Adaptive floating node device for cross-medium network energy transmission and scheduling

Abstract

The invention discloses a self-adaptive floating node device for cross-medium network energy transmission and scheduling. The self-adaptive floating node device comprises a spherical floating body module, an information acquisition module, a cross-medium energy transmission module, an energy scheduling module, a control and decision-making module and a self-adaptive movement module, the spherical floating body maintains a semi-submersible posture through a hollow layered structure and a dynamic gravity center, and provides a stable carrier; the information acquisition module acquires node data through the up-down communication unit and preprocesses the node data; the cross-medium energy transmission module adopts a dual-mode composite energy transmission design; the energy scheduling module dynamically distributes stored energy in combination with a multi-dimensional priority evaluation model; the control module formulates a fault self-healing scheme based on a CART decision tree and a random forest model; the adaptive movement module realizes optimal movement through an A * algorithm and federated Kalman filtering; according to the device, cross-medium energy transmission and information interaction are respectively realized through the double-function unit, a cross-medium collaborative core of energy and information is constructed, and the device has wide application value.

Description

Technical Field

[0001] This invention relates to the field of energy systems and Internet of Things (IoT) deployment, and more specifically to an adaptive floating node device for cross-media network energy transmission and scheduling. Background Technology

[0002] With the deepening development of marine development and environmental monitoring, cross-medium networks have become a core support for data acquisition and energy replenishment. The efficiency of their energy transmission and coordinated scheduling directly determines the stability of network operation. Currently, wireless power transmission technology has been extensively researched in fields such as transportation and electronic equipment, with related patents mainly focusing on the design of magnetic coupling mechanisms, optimization of single-medium control strategies, and improvement of circuit topology. However, dedicated node devices and intelligent decision-making models for cross-medium scenarios still face significant technical bottlenecks, making it difficult to meet the precise scheduling requirements in complex environments.

[0003] Chinese patent (publication number CN119051285A) discloses a wireless charging method for lightweight underwater vehicles without communication based on BPNN prediction. This method uses a neural network to predict the output parameters of underwater charging. However, the model only relies on electrical data such as primary side current and voltage, and does not integrate environmental parameters such as solar intensity and water flow speed, as well as features such as mission urgency. This makes it unable to adapt to the needs of multi-factor dynamic interference in cross-medium scenarios.

[0004] Chinese patent (publication number CN114050796B) discloses a digital predistortion system and method based on random forest algorithm. The method uses random forest for power amplifier linearization control, and the number of decision trees is fixed at 60. It does not dynamically adjust according to the interference intensity of different media. If it is directly applied to cross-media scenarios, the prediction error will exceed 8% in the deep water area with strong interference due to insufficient number of trees, while the computational efficiency will be reduced by more than 30% in the land with weak interference due to redundant number of trees.

[0005] Chinese patent (publication number CN121052541A) discloses a multi-type energy dispatching method and system based on trusted data space. The system constructs a federated learning prediction model, but focuses on land energy scenarios such as power grids and does not involve hardware adaptation for underwater and above-water cross-medium energy transmission. Its optimization objective is energy absorption efficiency and energy curtailment rate, which does not match the multi-objective requirements of energy transmission efficiency, link quality and attitude stability in cross-medium scenarios, resulting in a disconnect between prediction, decision-making and execution of the cross-medium energy dispatching system.

[0006] In summary, to address the technical pain points of insufficient accuracy in predicting energy supply and demand and communication quality in cross-media networks, poor model adaptability, and low hardware-software synergy, it is crucial to develop an intelligent prediction and scheduling system that integrates multi-source data, dynamically optimizes parameters, and deeply collaborates with hardware. This is a key requirement for improving the energy transmission efficiency and operational stability of cross-media networks. Summary of the Invention

[0007] Based on the aforementioned technical problems, this application discloses an adaptive floating node device for cross-medium network energy transmission and scheduling, comprising a spherical floating body, an information acquisition module, a cross-medium energy transmission module, an energy scheduling module, a control and decision-making module, and an adaptive movement module, specifically:

[0008] The spherical floating body is constructed with a hollow layered structure and a dynamically adjustable center of gravity, maintaining a semi-submersible attitude with half above water and half below water. It provides a stable carrier for the information acquisition module and the cross-medium energy transmission system to be deployed on the surface and underwater, ensuring the basic environmental stability of cross-medium data interaction and energy transmission.

[0009] The information acquisition module consists of a top communication unit and a bottom communication unit, which respectively collect energy status data and environmental monitoring data from land nodes and underwater nodes, receive and transmit control commands, and transmit multi-dimensional data to the control and decision-making module after preprocessing.

[0010] The cross-medium energy transfer module is constructed from a top-mounted underwater energy transfer unit and a bottom-mounted underwater energy transfer unit. The module uses a dual-mode structure of wireless sensing and contact in the top-mounted underwater energy transfer unit, combined with the positioning guidance and composite energy transfer design of the bottom-mounted underwater energy transfer unit, to transfer energy to land nodes and underwater nodes respectively under the coordination of the control and decision module, while simultaneously feeding back the energy transfer link parameters to the control and decision module in real time.

[0011] The energy scheduling module stores adaptively acquired solar energy through the energy storage unit. Through the intelligent energy scheduling unit, combined with the data transmitted by the control and decision-making module and the information acquisition module, a multi-dimensional priority evaluation model is constructed. The energy storage unit is divided into a basic energy storage area and an energy transmission buffer zone, and energy is dynamically allocated to the device's own core module and the cross-medium energy transmission module.

[0012] Based on the data transmitted by the information acquisition module, the control and decision-making module constructs an energy supply and demand prediction model and a communication quality prediction model through the edge decision tree algorithm. Based on the model output, it formulates a cross-medium energy transmission strategy and a node collaborative scheduling scheme. At the same time, based on the cross-medium energy transmission status and node data, it optimizes the scheduling algorithm and generates a fault self-healing strategy.

[0013] The adaptive movement module acquires the position and environmental data of the device itself and the target node through the attitude perception unit, plans the optimal movement path through the movement decision algorithm unit, drives the composite propulsion unit to move the device to the best position for cross-medium energy transmission, and dynamically adjusts the movement strategy according to the link status fed back by the information acquisition module.

[0014] Preferably, the hollow layered structure and dynamically adjustable center of gravity are specifically: a hollow layered structure of a spherical floating body with a three-layer composite design, the outer layer being a high-strength corrosion-resistant composite material, the middle layer being a sealing and heat-insulating layer, and the inner layer being a conductive shielding layer; the dynamically adjustable center of gravity is achieved through a built-in center of gravity adjustment architecture, which uses a vertical slide rail arranged along the axis of the spherical floating body and a stepper motor to drive a movable counterweight, so that the counterweight moves on the slide rail to achieve dynamic adjustment of the center of gravity, and is linked with an attitude sensor, when the tilt angle exceeds a preset tilt angle threshold. The stepper motor starts immediately, driving the counterweight to move and adjust the center of gravity.

[0015] Preferably, the preprocessing specifically involves filtering and denoising the data using a Kalman filter algorithm, as shown in the formula:

[0016]

[0017]

[0018] in, for The system state vector at any given time. Here is the state transition matrix. To control the input matrix, for Time-based control input, For process noise, for Time-based observations For the observation matrix, To observe noise;

[0019] The denoised data is processed by Criteria for eliminating more than outliers (where The arithmetic mean of the denoised data samples. The standard deviation of the denoised data sample is used to obtain the calibration coefficient through standard equipment calibration. The data is then linearly calibrated to obtain a representative dataset.

[0020] Preferably, the bottom underwater energy transfer unit specifically achieves positioning guidance through two-way ranging between underwater acoustic communication and underwater nodes. The resonant coupling energy transfer adjustment formula for underwater acoustic communication is:

[0021]

[0022] in, The resonant frequency, For inductance, For capacitors;

[0023] The formula for two-way distance measurement at underwater nodes is:

[0024]

[0025] in The distance between the device and the underwater node. Let be the speed at which sound waves travel in water. This is the round-trip time difference of the sound wave signal.

[0026] Preferably, the multi-dimensional priority evaluation model specifically involves: converting multi-dimensional information into a single-dimensional score value through a preset weight allocation, and then converting the multi-dimensional information into a single-dimensional score value through linear summation, thereby achieving an intuitive comparison and ranking of node priorities and adapting to the real-time scheduling requirements of cross-media networks. The formula is as follows:

[0027]

[0028] in, Score based on priority. The maximum energy storage capacity of the node. This represents the node's current remaining battery power. This represents the task importance coefficient. This is the link quality coefficient. The remaining energy percentage of the node. , , These are the weighting coefficients for node energy, task importance, and link quality, respectively.

[0029] Preferably, the weighting coefficients for node energy, task importance, and link quality are specifically determined by dynamically adjusting the weighting coefficients of the three dimensions using a gradient descent algorithm, with the optimization objective of minimizing the difference in demand among nodes after energy allocation. , , The loss function is defined to quantify the deviation between the model's predicted values ​​and the actual allocation results. The formula is as follows:

[0030]

[0031] in, For loss function, For nodes Priority score, This represents the actual energy distribution ratio. The number of nodes participating in the energy allocation;

[0032] By calculating the loss function , , The partial derivatives are used to determine the direction of weight adjustment, and the weight coefficients are updated step by step until the loss function converges, thereby achieving dynamic optimization of the weights.

[0033] Preferably, the construction of the energy supply and demand prediction model and the communication quality prediction model specifically involves: collecting input feature variables and output target variables of historical cross-medium energy transmission. The input feature variables include the remaining energy of the node, task type, ambient temperature, solar radiation intensity and historical energy transmission frequency. The output target variable is the node energy demand value. After preprocessing, a supply and demand prediction training set and a communication prediction training set are constructed.

[0034] The energy supply and demand forecasting model is constructed using a CART decision tree, with the Gini coefficient as the splitting criterion. The formula is as follows:

[0035]

[0036] in, The Gini coefficient for the sample. For the first The proportion of class samples in the nodes;

[0037] The communication quality prediction model is constructed using the random forest algorithm, where the number of feature subsets for each tree in the forest is [number missing]. ,in The total number of characteristics;

[0038] The fault self-healing strategy is generated through fault tree analysis (FTA). The top event is the interruption of cross-medium energy transmission, and the bottom events include energy transfer module failure, communication link interruption, and insufficient energy storage. The critical fault path is determined by the minimum cut set method.

[0039] Preferably, the energy supply and demand forecasting model is constructed using a CART decision tree, specifically as follows:

[0040] Based on the splitting criterion of minimizing the Gini coefficient, the training set is recursively split. During each round of splitting, the Gini coefficient gain of each feature of the current node is calculated, and the feature with the largest gain and its corresponding split point are selected for node partitioning. The formula is as follows:

[0041]

[0042] in For Gini coefficient gain, Let Gini coefficient be the parent node. is the Gini coefficient of the child node. The number of child node samples. The number of samples in the parent node;

[0043] Once the model converges, the splitting is terminated by pre-pruning that limits the tree depth and the minimum number of leaf node samples. Redundant subtrees are then removed by post-pruning based on the error minimization principle, using a test set to verify the model, thus obtaining the target model.

[0044] Preferably, the communication quality prediction model is constructed using the random forest algorithm, specifically by setting the number of decision trees in the forest and the feature subset trees of each tree, and by using the Bootstrap sampling method to randomly extract samples of the same size as the original training set for each tree, while simultaneously performing random subset extraction of the input features to ensure diversity among trees.

[0045] Based on the feature stratification of the current tree sampling, the tree structure parameters are dynamically adjusted. If the tree contains many environmental interference layer features, the tree depth is increased and the number of minimum leaf node samples is reduced to capture subtle changes in the interference layer features.

[0046] Each tree is independently constructed and trained using CART decision trees based on sampled data and features, according to the Gini coefficient splitting criterion. After all trees are trained, the prediction results are integrated using a simple majority voting method to obtain the communication link quality coefficient. The predicted value;

[0047] Verify the predicted values ​​and actual communication link quality coefficients. If the deviation between the predicted communication link quality coefficient Q and the corresponding actual value exceeds a preset error threshold, the parameters of the random forest model are readjusted and retrained. After the model converges, the target communication quality prediction model is obtained. The error threshold is a preset upper limit for measuring the difference between the predicted communication link quality coefficient and the corresponding actual value.

[0048] Preferably, the attitude and position sensing unit of the adaptive motion module integrates multi-sensor data to obtain position and environmental data using a federated Kalman filter algorithm. The local filtering formula is as follows:

[0049]

[0050]

[0051]

[0052]

[0053] in , , , The first One sensor Prior state estimate at time step, posterior local estimate after fusion of observations, local estimate covariance, Kalman gain, sensor observations. , , These represent the actual energy allocation ratio, control input matrix, observation matrix, and link quality coefficient for the i-th sample, respectively.

[0054] The federal fusion formula is:

[0055]

[0056]

[0057] in This is the global fusion estimate. To globally integrate covariance, For fusion weights;

[0058] After obtaining location and environmental data, the mobile decision algorithm unit uses the A* path planning algorithm to calculate the mobile decision of the adaptive mobile module. The heuristic function formula is as follows:

[0059]

[0060] in For nodes The heuristic cost to the target node, Distance weights For nodes The Euclidean distance to the target node. As energy consumption weight, For nodes Energy consumption per unit distance traveled;

[0061] Based on movement decisions, the composite propulsion unit is driven to move adaptively.

[0062] Compared with the prior art, the technical solution of this application has the following technical effects:

[0063] This invention reduces energy supply and demand prediction errors and improves generalization ability by constructing multi-source dynamic datasets, hierarchical feature selection and adaptive standardization, two-stage pruning and error feedback iteration, breaking through the limitations of traditional prediction models and achieving accurate prediction across media scenarios.

[0064] This invention utilizes a dual-core architecture for control and decision-making modules. The main controller coordinates the timing of the module, while the edge decision-making module integrates the outputs of two types of prediction models to construct a multi-dimensional priority evaluation model. The resulting scheduling strategy directly drives the cross-medium energy transmission module to switch energy transmission modes, the adaptive movement module to adjust its position, and the spherical floating body to fine-tune its center of gravity, thereby achieving real-time linkage between prediction results and hardware actions.

[0065] This invention significantly improves underwater energy transfer efficiency, reduces the interruption rate of cross-medium energy transfer, and substantially enhances network operational stability by using a cross-medium energy transfer module for positioning guidance and a composite energy transfer design, combined with a predictive model to optimize energy transfer location and power in advance.

[0066] This invention coordinates the energy status, environmental data, and task requirements of each medium node through an information module, and combines edge decision-making capabilities to complete energy supply and demand matching and communication link optimization, thereby achieving directional transmission of surplus energy and efficient data flow, improving the overall system utilization efficiency, and realizing precise coordination of energy and information.

[0067] This invention responds quickly to the urgent energy needs of nodes through an intelligent scheduling mechanism, effectively copes with sudden power shortages based on energy priority allocation and dynamic power adjustment, improves the operational resilience of the entire cross-media network, and strengthens the system's emergency response capability.

[0068] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0069] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0070] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0071] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0072] Figure 1 This is a general structural diagram of an adaptive floating node device for energy transfer and scheduling in cross-media networks;

[0073] Figure 2 A neural network architecture diagram for a multi-dimensional priority evaluation model;

[0074] Figure 3A diagram of the neural network architecture for an energy supply and demand forecasting model;

[0075] Figure 4 A diagram of the neural network architecture for a communication quality prediction model;

[0076] Figure 5 This is an experimental architecture diagram for testing the present invention at a marine environmental monitoring station;

[0077] Figure 6 This is a comparison chart of the accuracy and transmission performance indicators of different testing methods in the experiment;

[0078] Figure 7 This is a comparison chart of the error rate and imbalance rate of different testing methods in the experiment;

[0079] Figure 8 This is a comparison chart of the predicted link stability duration data for different testing methods in the experimental group. Detailed Implementation

[0080] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0081] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0082] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0083] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0084] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0085] It should also 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.

[0086] Example 1 describes an adaptive floating node device for energy transfer and scheduling in cross-media networks, such as... Figure 1 As shown, it includes a spherical floating body, an information acquisition module, a cross-medium energy transmission module, an energy scheduling module, a control and decision-making module, and an adaptive movement module, specifically:

[0087] The spherical floating body is constructed with a hollow layered structure and a dynamically adjustable center of gravity, maintaining a semi-submersible attitude with half above water and half below water. It provides a stable carrier for the information acquisition module and the cross-medium energy transmission system to be deployed on the surface and underwater, ensuring the basic environmental stability of cross-medium data interaction and energy transmission.

[0088] The information acquisition module consists of a top communication unit and a bottom communication unit, which respectively collect energy status data and environmental monitoring data from land nodes and underwater nodes, receive and transmit control commands, and transmit multi-dimensional data to the control and decision-making module after preprocessing.

[0089] The cross-medium energy transfer module is constructed from a top-mounted underwater energy transfer unit and a bottom-mounted underwater energy transfer unit. The module uses a dual-mode structure of wireless sensing and contact in the top-mounted underwater energy transfer unit, combined with the positioning guidance and composite energy transfer design of the bottom-mounted underwater energy transfer unit, to transfer energy to land nodes and underwater nodes respectively under the coordination of the control and decision module, while simultaneously feeding back the energy transfer link parameters to the control and decision module in real time.

[0090] The energy scheduling module stores adaptively acquired solar energy through the energy storage unit. Through the intelligent energy scheduling unit, combined with the data transmitted by the control and decision-making module and the information acquisition module, a multi-dimensional priority evaluation model is constructed. The energy storage unit is divided into a basic energy storage area and an energy transmission buffer zone, and energy is dynamically allocated to the device's own core module and the cross-medium energy transmission module.

[0091] Based on the data transmitted by the information acquisition module, the control and decision-making module constructs an energy supply and demand prediction model and a communication quality prediction model through the edge decision tree algorithm. Based on the model output, it formulates a cross-medium energy transmission strategy and a node collaborative scheduling scheme. At the same time, based on the cross-medium energy transmission status and node data, it optimizes the scheduling algorithm and generates a fault self-healing strategy.

[0092] The adaptive movement module acquires the position and environmental data of the device itself and the target node through the attitude perception unit, plans the optimal movement path through the movement decision algorithm unit, drives the composite propulsion unit to move the device to the best position for cross-medium energy transmission, and dynamically adjusts the movement strategy according to the link status fed back by the information acquisition module.

[0093] Furthermore, the spherical floating body adopts a hollow structure, and its size can be adjusted according to the deployment scenario. Its material is a layered design, with the outer layer being a high-strength corrosion-resistant composite material that is resistant to ultraviolet rays, seawater erosion, and impact. The middle layer is a sealing and heat-insulating layer, and the inner layer is a conductive shielding layer. For the underwater functional areas, a composite sealing solution combining welding, multi-level sealant, and waterproof joints is adopted. The spherical floating body has a slide rail set along the vertical direction inside, and a movable counterweight is installed on the slide rail. Together with the drive unit, it forms a center of gravity adjustment mechanism. At the same time, the lower half of the sphere is filled with low-density buoyancy material to control the buoyancy center and achieve a stable semi-submersible attitude for the device, that is, half of it is above water and half of it is underwater.

[0094] The sphere material adopts a layered design: the outer layer is a high-strength corrosion-resistant composite material, which is resistant to ultraviolet rays, seawater erosion and impact; the middle layer is a sealed heat insulation layer, which blocks the exchange of internal and external environments and ensures the stable operation of the internal energy storage, energy transmission and communication modules; the inner layer is a conductive shielding layer, which reduces the impact of external interference on the energy control circuit and communication signals.

[0095] The sealing design focuses on underwater functional areas, employing a composite solution of "welding + multi-level sealant + waterproof joints" to ensure the underwater unit's sealing performance under different water pressure environments, preventing seawater infiltration from affecting power transmission efficiency, communication quality, and equipment safety.

[0096] Furthermore, the preprocessing specifically involves filtering and denoising the data using a Kalman filter algorithm, with the following formula:

[0097]

[0098]

[0099] in, for The system state vector at any given time. Here is the state transition matrix. To control the input matrix, for Time-based control input, For process noise, for Time-based observations For the observation matrix, To observe noise;

[0100] The denoised data is processed by Criteria for eliminating more than outliers (where The arithmetic mean of the denoised data samples. The standard deviation of the denoised data sample is used to obtain the calibration coefficient through standard equipment calibration. The data is then linearly calibrated to obtain a representative dataset.

[0101] Furthermore, the bottom underwater energy transfer unit specifically achieves positioning guidance through two-way ranging between underwater acoustic communication and underwater nodes. The resonant coupling energy transfer adjustment formula for underwater acoustic communication is:

[0102]

[0103] in, The resonant frequency, For inductance, For capacitors;

[0104] The formula for two-way distance measurement at underwater nodes is:

[0105]

[0106] in The distance between the device and the underwater node. Let be the speed at which sound waves travel in water. This is the round-trip time difference of the sound wave signal.

[0107] Furthermore, such as Figure 2 As shown, the multi-dimensional priority evaluation model specifically involves: converting multi-dimensional information into a single-dimensional score value through a preset weight allocation, and then converting the multi-dimensional information into a single-dimensional score value through linear summation, thereby achieving an intuitive comparison and ranking of node priorities, adapting to the real-time scheduling requirements of cross-media networks. The formula is as follows:

[0108]

[0109] in, Score based on priority. The maximum energy storage capacity of the node. This represents the node's current remaining battery power. This represents the task importance coefficient. This is the link quality coefficient. The remaining energy percentage of the node. , , These are the weighting coefficients for node energy, task importance, and link quality, respectively.

[0110] Furthermore, the weighting coefficients for node energy, task importance, and link quality are specifically adjusted using a gradient descent algorithm, with the optimization objective of minimizing the difference in demand among nodes after energy allocation. , , The loss function is defined to quantify the deviation between the model's predicted values ​​and the actual allocation results. The formula is as follows:

[0111]

[0112] in, For loss function, For nodes Priority score, This represents the actual energy distribution ratio. The number of nodes participating in the energy allocation;

[0113] By calculating the loss function , , The partial derivatives are used to determine the direction of weight adjustment, and the weight coefficients are updated step by step until the loss function converges, thereby achieving dynamic optimization of the weights.

[0114] Furthermore, such as Figure 3 , Figure 4 As shown, the construction of the energy supply and demand prediction model and the communication quality prediction model specifically involves: collecting input feature variables and output target variables of historical cross-medium energy transmission. The input feature variables include the remaining energy of the node, task type, ambient temperature, solar radiation intensity and historical energy transmission frequency. The output target variable is the node energy demand value. After preprocessing, the supply and demand prediction training set and the communication prediction training set are constructed.

[0115] The energy supply and demand forecasting model is constructed using a CART decision tree, with the Gini coefficient as the splitting criterion. The formula is as follows:

[0116]

[0117] in, The Gini coefficient for the sample. For the first The proportion of class samples in the nodes;

[0118] The communication quality prediction model is constructed using the random forest algorithm, where the number of feature subsets for each tree in the forest is [number missing]. ,in The total number of characteristics;

[0119] The fault self-healing strategy is generated through fault tree analysis (FTA). The top event is the interruption of cross-medium energy transmission, and the bottom events include energy transfer module failure, communication link interruption, and insufficient energy storage. The critical fault path is determined by the minimum cut set method.

[0120] Furthermore, the energy supply and demand forecasting model is constructed using a CART decision tree as follows:

[0121] Based on the splitting criterion of minimizing the Gini coefficient, the training set is recursively split. During each round of splitting, the Gini coefficient gain of each feature of the current node is calculated, and the feature with the largest gain and its corresponding split point are selected for node partitioning. The formula is as follows:

[0122]

[0123] in For Gini coefficient gain, Let Gini coefficient be the parent node. is the Gini coefficient of the child node. The number of child node samples. The number of samples in the parent node;

[0124] Once the model converges, the splitting is terminated by pre-pruning that limits the tree depth and the minimum number of leaf node samples. Redundant subtrees are then removed by post-pruning based on the error minimization principle, using a test set to verify the model, thus obtaining the target model.

[0125] Furthermore, the communication quality prediction model is constructed using the random forest algorithm, specifically: setting the number of decision trees in the forest and the feature subset trees of each tree, and randomly sampling samples of the same size as the original training set for each tree using the Bootstrap sampling method, while simultaneously performing random subset sampling of the input features to ensure diversity among trees;

[0126] Based on the feature stratification of the current tree sampling, the tree structure parameters are dynamically adjusted. If the tree contains many environmental interference layer features, the tree depth is increased and the number of minimum leaf node samples is reduced to capture subtle changes in the interference layer features.

[0127] Each tree is independently constructed and trained using CART decision trees based on sampled data and features, according to the Gini coefficient splitting criterion. After all trees are trained, the prediction results are integrated using a simple majority voting method to obtain the communication link quality coefficient. The predicted value;

[0128] Verify the predicted values ​​and actual communication link quality coefficients. If the deviation between the predicted communication link quality coefficient Q and the corresponding actual value exceeds a preset error threshold, the parameters of the random forest model are readjusted and retrained. After the model converges, the target communication quality prediction model is obtained. The error threshold is a preset upper limit for measuring the difference between the predicted communication link quality coefficient and the corresponding actual value.

[0129] Furthermore, through the attitude and position sensing unit of the adaptive motion module, a federated Kalman filter algorithm is used to integrate multi-sensor data to obtain position and environmental data. The local filtering formula is:

[0130]

[0131]

[0132]

[0133]

[0134] in , , , The first One sensor Prior state estimate at time step, posterior local estimate after fusion of observations, local estimate covariance, Kalman gain, sensor observations. , , These represent the actual energy allocation ratio, control input matrix, observation matrix, and link quality coefficient for the i-th sample, respectively.

[0135] The federal fusion formula is:

[0136]

[0137]

[0138] in This is the global fusion estimate. To globally integrate covariance, For fusion weights;

[0139] After obtaining location and environmental data, the mobile decision algorithm unit uses the A* path planning algorithm to calculate the mobile decision of the adaptive mobile module. The heuristic function formula is as follows:

[0140]

[0141] in For nodes The heuristic cost to the target node, Distance weights For nodes The Euclidean distance to the target node. As energy consumption weight, For nodes Energy consumption per unit distance traveled;

[0142] Based on movement decisions, the composite propulsion unit is driven to move adaptively.

[0143] This embodiment details an adaptive floating node device for cross-medium network energy transmission and scheduling, comprising six modules: a spherical floating body, information acquisition, cross-medium energy transmission, energy scheduling, control and decision-making, and adaptive movement. The spherical floating body maintains a semi-submerged attitude with a hollow layered structure and a dynamically adjustable center of gravity, providing a stable carrier for cross-medium functions. The information acquisition module collects and preprocesses node data through upper and lower communication units. The cross-medium energy transmission module adopts a dual-mode composite energy transmission design, accurately transmitting energy with the cooperation of the control module. The energy scheduling module dynamically allocates energy storage using a multi-dimensional priority evaluation model. The control module formulates strategies and generates fault self-healing schemes based on CART decision trees and random forest models. The adaptive movement module achieves optimal movement through the A* algorithm and federated Kalman filtering.

[0144] Example 2 describes in detail the experiment conducted on the present invention at a marine environmental monitoring station. The test area included shore-based monitoring points, nearshore aquaculture areas (diving), and seabed ecological areas (deep water). The specific experimental process is as follows:

[0145] The adaptive floating node device of this invention is deployed based on a spherical floating body with a diameter of 1.2m, an outer layer thickness of 15mm, a middle heat insulation layer of 8mm, and an inner shielding layer of 5mm. The information acquisition module uses a RAK3172 top LoRa communication unit and a Teledyne Benthos ATM900 bottom underwater acoustic communication unit; the cross-medium energy transmission module uses a top wireless inductive energy transfer coil with a diameter of 30cm and a coupling coefficient of 0.8 and a bottom underwater acoustic resonant coupler with a resonant frequency of 20kHz; the adaptive movement module uses a T-Motor U8 brushless DC propeller; and the control and decision module uses an STM32H743VI main controller and an RK3399Pro edge decision AI chip.

[0146] Five land-based monitoring nodes, numbered L1-L5, were set up at the shore-based monitoring point; three shallow-water monitoring nodes, numbered S1-S3, were set up in the nearshore aquaculture area; and five deep-water monitoring nodes, numbered D1-D5, were set up in the seabed ecological area.

[0147] Temperature and humidity data were collected every 10 minutes at the land node, water quality data were collected every 15 minutes at the shallow water node, and seabed ecological data were collected every 5 minutes at the deep water node. Comparative tests were conducted using existing commonly used methods, such as MCR-WPT, GCN-GRU, and PTMR-Comm, under the same environment.

[0148] Table 1 illustrates the test results of the method of the present invention compared to existing methods for various indicators.

[0149]

[0150] As shown in Table 1 above, the present invention achieves low error rates of 3.2% and 2.1% in energy supply and demand prediction and cross-medium transmission interruption rate, respectively. Furthermore, the accuracy rates of communication quality prediction and low-quality link prediction reach 92.5% and 87.1%, respectively, representing significant breakthroughs compared to the three existing methods. In addition, the present invention also surpasses the three existing methods in terms of predicted link stability duration deviation and total power replenishment time, demonstrating its significant advantages in accuracy and efficiency.

[0151] according to Figure 6 - Figure 8 It can be seen that, through multi-source data fusion and dynamic optimization, the present invention reduces the energy prediction error by 4.1% compared with GCN-GRU, increases the low-quality link communication prediction accuracy by 22.3% compared with PTMR-Comm, reduces the transmission interruption rate by 14.7% compared with PTMR-Comm, and lowers the maximum imbalance rate of deep-sea devices to 1.8%. It has significant advantages over existing methods in all aspects of data.

[0152] This embodiment describes in detail the experiment conducted on the present invention at a marine environmental monitoring station. The experiment fully verifies the significant advantages of the present invention in terms of energy prediction error, low-quality link communication prediction accuracy, energy replenishment time, transmission interruption rate, emergency data integrity, and hardware attitude stability through multi-source data fusion and dynamic optimization. It also demonstrates the excellent prediction performance, efficient energy transmission scheduling, and wide scenario adaptability of the present invention.

[0153] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. An adaptive floating node device for energy transfer and scheduling across media networks, characterized in that, It includes a spherical floating body, an information acquisition module, a cross-medium energy transfer module, an energy scheduling module, a control and decision-making module, and an adaptive movement module, specifically: The spherical floating body is constructed with a hollow layered structure and a dynamically adjustable center of gravity, maintaining a semi-submersible attitude with half above water and half below water. It provides a stable carrier for the information acquisition module and the cross-medium energy transmission system to be deployed on the surface and underwater, ensuring the basic environmental stability of cross-medium data interaction and energy transmission. The information acquisition module consists of a top communication unit and a bottom communication unit, which respectively collect energy status data and environmental monitoring data from land nodes and underwater nodes, receive and transmit control commands, and transmit multi-dimensional data to the control and decision-making module after preprocessing. The cross-medium energy transfer module is constructed from a top-mounted underwater energy transfer unit and a bottom-mounted underwater energy transfer unit. The module uses a dual-mode structure of wireless sensing and contact in the top-mounted underwater energy transfer unit, combined with the positioning guidance and composite energy transfer design of the bottom-mounted underwater energy transfer unit, to transfer energy to land nodes and underwater nodes respectively under the coordination of the control and decision module, while simultaneously feeding back the energy transfer link parameters to the control and decision module in real time. The energy scheduling module stores adaptively acquired solar energy through the energy storage unit. Through the intelligent energy scheduling unit, combined with the data transmitted by the control and decision-making module and the information acquisition module, a multi-dimensional priority evaluation model is constructed. The energy storage unit is divided into a basic energy storage area and an energy transmission buffer zone, and energy is dynamically allocated to the device's own core module and the cross-medium energy transmission module. Based on the data transmitted by the information acquisition module, the control and decision-making module constructs an energy supply and demand prediction model and a communication quality prediction model through the edge decision tree algorithm. Based on the model output, it formulates a cross-medium energy transmission strategy and a node collaborative scheduling scheme. At the same time, based on the cross-medium energy transmission status and node data, it optimizes the scheduling algorithm and generates a fault self-healing strategy. The adaptive movement module acquires the position and environmental data of the device itself and the target node through the attitude perception unit, plans the optimal movement path through the movement decision algorithm unit, drives the composite propulsion unit to move the device to the best position for cross-medium energy transmission, and dynamically adjusts the movement strategy according to the link status fed back by the information acquisition module.

2. The apparatus according to claim 1, characterized in that, The hollow layered structure and dynamically adjustable center of gravity are specifically described as follows: The spherical floating body features a hollow layered structure with a three-layer composite design. The outer layer is a high-strength, corrosion-resistant composite material, the middle layer is a sealed and heat-insulating layer, and the inner layer is a conductive shielding layer. The dynamically adjustable center of gravity is achieved through a built-in center of gravity adjustment architecture. This architecture uses a vertical slide rail arranged along the axis of the spherical floating body and a stepper motor to drive a movable counterweight, allowing the counterweight to move on the slide rail to achieve dynamic adjustment of the center of gravity. This is linked to an attitude sensor; when the tilt angle exceeds a preset tilt angle threshold... The stepper motor starts immediately, driving the counterweight to move and adjust the center of gravity.

3. The apparatus according to claim 1, characterized in that, The preprocessing specifically involves filtering and denoising the data using the Kalman filter algorithm, with the following formula: ； ； in, for The system state vector at any given time. Here is the state transition matrix. To control the input matrix, for Time-based control input, For process noise, for Time-based observations For the observation matrix, To observe noise; The denoised data is processed by Criteria for eliminating more than outliers, among which The arithmetic mean of the denoised data samples. The standard deviation of the denoised data sample is used to obtain the calibration coefficient through standard equipment calibration. The data is then linearly calibrated to obtain a representative dataset.

4. The apparatus according to claim 1, characterized in that, The bottom underwater energy transfer unit specifically achieves positioning guidance through two-way ranging between underwater acoustic communication and underwater nodes. The resonant coupling energy transfer adjustment formula for underwater acoustic communication is as follows: ； in, The resonant frequency, For inductance, For capacitors; The formula for two-way distance measurement at underwater nodes is: ； in The distance between the device and the underwater node. Let be the speed at which sound waves travel in water. This is the round-trip time difference of the sound wave signal.

5. The apparatus according to claim 1, characterized in that, The multi-dimensional priority evaluation model specifically involves: converting multi-dimensional information into a single-dimensional score value through a preset weight allocation, and then converting the multi-dimensional information into a single-dimensional score value through linear summation, thereby achieving an intuitive comparison and ranking of node priorities and adapting to the real-time scheduling requirements of cross-media networks. The formula is as follows: ； in, Score based on priority. The maximum energy storage capacity of the node. This represents the node's current remaining battery power. This represents the task importance coefficient. For link quality coefficient, The remaining energy percentage of the node. , , These are the weighting coefficients for node energy, task importance, and link quality, respectively.

6. The apparatus according to claim 5, characterized in that, The weighting coefficients for node energy, task importance, and link quality are specifically determined by dynamically adjusting the weighting coefficients of the three dimensions using a gradient descent algorithm, with the optimization objective of minimizing the difference in demand among nodes after energy allocation. , , The loss function is defined to quantify the deviation between the model's predicted values ​​and the actual allocation results. The formula is as follows: ； in, For loss function, For nodes Priority score, This represents the actual energy distribution ratio. The number of nodes participating in the energy allocation; By calculating the loss function , , The partial derivatives are used to determine the direction of weight adjustment, and the weight coefficients are updated step by step until the loss function converges, thereby achieving dynamic optimization of the weights.

7. The apparatus according to claim 1, characterized in that, The construction of the energy supply and demand prediction model and the communication quality prediction model specifically involves: collecting input feature variables and output target variables from historical cross-medium energy transmission. The input feature variables include the remaining energy of the node, task type, ambient temperature, solar radiation intensity, and historical energy transmission frequency. The output target variable is the node energy demand value. After preprocessing, a supply and demand prediction training set and a communication prediction training set are constructed. The energy supply and demand forecasting model is constructed using a CART decision tree, with the Gini coefficient as the splitting criterion. The formula is as follows: ； in, The Gini coefficient for the sample. For the first The proportion of class samples in the nodes; The communication quality prediction model is constructed using the random forest algorithm, where the number of feature subsets for each tree in the forest is [number missing]. ,in The total number of characteristics; The fault self-healing strategy is generated through fault tree analysis (FTA). The top event is the interruption of cross-medium energy transmission, and the bottom events include energy transfer module failure, communication link interruption, and insufficient energy storage. The critical fault path is determined by the minimum cut set method.

8. The apparatus according to claim 7, characterized in that, The energy supply and demand forecasting model is constructed using a CART decision tree, specifically as follows: Based on the splitting criterion of minimizing the Gini coefficient, the training set is recursively split. During each round of splitting, the Gini coefficient gain of each feature of the current node is calculated, and the feature with the largest gain and its corresponding split point are selected for node partitioning. The formula is as follows: ； in For Gini coefficient gain, Let Gini coefficient be the parent node. is the Gini coefficient of the child node. The number of child node samples. The number of samples in the parent node; Once the model converges, the splitting is terminated by pre-pruning that limits the tree depth and the minimum number of leaf node samples. Redundant subtrees are then removed by post-pruning based on the error minimization principle, using a test set to verify the model, thus obtaining the target model.

9. The apparatus according to claim 7, characterized in that, The communication quality prediction model is constructed using the random forest algorithm. Specifically, the number of decision trees in the forest and the feature subset trees of each tree are set. The Bootstrap sampling method is used to randomly extract samples of the same size as the original training set for each tree. At the same time, random subsets of the input features are extracted to ensure diversity among trees. Based on the feature layering of the current tree sampling, the tree structure parameters are dynamically adjusted. If the proportion of environmental interference-related features used for splitting in the current decision tree exceeds a preset threshold, the tree depth is increased and the number of minimum leaf node samples is reduced to capture subtle changes in the interference layer features. Each tree is independently constructed and trained using CART decision trees based on sampled data and features, according to the Gini coefficient splitting criterion. After all trees are trained, the prediction results are integrated using a simple majority voting method to obtain the communication link quality coefficient. The predicted value; Verify the predicted values ​​and actual communication link quality coefficients. When the deviation between the predicted communication link quality coefficient Q and the corresponding actual value exceeds a preset error threshold, the parameters of the random forest model are readjusted and retrained. After the model converges, the target communication quality prediction model is obtained. The error threshold is a preset upper limit of error used to measure the difference between the predicted communication link quality coefficient and the corresponding actual value.

10. The apparatus according to claim 1, characterized in that, The adaptive motion module's attitude and position sensing unit integrates multi-sensor data to obtain position and environmental data using a federated Kalman filter algorithm. The local filtering formula is as follows: ； ； ； ； in , , , The first One sensor Prior state estimate at time step, posterior local estimate after fusion of observations, local estimate covariance, Kalman gain, sensor observations. , , These represent the actual energy allocation ratio, control input matrix, observation matrix, and link quality coefficient for the i-th sample, respectively. The federal fusion formula is: ； ； in This is the global fusion estimate. To globally integrate covariance, For fusion weights; After obtaining location and environmental data, the mobile decision algorithm unit uses the A* path planning algorithm to calculate the mobile decision of the adaptive mobile module. The heuristic function formula is as follows: ； in For nodes The heuristic cost to the target node, Distance weights For nodes The Euclidean distance to the target node. As energy consumption weight, For nodes Energy consumption per unit distance traveled; Based on movement decisions, the composite propulsion unit is driven to move adaptively.

Images