risks of link interruption, node failure, network islands, and network congestion are predicted in advance, and hierarchical self-healing reconstruction plans are generated in advance.

[0090] The fault risk prediction algorithm is based on a marine-specific fault classification model. The model input includes link health, node remaining energy, spectrum occupancy, and marine environmental disturbance data from a digital twin mirror. The output is the probability of occurrence and risk level of four types of faults. The model adopts a lightweight CNN+SVM structure. The CNN layer extracts fault features, and the SVM layer performs binary classification of risk levels. Both model training and testing use real-world marine network fault data. The fault identification accuracy on the test set is ≥95%. The risk level is bound to the link health threshold: Level 1 warning (low risk) link health ∈ [0.7, 1], Level 2 warning (medium risk) ∈ [0.5, 0.7), Level 3 warning (high risk) ∈ [0.3, 0.5], and fault state <0.3. The risk level output by the model is directly mapped to the corresponding self-healing reconstruction plan.

[0091] Link health is quantified and calculated based on four core indicators: channel signal-to-noise ratio (SNR), link packet loss rate, transmission delay, and link connectivity. The weights of the four indicators are 0.3, 0.3, 0.2, and 0.2, respectively. Each indicator is first normalized to the range [0, 1], and then the weighted sum is obtained to obtain the link health value, which ranges from [0, 1]. Specifically, a channel SNR ≥ 20dB scores 1 point, < 10dB scores 0 points; a link packet loss rate < 1% scores 1 point, > 10% scores 0 points; a transmission delay < 100ms scores 1 point, > 500ms scores 0 points; and a link connectivity ≥ 99% scores 1 point, < 90% scores 0 points. Nodes collect indicator data every 5 seconds and calculate the average link health value every 10 minutes, then upload it to the shore-based gateway.

[0092] A three-tiered early warning and fault classification mechanism is set up. Based on the results of digital twin simulation and real-time monitoring data, when the link health is lower than the first-level early warning threshold (low risk), the routing cost function weight is adjusted in advance, the path selection and resource allocation are optimized, and when it is lower than the second-level early warning threshold (medium risk), the backup route is activated in advance, the transmission resources are reserved, and the node sleep cycle is adjusted, and when it is lower than the third-level early warning threshold (high risk), the topology pre-reconstruction is performed in advance. For example, when it is predicted that a relay node is about to fail, a new relay node is elected in advance, and the routing and resource pre-configuration is completed. When the link health is lower than the preset fault threshold and the remaining energy of the node is lower than the power supply critical value, a fault state is triggered.

[0093] In the early warning state, nodes autonomously execute self-healing plans; in the fault state, cross-cluster distributed reconstruction occurs; when ordinary nodes fail, cluster heads adjust resources and routes; when the primary cluster head fails, backup cluster heads take over; when relay links fail, paths are rebuilt; and remote islands access the backbone network through multi-cluster collaboration. After topology reconstruction is completed, nodes report reconstruction information and update the cloud-based digital twin image and the entire network information.

[0094] In fault scenarios, the retransmission of lost monitoring data adopts a hierarchical caching, breakpoint resumption, and cross-cluster retransmission implementation process. Nodes set up local data caches, with first-level business data cached for ≥24 hours and second- and third-level business data cached for ≥12 hours. All data collected during the fault is stored in the cache. After topology reconstruction and link recovery, nodes first detect lost data in their local caches and initiate breakpoint resumption requests according to business priority from high to low. The cluster head allocates a dedicated temporary time slot for the retransmission data. If the node's own link still has transmission limitations, the cluster head coordinates with cross-cluster relay nodes to complete the retransmission of lost data via cross-cluster links. After retransmission, the node clears the transmitted data from its cache. Data that failed to be retransmitted will be retried, with a maximum of 3 retries. After a retransmission failure, retries are performed at 1-minute, 2-minute, and 5-minute intervals. After 3 failed retries, a data retransmission fault warning is generated and added to the node's maintenance checklist.

[0095] The optimization management phase is specifically as follows:

[0096] The shore-based gateway collects data from the entire network in real time and uploads it to the cloud. It performs cloud-edge-device collaborative closed-loop optimization driven by federated learning. Based on the collected data, the cloud platform analyzes the channel variation patterns, node energy consumption patterns, and service transmission characteristics in different sea areas, seasons, and environments. It identifies performance bottlenecks such as network congestion areas, high-interference channels, and high-fault-risk nodes, and uses a federated learning framework to build a distributed network optimization model.

[0097] The federated learning-driven distributed networking optimization model has a layered architecture, including a local model layer at shore-based edge nodes and a global model layer in the cloud. Both layers adopt a fully connected neural network structure. The inputs are marine channel parameters, node energy consumption data, and service transmission characteristics. The outputs are networking cluster size, routing cost weights, time-frequency resource allocation coefficients, and other networking optimization parameters. The training steps for the local model layer are as follows:

[0098] 1. The shore-based gateway collects network data from the sea area under its jurisdiction as a local dataset, trains it with a batch size of 64, a learning rate of 0.001, and 50 iterations;

[0099] 2. Calculate the gradient of the model parameters, upload only the gradient and model weight parameters to the cloud, and keep the raw data locally;

[0100] 3. The cloud receives all edge node parameters and aggregates them using the FedAvg algorithm. The aggregation weight is distributed according to the proportion of the number of sea area nodes under the jurisdiction of each shore-based node.

[0101] 4. A global model is generated in the cloud and distributed to each shore-based edge node and all network nodes. The local model is fine-tuned based on the global model until the validation set loss converges, completing one iteration of optimization. The iteration cycle of the global model is 24 hours.

[0102] The shore-based edge gateway trains the distributed optimization model locally based on network data of the local sea area, and only uploads the model parameters to the cloud without transmitting the original monitoring data, which greatly reduces the bandwidth pressure. The cloud aggregates the model parameters of all edge nodes to generate a global optimization model, and then distributes it to each edge node and all network nodes. Based on the global model, the core network parameters are iteratively optimized, and the channel, interference, and route prediction models are iteratively updated simultaneously to form a sea area-specific network knowledge base, thereby improving network performance.

[0103] The cloud-based digital twin system generates a node operation and maintenance early warning list, simulates node replacement and addition scenarios, and automatically completes initialization after the new node is powered on, synchronizing the network configuration of the entire network of the marine ecological monitoring terminal wireless networking, achieving plug-and-play functionality.

[0104] The automatic network access process for new nodes involves five steps: initialization, discovery, authentication, configuration, and network access.

[0105] 1. After a new node is powered on, it automatically completes four-dimensional holographic calibration and generates a unique identifier for the node;

[0106] 2. The node sends a network discovery broadcast through the control channel. After receiving the broadcast, the surrounding cluster heads report the node information to the shore-based gateway.

[0107] 3. The shore-based gateway performs identity security verification on newly added nodes. After successful verification, it assigns the node to a cluster and topology role based on the node's geographical location and sea area partitioning results.

[0108] 4. The gateway distributes the network configuration, including the spatiotemporal spectrum database of the entire network, the clustering rules of the cluster to which it belongs, the service priority rules, the routing configuration, the hibernation rules, etc., to the newly added nodes, and the nodes complete the local configuration synchronization;

[0109] 5. New nodes establish communication links with their cluster head and neighboring nodes. The cluster head synchronizes the new node information to the cloud digital twin mirror and other nodes in the network to complete the automatic network access. The entire network access process takes ≤5 minutes and requires no manual intervention.

[0110] The node operation and maintenance early warning list includes six core components: unique identifier of the warning node, node geographical location, warning type, risk level, recommended handling measures, and warning validity period. Warning types are divided into four categories: low energy warning, high-risk link failure warning, hardware anomaly warning, and continuous channel interference warning. Risk levels are categorized as high, medium, and low. The list is dynamically generated every 24 hours by a cloud-based digital twin system, based on real-time node monitoring data and fault risk prediction algorithms. Nodes are automatically added to the list when: remaining energy < 20%, link health < 0.4 for one consecutive hour, hardware operating parameters exceed normal range, or channel interference intensity > -60dBm for two consecutive hours. High-risk warning nodes are pinned to the top of the list, and the system pushes real-time alerts to the operation and maintenance team. Warning nodes are automatically removed from the list after the fault is resolved or the indicators return to normal.

[0111] High-risk warnings are valid for 72 hours, medium-risk warnings for 48 hours, and low-risk warnings for 24 hours. If the indicators return to normal within the validity period, the warning will be removed from the list. If the warning is not addressed within the validity period, the warning level will be upgraded.

[0112] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0113] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A wireless networking method for marine ecological monitoring terminals, characterized in that, include: During the initialization phase, after the network monitoring terminal is powered on, the node holographic calibration is completed. Each node performs joint spatiotemporal spectrum mapping and interference prediction in the sea area and synchronizes relevant data. The shore-based edge gateway aggregates and constructs the initial spatiotemporal spectrum database of the entire network. During the dynamic clustering construction phase, the shore-based edge gateway performs dynamic partitioning and heterogeneous clustering based on the full network spatiotemporal spectrum database and node holographic calibration data. It divides the sea area into regions according to channel quality, node density and service requirements and dynamically adjusts the boundaries. Each region completes topology construction according to differentiated clustering rules, and the clustering results are synchronized to the shore-based gateway and cloud platform. During the cross-domain routing decision-making phase, each cluster head initiates route discovery and constructs an end-to-end routing path. It adopts a dynamic routing mechanism driven by both channel prediction and service awareness, constructs a routing cost function based on service priority, sets a sea area-specific prediction factor, selects primary and backup transmission paths, and synchronizes routing information to relevant nodes. During the resource scheduling phase, the shore-based gateway coordinates and each cluster head performs distributed time-frequency resource scheduling, prioritizes services, and performs differentiated scheduling based on the network-wide spectrum database to complete the coordinated configuration of network-wide resources. During the hibernation control phase, all network nodes execute differentiated hibernation rules according to their topology roles, and dynamically adjust the hibernation period based on service priority, energy status, and channel prediction to achieve intra-cluster and inter-cluster timing coordination. During the self-healing and reconstruction phase, all network nodes monitor link health in real time, construct a collaborative self-healing and reconstruction mechanism based on digital twin pre-simulation, predict fault risks and set up hierarchical self-healing contingency plans to achieve pre-fault reconstruction and post-fault distributed reconstruction. During the optimization and management phase, the shore-based gateway collects data from the entire network and uploads it to the cloud. It performs closed-loop optimization of the entire lifecycle of cloud-edge-device collaboration, iteratively optimizes the core network parameters and related prediction models, and generates a list of node operation and maintenance early warnings.

2. The wireless networking method for marine ecological monitoring terminals according to claim 1, characterized in that, In the initialization phase, the node holographic calibration is a four-dimensional calibration, including the node's inherent attributes, state attributes, channel environment prediction attributes, and heterogeneous network access attributes. Each node completes spectrum scanning and interference detection within the available frequency bands for maritime communication, identifies the characteristics of interference sources unique to the sea area, and generates a spectrum interference prediction map and a local spatiotemporal spectrum database. Then, the calibration information and local spectrum data are synchronized to the shore-based edge gateway through the control channel, and the shore-based edge gateway aggregates them to form the initial spatiotemporal spectrum database of the entire network.

3. The wireless networking method for marine ecological monitoring terminals according to claim 2, characterized in that, In the dynamic clustering construction phase, the shore-based edge gateway adopts a time-varying environment-driven dynamic partitioning and heterogeneous clustering mechanism to dynamically divide the sea area into three types of regions: nearshore high coverage area, nearshore relay area, and offshore sparse area. The partition boundaries are adjusted in real time according to changes in the sea environment and interference. Differentiated clustering rules are implemented in each region: the nearshore high coverage area adopts a hybrid topology of intra-cluster star and inter-cluster mesh and dynamically adjusts the cluster size, electing a master cluster head and a backup cluster head; the nearshore relay area adopts chain clustering and cross-cluster redundancy topology, constructs a chain cluster structure along the offshore distance gradient and reserves cross-system redundant relay links, and selects suitable nodes as cluster heads and gateway nodes; the offshore sparse area adopts redundant relay clustering and satellite backhaul collaborative topology, dynamically controls the cluster size and elects master and backup cluster heads, and achieves inter-cluster bridging through multi-hop relay nodes. The master cluster head integrates a satellite communication module and reserves a satellite backup link, while the backup cluster head is configured with a satellite communication module in a low-power standby state.

4. The wireless networking method for marine ecological monitoring terminals according to claim 3, characterized in that, In the cross-domain routing decision-making stage, a cross-layer and cross-domain dynamic routing mechanism driven by both channel prediction and service awareness is adopted. Link parameters and service priority information are collected, and a three-dimensional joint routing cost function of service, channel, and energy is constructed. Dynamic differentiated weights are set according to service priority. The routing cost function sets a marine-specific multipath fading margin factor and a link availability prediction factor, and predicts the risk of link interruption based on the spatiotemporal spectrum database. The path with the lowest cost is selected as the primary transmission path, and the path with node overlap not exceeding a preset threshold and different communication standards is selected as the backup path. When the risk of primary path interruption exceeds the threshold, low-priority services are switched in advance, and primary and backup routing information is synchronized and the routing table is updated and maintained.

5. The wireless networking method for marine ecological monitoring terminals according to claim 4, characterized in that, During the resource scheduling phase, monitoring services are divided into three priority levels. A spatiotemporal joint dynamic multiplexing mechanism is used to allocate channels. The available maritime frequency bands are divided into multiple time-frequency resource blocks. Spatiotemporal multiplexing of channels is achieved based on the spectrum interference prediction map. Corresponding channels are allocated according to service priorities, and temporary multiplexing and conflict-free avoidance of the licensed channel idle window are implemented. A TDMA adaptive mechanism is used to schedule time slots, allocating dedicated, adaptive, and competitive differentiated time slots for services with different priorities. Time slot aggregation and optimal channel scheduling are used for low-power nodes in the open sea. The shore-based gateway coordinates interference, and each cluster head broadcasts the scheduling results to nodes within the cluster to complete the coordinated configuration of network resources.

6. The wireless networking method for marine ecological monitoring terminals according to claim 5, characterized in that, During the hibernation control phase, all network nodes adopt a three-dimensional collaborative adaptive hibernation and wake-up mechanism based on channels, services, and energy, and execute differentiated hibernation rules according to topology roles: ordinary nodes within a cluster only wake up in allocated time slots, and the hibernation period is adjusted in conjunction with service priority, remaining energy, and channel prediction results; backup cluster heads and primary cluster heads use staggered wake-up, and the wake-up windows of adjacent cluster heads are staggered, while the primary cluster head is in normal shallow hibernation and listens to the control channel; in emergency situations, when the marine ecological parameters and network operation parameters monitored by the node approach the warning threshold, it switches to shallow hibernation, and after triggering an alarm, it switches to high-frequency wake-up, and returns to normal after the emergency is lifted; the cluster head broadcasts the hibernation and wake-up timing information of the entire network to ensure that the node wake-up windows are aligned.

7. The wireless networking method for marine ecological monitoring terminals according to claim 6, characterized in that, During the self-healing reconstruction phase, the shore-based edge gateway collects the status of all network nodes, link quality, environmental data, and service transmission data in real time, constructs a digital twin image of the marine network in the cloud, and maps the core status of the entire network in real time; based on the predicted data, it simulates network status changes, predicts various fault risks, and generates hierarchical self-healing reconstruction plans in advance. Set up a three-level early warning and fault classification mechanism based on link health, and perform pre-adjustment actions such as route optimization, backup route activation, and topology pre-reconstruction according to different early warning levels; In the event of an early warning or fault, the node autonomously executes a self-healing plan and cross-cluster distributed reconstruction. After reconstruction, it reports information and updates the cloud mirror and the entire network information.

8. The wireless networking method for marine ecological monitoring terminals according to claim 7, characterized in that, During the optimization management phase, the cloud platform identifies network performance bottlenecks based on the data from the entire network and constructs a distributed network optimization model using a federated learning framework; the shore-based edge gateway trains a local distributed optimization model based on the data of the sea area under its jurisdiction, and only uploads the model parameters to the cloud. The cloud aggregates the parameters to generate a global optimization model and distributes it to each shore-based edge node and the entire network monitoring terminal node. Based on the iterative optimization of the core network parameters and related prediction models using a global model, a dedicated network knowledge base for marine areas is formed; a node operation and maintenance early warning list is generated through a cloud-based digital twin system, supporting the automatic network access of new nodes and the synchronization of network-wide configuration.

Images