Ad hoc power transmission line monitoring device data communication module and monitoring data backhaul system

Abstract

This invention discloses a data communication module and monitoring data feedback system for ad hoc network transmission line monitoring equipment, relating to the field of transmission line monitoring data communication technology. It includes a physical layer sensing unit, a signal feature extraction unit, an inference and calculation unit, a state evolution unit, and a protocol adaptation unit. The physical layer sensing unit collects micro-meteorological parameters, conductor vibration frequency, and tower tilt angle; the signal feature extraction unit extracts multi-dimensional feature vectors characterizing channel quality; the inference and calculation unit infers the link connection strategy of adjacent nodes using a built-in deep reinforcement learning link decision model; the state evolution unit maintains and updates the ad hoc network connectivity state matrix; and the protocol adaptation unit adjusts the transmission format and routing hop count based on the state matrix. This invention can adapt to the complex environment of transmission lines, reduce link anomalies, lower transmission latency, and improve the stability and efficiency of ad hoc network data communication.

Description

Technical Field

[0001] This invention belongs to the field of data communication technology for transmission line monitoring, specifically a data communication module and monitoring data feedback system for self-organizing network transmission line monitoring equipment. Background Technology

[0002] Data communication of transmission line monitoring equipment is crucial for ensuring the stable operation of the monitoring system. Currently, most mainstream monitoring equipment uses a self-organizing network mode for data communication, which achieves data transmission through monitoring nodes deployed along the transmission line. In existing technologies, monitoring nodes typically only collect basic monitoring parameters of the transmission line and use fixed link connection methods and transmission protocols for data transmission. Link decisions mostly rely on preset fixed rules and are not dynamically adjusted based on the environmental and equipment status parameters during the actual operation of the transmission line.

[0003] In existing technical solutions, link decision-making does not fully consider the impact of transmission line-specific micro-meteorological changes, conductor vibration, tower tilt, and other parameters on channel quality. This leads to a mismatch between link connection strategies and actual channel conditions, easily resulting in link interruptions and data transmission delays. Furthermore, existing ad hoc network communication modules do not dynamically track and update network connectivity status, making it impossible to promptly grasp the network connectivity situation at the next moment. This results in a lack of precise basis for adjusting data frame transmission formats and routing hop counts, making it difficult to adapt to the complex and ever-changing operating environment of transmission lines.

[0004] The operating environment of power transmission lines is complex. Factors such as micro-meteorological parameters, conductor vibration frequency, and tower tilt angle continuously affect the channel quality of ad hoc networks, thereby affecting the stability and reliability of data transmission. How to dynamically optimize link connection strategies based on actual monitoring parameters of power transmission lines, and at the same time monitor network connectivity status in real time to adjust transmission parameters, has become an urgent problem to be solved by existing technologies. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a data communication module for self-organizing network transmission line monitoring equipment, comprising: Physical layer sensing units are deployed at monitoring nodes along the transmission line. These units are used to collect micro-meteorological parameters, conductor vibration frequency, and tower tilt angle of the transmission line. A signal feature extraction unit connected to the physical layer sensing unit extracts a multi-dimensional feature vector from the micro-meteorological parameters, conductor vibration frequency, and tower tilt angle to characterize channel quality. The inference computing unit is connected to the signal feature extraction unit. The inference computing unit has a built-in link decision model based on deep reinforcement learning. The inference computing unit infers the link connection strategy between each adjacent node in the ad hoc network topology in real time based on the multi-dimensional feature vector. A state evolution unit connected to the inference computing unit maintains a state matrix reflecting the current connectivity state of the ad hoc network. The state evolution unit updates the state matrix according to the link connection strategy so that the state matrix can reflect the network connectivity status at the next moment. A protocol adaptation unit connected to the state evolution unit reads the updated state matrix and dynamically adjusts the transmission format and routing hop count of the data frame based on the connectivity information in the state matrix.

[0006] Furthermore, the signal feature extraction unit includes: A wavelet denoising subunit connected to the physical layer sensing unit performs noise reduction processing on the conductor vibration frequency signal to eliminate interference caused by environmental noise. The time-frequency analysis subunit connected to the wavelet denoising subunit performs joint time-frequency analysis on the denoised conductor vibration frequency signal and the micro-meteorological parameters to extract the variation characteristics of the channel impulse response. The vector construction subunit connected to the time-frequency analysis subunit normalizes and splices together the channel impulse response variation characteristics, the tower tilt angle, and the current signal-to-noise ratio measurement value to generate the multidimensional feature vector.

[0007] Furthermore, the inference computing unit includes: A model loading subunit connected to the signal feature extraction unit loads the pre-trained deep reinforcement learning link decision model and uses the multi-dimensional feature vector as the input state. The action space generation subunit, which is connected to the model loading subunit, generates a set of candidate actions containing multiple routing schemes based on the current list of neighbor nodes of the ad hoc network. A benefit evaluation subunit is connected to the action space generation subunit. The benefit evaluation subunit simulates and deduces each routing scheme in the candidate action set and calculates the expected benefit value corresponding to each scheme based on the deduction results. The policy output subunit, which is connected to the benefit evaluation subunit, selects the routing scheme with the largest expected benefit value and outputs it as the link connection policy.

[0008] Furthermore, the state evolution unit includes: A matrix initialization subunit connected to the inference calculation unit generates an initial state matrix based on the preset node deployment positions when the monitoring device is started. The element values ​​in the state matrix represent the theoretical connection state between nodes. A state update subunit connected to the matrix initialization subunit receives the link connection strategy and maps the connection and disconnection instructions contained in the link connection strategy into modification operations on the corresponding element values ​​in the state matrix. The evolution prediction subunit, connected to the state update subunit, uses a Markov chain model to predict the network topology change trend in the next few time steps based on the modified state matrix, and feeds the prediction results back to the state matrix for secondary correction.

[0009] Furthermore, the protocol adaptation unit includes: A routing table reconstruction subunit is connected to the state evolution unit. The routing table reconstruction subunit parses the state matrix, identifies the valid data transmission paths in the current network, and generates a real-time dynamic routing table accordingly. A frame structure adjustment subunit connected to the routing table reconstruction subunit automatically increases or decreases the length of the routing information field in the data frame header according to the transmission path length indicated in the dynamic routing table. The transmission control subunit, which is connected to the frame structure adjustment subunit, assigns different priorities to data frames based on the link quality level reflected in the state matrix, and initiates a retransmission mechanism when the link quality is detected to be below a threshold.

[0010] Furthermore, the physical layer sensing unit also integrates a power monitoring circuit, which is used to collect the remaining battery voltage of the monitoring device; When generating the link connection strategy, the inference computing unit will simultaneously read the remaining battery voltage and input the remaining battery voltage as a constraint into the deep reinforcement learning link decision model, so that the link connection strategy will prioritize the communication path with lower energy consumption while considering channel quality.

[0011] Furthermore, the state evolution unit is also configured with abnormal state detection logic, which continuously monitors the node connection values ​​representing critical links in the state matrix. When the node connection value of the critical link changes multiple times in a row, the state evolution unit will trigger an alarm flag, and package the alarm flag and the current state matrix together, and send them out through the protocol adaptation unit.

[0012] Furthermore, the data communication module also includes a data compression submodule, which is located after the physical layer sensing unit. The data compression submodule performs lossless compression on the raw monitoring data collected by the physical layer sensing unit. When processing data, the signal feature extraction unit directly decompresses and extracts features from the compressed data stream processed by the data compression submodule. The decompression process of the data compression submodule outputs the decompressed data to the signal feature extraction unit for subsequent processing.

[0013] Furthermore, the data communication module also includes a security authentication submodule, which is connected to the protocol adaptation unit. Before each data transmission, the security authentication submodule obtains a list of active nodes in the current network from the state evolution unit. The security authentication submodule only allows data to be sent to addresses in the list of active nodes, and adds an encrypted signature based on the node identity to the data frame. The encrypted signature is used by the receiving end to verify the legitimacy of the data source.

[0014] Furthermore, the present invention also includes a self-organizing network transmission line monitoring data feedback system, the system including the self-organizing network transmission line monitoring equipment data communication module as described above, and an aggregation gateway deployed in the monitoring center; The data communication modules of the several self-organizing network transmission line monitoring equipment are distributed in different spans of the transmission line. The modules form a self-organizing network through wireless multi-hop. The modules dynamically maintain the communication links between each other through the inference calculation unit and the state evolution unit. The aggregation gateway periodically broadcasts a network-wide status request instruction to the ad hoc network. After receiving the network-wide status request instruction, the data communication modules of each ad hoc network transmission line monitoring device summarize the status matrix they maintain and send it back to the aggregation gateway through multi-hop routing. The aggregation gateway constructs a communication health status map of the entire transmission line based on all the collected status matrices, and issues new communication parameter configuration instructions to the data communication modules of the designated self-organizing network transmission line monitoring equipment according to the communication health status map. After receiving the communication parameter configuration instructions, each module adjusts its local communication behavior through the protocol adaptation unit.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The system incorporates a deep reinforcement learning-based link decision model, which extracts multi-dimensional feature vectors to characterize channel quality from micro-meteorological parameters, conductor vibration frequency, and tower tilt angle. Based on these multi-dimensional feature vectors, it infers the link connection strategies between adjacent nodes in the ad hoc network topology in real time. This fully captures the dynamic changes in channel quality during transmission line operation, ensuring that the link connection strategies are highly adapted to the actual channel conditions. This avoids the problem of mismatch between fixed link decision rules and actual channels in conventional technologies, reduces link interruptions and data transmission lag, and improves the adaptability and stability of link connections between adjacent nodes.

[0016] Maintain a state matrix reflecting the current connectivity status of the ad hoc network. Update the state matrix based on the inferred link connection strategy so that the state matrix can reflect the network connectivity status at the next moment. This allows protocol adaptation units to directly read the updated state matrix, obtain accurate network connectivity information, and then dynamically adjust the transmission format and routing hop count of data frames. This avoids the problem of untimely and inaccurate adjustment of transmission parameters due to the lag in network connectivity status perception in conventional technologies, optimizes the data transmission process, reduces data transmission latency, and improves the overall data communication efficiency and stability of the ad hoc network. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of the data communication module of the self-organizing network transmission line monitoring equipment described in this invention. Figure 2 This is a flowchart of the internal structure of the signal feature extraction unit; Figure 3 A flowchart for the decision-making process of the inference computing unit link; Figure 4 For transmission control – link quality level and stability analysis; Figure 5 The curves verify the relationship between routing hop count and transmission success rate. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0019] See Figure 1 The data communication module of the self-organizing network transmission line monitoring equipment includes a physical layer sensing unit, a signal feature extraction unit, an inference calculation unit, a state evolution unit, and a protocol adaptation unit.

[0020] The physical layer sensing unit is deployed at monitoring nodes along the transmission line. This unit integrates micro-meteorological sensors, conductor vibration sensors, and tilt sensors to continuously collect micro-meteorological parameters of the transmission line environment, frequency signals of conductor vibration caused by wind, and tilt angle data of the tower structure. The signal feature extraction unit is electrically connected to the physical layer sensing unit. It receives raw sensor data from the physical layer sensing unit and, through processing and analysis of micro-meteorological parameters, conductor vibration frequency signals, and tower tilt angle data, extracts a multi-dimensional feature vector that comprehensively reflects the time-varying characteristics of the wireless channel. The inference computing unit is connected to the signal feature extraction unit. Internally, it stores a pre-trained link decision model based on a deep reinforcement learning algorithm. This inference computing unit uses the multi-dimensional feature vector output by the signal feature extraction unit as the model input. Through forward inference of the model, it calculates and outputs in real time the link connection strategy that should be adopted between adjacent nodes in the ad hoc network topology under the current network environment. This strategy is specifically manifested as the selection of the next-hop relay node or the establishment and disconnection of the link. The State Evolution Unit is connected to the Inference Computation Unit. This unit maintains a dynamic state matrix in memory, where rows and columns correspond to nodes in the network, and element values ​​represent the theoretical connection state between corresponding node pairs. The State Evolution Unit receives link connection strategies from the Inference Computation Unit and updates the corresponding element values ​​in the state matrix based on the connection and disconnection instructions contained in the strategy. This allows the state matrix to reflect the network connectivity status in real time after implementing the new strategy. The Protocol Adaptation Unit is connected to the State Evolution Unit. This unit periodically reads the updated state matrix, parses the network connectivity information contained in the matrix, and dynamically adjusts the underlying parameters of data communication based on this information. Specifically, the Protocol Adaptation Unit adjusts the transmission format of data frames based on the valid path information derived from the state matrix, such as modifying the length of the field in the frame header used to record the routing hop count. Simultaneously, it also allocates differentiated transmission priorities and routing paths for different types or urgency levels of monitoring data based on the link quality reflected in the state matrix.

[0021] In one embodiment of the present invention, the signal feature extraction unit includes a wavelet denoising subunit, a time-frequency analysis subunit, and a vector construction subunit. See also... Figure 2The wavelet denoising subunit is connected to the physical layer sensing unit. It receives the original conductor vibration frequency signal acquired by the physical layer sensing unit. This subunit uses a preset wavelet basis function and threshold strategy to decompose and reconstruct the signal, filtering out interference components caused by environmental background noise to obtain the denoised conductor vibration frequency signal. The time-frequency analysis subunit is connected to the wavelet denoising subunit. This subunit performs joint time-frequency analysis on the denoised conductor vibration frequency signal and synchronously acquired micro-meteorological parameters. Using short-time Fourier transform or wavelet transform methods, it calculates the energy distribution of the signal in the time-frequency domain and extracts channel impulse response variation features that characterize channel multipath effects and Doppler shift changes. The vector construction subunit is connected to the time-frequency analysis subunit. This subunit normalizes the channel impulse response variation features extracted by the time-frequency analysis subunit, the tower tilt angle data directly acquired by the physical layer sensing unit, and the current signal-to-noise ratio measurement value read from the communication chip. After eliminating the influence of dimensions, these features are concatenated in a predetermined order to generate a fixed-dimensional multidimensional feature vector.

[0022] In practical implementation, the wavelet denoising subunit directly receives the conductor vibration frequency signal output by the physical layer sensing unit, such as a discrete signal sequence containing environmental noise interference in the time domain. Specifically, the wavelet denoising subunit uses the Daubechies wavelet basis functions to perform multi-scale decomposition on the input original conductor vibration frequency signal, decomposing the signal into approximate coefficients and detail coefficients at different frequency subbands. Subsequently, the wavelet denoising subunit applies a soft thresholding function to the detail coefficients obtained from each decomposition, setting coefficients with amplitudes below a preset threshold to zero or shrinking them to suppress noise components. Finally, the wavelet denoising subunit uses the processed coefficients to reconstruct the signal, outputting the denoised conductor vibration frequency signal. Before and after this process, the data characteristics of the conductor vibration frequency signal change significantly. For example, the original signal may have a broadband noise floor with an amplitude of approximately 0.5mV in the 100Hz to 150Hz frequency band, while the noise floor amplitude of the signal in the same frequency band after denoising can be reduced to below 0.05mV, highlighting the dominant vibration frequency component of the conductor itself around 8Hz.

[0023] The time-frequency analysis subunit receives the denoised conductor vibration frequency signal from the wavelet denoising subunit and micro-meteorological parameters, such as wind speed and direction time series, synchronously collected by the physical layer sensing unit. In practice, the time-frequency analysis subunit performs a windowed short-time Fourier transform on the denoised conductor vibration frequency signal, calculates the energy distribution of the signal in the time-frequency domain, and obtains the signal's time-frequency spectrum. Simultaneously, the time-frequency analysis subunit uses wind speed parameters as a modulation factor and performs correlation analysis with the time-varying characteristics of the signal. It can be understood that the time-frequency analysis subunit quantifies the drastic changes in the channel impulse response using a time-frequency clustering metric formula. This metric formula is expressed as: Where: characters Quantized values ​​representing the characteristics of channel impulse response changes, characters Represents the total number of time windows, characters Represents the total number of frequency modules. Indicates at a point in time and frequency point The magnitude of the short-time Fourier transform coefficients at that point. Indicates at a point in time The normalized wind speed value. The quantized value calculated by the time-frequency analysis subunit. Together with the identified signal frequency offset, they constitute the changing characteristics of the channel impulse response.

[0024] The vector construction subunit is responsible for integrating feature data from different sources. It receives channel impulse response variation characteristics output from the time-frequency analysis subunit, raw tower tilt angle values ​​directly acquired by the physical layer sensing unit, and current signal-to-noise ratio measurements obtained from the communication module's RF front-end. In practice, the vector construction subunit performs normalization preprocessing on these four heterogeneous feature data. For the quantized values ​​of the channel impulse response variation characteristics... The vector construction subunit uses a max-min normalization method to map it to the [0,1] interval. The original values ​​of the tower tilt angle and the signal-to-noise ratio measurement are also normalized using a linear scaling method based on their preset physical range. Optionally, the main frequency offset is normalized using its absolute value. After normalizing all features, the vector construction subunit concatenates the processed quantized values ​​of the channel impulse response change features, the normalized values ​​of the tower tilt angle, the normalized values ​​of the signal-to-noise ratio, and the normalized values ​​of the main frequency offset in a preset order to generate a four-dimensional feature vector, such as a multi-dimensional feature vector in the form of [0.75, 0.12, 0.95, 0.03], for use by the inference calculation unit.

[0025] In one embodiment of the present invention, the inference computation unit includes a model loading subunit, an action space generation subunit, a benefit evaluation subunit, and a policy output subunit. See also... Figure 3The model loading subunit is connected to the signal feature extraction unit. This subunit loads the pre-trained deep reinforcement learning link decision model from non-volatile memory and uses the multi-dimensional feature vector generated in real time by the signal feature extraction unit as the model's input state at the current moment. The action space generation subunit is connected to the model loading subunit. This subunit generates a candidate action set containing multiple possible routing schemes based on the local neighbor node list maintained by the ad hoc network protocol stack, such as choosing a different neighbor node as the next hop or remaining silent. The benefit evaluation subunit is connected to the action space generation subunit. For each routing scheme in the candidate action set, this subunit performs rapid simulation within the model, predicts the network state changes over a period of time after adopting the action, and calculates the expected benefit value for each scheme according to a preset reward function, which typically considers link stability, throughput, and latency. The policy output subunit is connected to the benefit evaluation subunit. This subunit compares the expected benefit values ​​calculated for all candidate action schemes, selects the routing scheme with the highest expected benefit value, and outputs it as the final link connection policy to the state evolution unit.

[0026] In specific implementation, the model loading subunit is connected to the signal feature extraction unit and receives multi-dimensional feature vectors from the signal feature extraction unit, such as a four-dimensional vector [0.8, 0.1, 0.9, 0.05]. The model loading subunit loads a pre-trained deep reinforcement learning link decision model from the device's flash memory. In some embodiments, the deep reinforcement learning link decision model adopts a deep Q-network architecture. The model loading subunit inputs the multi-dimensional feature vector [0.8, 0.1, 0.9, 0.05] as the current environment state into the input layer of the deep reinforcement learning link decision model. It can be understood that the deep reinforcement learning link decision model has been trained offline in the simulation environment before deployment, and its network parameters have been fixed. The action space generation subunit works in conjunction with the model loading subunit. The action space generation subunit accesses the local neighbor discovery table maintained by the ad hoc networking protocol stack. This table records the identifiers of currently communicable neighboring nodes. For example, the neighbor table of node A includes nodes B, C, and D. The action space generation subunit generates a candidate action set based on this neighbor list. In some embodiments, the candidate action set contains multiple routing schemes, which may be in the form of {"Next hop: Node B", "Next hop: Node C", "Next hop: Node D", "Silent"}. Each action in the candidate action set represents an immediate link connection selection.

[0027] The revenue evaluation subunit evaluates each routing scheme in the candidate action set. Within a simulation environment built on a deep reinforcement learning link decision model, the revenue evaluation subunit performs forward inference for each candidate action. For example, for the candidate action "Next hop: Node B," the revenue evaluation subunit simulates the execution of this action and predicts the network state changes caused by selecting Node B as a relay within a future decision cycle, including changes in link stability, packet delivery rate, and end-to-end latency. The revenue evaluation subunit calculates the expected revenue value for each action based on a preset reward function. In practice, the reward function can be quantified as a combination of multiple network performance indicators, and the formula for calculating the expected revenue value is expressed as: Where: characters Representative of the candidate action set The expected revenue value calculated from each routing scheme, character The character represents the link stability factor derived from simulation. The throughput factor derived from simulation is represented by the number of characters. This represents the time delay factor derived from simulation. (Character) , , These represent the weighting coefficients of the stability factor, throughput factor, and latency factor in the total revenue, respectively. These coefficients are determined during the training phase of the deep reinforcement learning link decision model. Optionally, for the action "next hop: node B", the stability factor may be obtained after simulation in the revenue evaluation sub-unit. Throughput factor Delay factor If the weighting coefficient Then the expected return value The revenue assessment subunit calculates the expected revenue value of the action "Next Jump: Node C" using the same method. The expected reward value of the action "Next hop: Node D" And the expected return value of "silent" action. The strategy output subunit receives the calculation results from the return evaluation subunit. The strategy output subunit compares all expected return values. , , , The routing scheme corresponding to the highest expected return value is selected from these options. In specific implementation, the expected return values ​​of 0.60, 0.55, 0.40, and 0.10 are compared, and the strategy output sub-unit determines the expected return value. It is the maximum value. Therefore, the strategy output subunit will be... The corresponding routing scheme "Next Hop: Node B" is determined as the final link connection strategy, and this strategy is output to the state evolution unit. It can be understood that through the above process, the inference and computation unit completes the closed-loop computation from environmental state perception to optimal link decision.

[0028] In one embodiment of the present invention, the state evolution unit includes a matrix initialization subunit, a state update subunit, and an evolution prediction subunit. The matrix initialization subunit operates when the monitoring device is powered on. This subunit generates an initial state matrix based on pre-configured node deployment location information. The values ​​of the elements in the matrix are set according to the theoretical communication distance and line-of-sight between nodes, representing the initial theoretical connection state between nodes. The state update subunit is connected to the matrix initialization subunit and the inference calculation unit. This subunit receives the link connection strategy from the inference calculation unit, parses and maps the connection and disconnection instructions explicitly specified in the strategy into specific modification operations on the corresponding row and column element values ​​in the state matrix, for example, changing an element value from 0 (disconnected) to 1 (connected). The evolution prediction subunit is connected to the state update subunit. After the state update subunit completes the matrix modification for the current time step, this subunit uses the modified state matrix as the initial state and employs a built-in Markov chain model to predict the future evolution of the network topology. It calculates the probability distribution of changes in the state of each link within several future time steps and feeds this predicted trend back to the current state matrix for secondary correction, such as marking links with a high probability of disconnection in advance. In addition, the state evolution unit is also equipped with abnormal state detection logic. This logic continuously monitors the node connection values ​​representing critical links in the state matrix. When it detects multiple drastic jumps in the node connection value of a critical link within a short period, the abnormal state detection logic determines that the network is experiencing abnormal oscillations, triggers an internal alarm flag, and packages this alarm flag along with the complete state matrix data at the trigger time, sending it to the network management terminal via the protocol adaptation unit.

[0029] In specific implementation, the matrix initialization subunit is executed during the monitoring device startup phase. The subunit reads the node deployment location configuration file stored in the device. This configuration file contains the latitude and longitude coordinates of all monitoring nodes in the network, for example, nodes 1 to 4. Based on the theoretical communication distance and line-of-sight connectivity between nodes, the matrix initialization subunit calculates and generates an initial state matrix. In some embodiments, for a network containing four nodes, the initial state matrix is ​​a 4x4 square matrix, where row index i represents the source node, column index j represents the destination node, and matrix element values... This represents the theoretical connection state from node i to node j. If node i and node j are within the theoretical communication range and there are no obstructions, the matrix element values ​​are... Set to 1; if node i and node j are outside the communication range or blocked by obstacles, the matrix element value is changed. Set to 0, and set the connection value from the node to itself to 1. Based on the position calculation, assuming that node 1 is connected to node 2 and node 3, but disconnected from node 4, the generated initial state matrix can be: first row [1,1,1,0], second row [1,1,0,1], third row [1,0,1,0], fourth row [0,1,0,1].

[0030] The state update subunit is connected to the matrix initialization subunit and the inference computation unit. The state update subunit receives link connection policies from the inference computation unit. These policies are in the form of instructions, such as "establish a connection with node 3" or "disconnect from node 2". The state update subunit parses the link connection policies, mapping the policy instructions to modification operations on specific elements in the state matrix. In a specific implementation, if the current state matrix is ​​as described in the previous example, and the link connection policy instruction is "connect node 1 to node 4", the state update subunit will modify the value of the element in the first row and fourth column of the state matrix. Change from 0 to 1. If the link connection policy instruction is "Node 2 disconnects from Node 4", the state update subunit will change the value of the element in the second row and fourth column of the state matrix. The state matrix is ​​changed from 1 to 0. After the state update subunit completes the modification, the state matrix is ​​updated from the initial [[1,1,1,0],[1,1,0,1],[1,0,1,0],[0,1,0,1]] to [[1,1,1,1],[1,1,0,0],[1,0,1,0],[0,1,0,1]]. This matrix reflects the network connectivity after the new strategy is implemented.

[0031] The evolutionary prediction subunit starts after the state update subunit completes its operation. Using the updated state matrix as the initial network state at the current moment, the evolutionary prediction subunit predicts the network topology change trend over several future time steps using a built-in Markov chain model. In specific implementation, the evolutionary prediction subunit maintains a state transition probability matrix for each element in the state matrix based on historical state transition statistics. The evolutionary prediction subunit calculates and predicts the probability that the states of each link will remain unchanged or change in the next time step. The calculation formulas involved in the prediction process are expressed as follows: Where: characters This represents the state probability vector formed after vectorizing the current state matrix. This represents the global state transition probability matrix learned from historical link stability data, and the character... This represents the predicted probability vector of the next state. The evolutionary prediction sub-unit will calculate this... The vector is converted back to matrix form and compared with the current state matrix. For example, the prediction result shows the elements. The probability of changing from 0 to 1 in the next time step is as high as 0.9. The evolutionary prediction sub-unit marks such high-probability predicted changes as "observable events" and uses the prediction results to perform secondary corrections on the current state matrix, for example, by adding elements to the metadata of the state matrix. Add a flag indicating a high probability of reconnection recovery to provide forward-looking information to the protocol adaptation unit.

[0032] Furthermore, the abnormal state detection logic configured in the state evolution unit operates independently and continuously. This logic continuously monitors the node connection values ​​representing critical links in the state matrix. These critical links are predefined by the configuration, such as the link from node 1 to node 4. Designated as a critical uplink. The anomaly detection logic samples the element values ​​of the critical link in the state matrix at fixed intervals. It can be understood that if multiple jumps in the node connection value of the critical link are detected within several consecutive sampling periods, for example, within 5 consecutive periods, the element... The value sequence is [1,0,1,0,0], with four switches between 0 and 1. The anomaly detection logic determines this pattern as an abnormal oscillation. Once such an anomaly is detected, the anomaly detection logic immediately triggers an internal alarm flag and sets its value to "true". Simultaneously, the anomaly detection logic captures a snapshot of the complete state matrix at the moment the alarm is triggered. Optionally, the state evolution unit packages the alarm flag and the state matrix snapshot data into an alarm message. Finally, this alarm message is transmitted to the protocol adaptation unit through the interface between the state evolution unit and the protocol adaptation unit, and then sent by the protocol adaptation unit to the network management terminal.

[0033] In one embodiment of the present invention, the protocol adaptation unit includes a routing table reconstruction subunit, a frame structure adjustment subunit, and a transmission control subunit. The routing table reconstruction subunit is connected to the state evolution unit. This subunit parses the latest state matrix maintained by the state evolution unit, uses graph theory algorithms to identify all possible valid data transmission paths in the current network environment, and generates or updates a real-time dynamic routing table based on path quality. The frame structure adjustment subunit is connected to the routing table reconstruction subunit. This subunit dynamically adjusts the encapsulation format of the data frame according to the hop count of the path selected by the dynamic routing table for the current data packet. If the path hop count is high, the length of the field used to store routing information in the frame header is increased; if the path hop count is low, the length of this field is reduced to save overhead. The transmission control subunit is connected to the frame structure adjustment subunit and the state evolution unit. This subunit assigns different transmission priorities and queue scheduling strategies to the data frames to be sent based on the link quality levels mapped by the link element values ​​in the state matrix. After data transmission, if a feedback mechanism detects that the quality level of a certain link is continuously lower than a preset threshold, a retransmission mechanism for the data frames of that link is automatically initiated. The protocol adaptation unit is also connected to a security authentication submodule. Before the protocol adaptation unit prepares to send each data frame, the security authentication submodule obtains a list of active nodes in the current network from the state evolution unit. Based on this list, the submodule filters the destination addresses of the data frames, allowing data to be sent only to node addresses on the list. Simultaneously, the security authentication submodule adds an encrypted signature generated based on the node's identifier and private key to a specific field in the data frame. This encrypted signature is used by the data receiver to verify the legitimacy of the data source, preventing unauthorized node access or data tampering.

[0034] In implementation, the routing table reconstruction subunit is connected to the state evolution unit and periodically reads the latest state matrix maintained by the state evolution unit. In this implementation, the state matrix is ​​a 4x4 matrix representing the connection status from node 1 to node 4, where a value of 1 indicates connectivity and 0 indicates disconnection. Assuming the current state matrix is: rows represent source nodes, columns represent destination nodes, and the first row [1,1,0,1] indicates that node 1 is connected to nodes 1, 2, and 4, but disconnected from node 3. The routing table reconstruction subunit parses this state matrix and uses the shortest path algorithm to identify all valid data transmission paths. The routing table reconstruction subunit calculates the path for each destination node and calculates a comprehensive quality score for each path based on the path hop count and link stability factor. The routing table reconstruction subunit updates the optimal path information to the dynamic routing table. In some embodiments, the dynamic routing table includes the destination node, next-hop node, total path hop count, and path quality score, as shown in Table 1.

[0035] Table 1: Dynamic Routing Table The frame structure adjustment subunit is connected to the routing table reconstruction subunit. When encapsulating a data frame to be sent, the frame structure adjustment subunit first queries the dynamic routing table to determine the path to the destination address of the data frame. The frame structure adjustment subunit reads the value of the "total hop count" field in the dynamic routing table. Based on the total hop count, the frame structure adjustment subunit dynamically calculates and adjusts the length of the routing information field in the data frame header. In specific implementations, the data frame header needs to include the temporary address of each hop on the path, and the frame structure adjustment subunit determines the address storage space that needs to be reserved based on the total hop count. It can be understood that the adjustment of the frame header length is achieved through a linear formula, expressed as: Where: characters This represents the total length of the routing information field in the calculated data frame header, in bytes and characters. Represents the total number of hops for the path retrieved from the dynamic routing table. (Character) This represents the standard length allocated to each hop address in the routing information field, for example, 4 bytes, characters. This represents the constant length of fixed fields in the frame header, excluding routing information, such as 6 bytes. If the current data frame needs to be sent to node 3, the total number of hops along the path is determined by querying the dynamic routing table. , preset , The frame structure adjustment subunit is calculated to obtain Bytes. Accordingly, the frame structure adjustment subunit configures the routing information field in the data frame header to be 14 bytes long, and reserves 2 address bits (4 bytes each) in it to record the path "node 1 -> node 4 -> node 3".

[0036] The transmission control subunit works in conjunction with the frame structure adjustment subunit and the state evolution unit. The transmission control subunit assigns priorities to data frames based on the link quality levels reflected in the state matrix. In some embodiments, the metadata of the state matrix includes an evaluation level for each link. The transmission control subunit reads the level of each hop link on the current transmission path and takes the minimum value as the link quality level of that frame. The transmission control subunit maps the link quality level to transmission priority according to a preset mapping rule. For example, level 5 is mapped to "highest priority", level 3 to "normal priority", and level 1 to "lowest priority". The transmission control subunit places data frames of different priorities into different transmission queues. It can be understood that the transmission control subunit continuously monitors acknowledgment signals from the receiving end after data frame transmission. If the transmission control subunit detects that a link to a specific next-hop node has not received acknowledgment after multiple consecutive data frame transmissions, and the level of that link in the state matrix has been updated to below a preset threshold, the transmission control subunit determines that the link quality does not meet the requirements. At this time, the transmission control subunit initiates a retransmission mechanism for data frames sent via the link but not acknowledged, extracts a copy from the sending queue, and selects another alternative path according to the dynamic routing table or waits for the original link to recover before retransmitting.

[0037] The security authentication submodule is connected to the protocol adaptation unit and plays a role in the data transmission process. Before the protocol adaptation unit prepares to send any data frame, the security authentication submodule sends a request to the state evolution unit to obtain the list of active nodes in the current network. The list of active nodes is dynamically generated by the state evolution unit based on the nodes whose connection values ​​are not zero at the current moment in the state matrix. For example, the state evolution unit returns the list [node 1, node 2, node 4]. The security authentication submodule checks the destination address of the data frame to be sent. If the destination address is not in the list of active nodes, the security authentication submodule blocks the transmission of the data frame. For data frames that are allowed to be sent, the security authentication submodule generates an encrypted signature for them. In specific implementations, the security authentication submodule uses the private key of the sending node to perform a hash operation on the payload of the data frame and digitally signs the hash value. Optionally, the security authentication submodule appends the generated encrypted signature to a specific field at the end of the data frame. After receiving the data frame, the receiving node uses the public key of the sending node to verify the encrypted signature, thereby confirming the legitimacy of the data source and whether the data has been tampered with during transmission.

[0038] See Figure 4In the data transmission control process of self-organizing network transmission lines, the joint analysis of link quality level and stability factor provides a core basis for dynamic routing scheduling and transmission strategy optimization. The figure uses a bar chart to represent the link quality level (range 1–5, higher values ​​indicate better physical layer communication quality) and a line chart to represent the link stability factor (range 0–1, values ​​closer to 1 indicate smaller fluctuations in link connectivity and higher reliability). It covers five typical communication links between nodes 1 and 4: Link 1→2: Link quality level is 5, stability factor reaches 0.95, a backbone link with excellent quality and stability in the current network, capable of carrying high-priority monitoring data and real-time control commands. Link 1→4: Link quality level is 4, stability factor is 0.90, a suboptimal link, suitable as an alternative path for the backbone link. Link 2→3: Link quality level is 3, stability factor drops to 0.80, a bottleneck link with weak quality and stability in the network, requiring triggering a retransmission mechanism and path redundancy backup strategy. Link 3→4: With a link quality level of 4 and a stability factor of 0.85, it exhibits good quality but moderate stability, suitable for carrying non-real-time monitoring data. Link 4→3: With a link quality level of 5 and a stability factor of 0.92, it, together with link 1→2, forms a highly reliable core transmission channel in the network, and can be used for priority forwarding of critical alarm data. Based on the combined distribution of the link quality levels and stability factors, the transmission control subunit assigns differentiated priorities and queue scheduling strategies to data frames on different paths: High-level, high-stability links (such as 1→2 and 4→3) are assigned the highest priority, allowing low-latency forwarding; low-level, low-stability links (such as 2→3) have their priority reduced, and retransmission and path switching mechanisms are initiated when the link quality consistently falls below a threshold. This ensures reliable backhaul of critical monitoring data while optimizing overall network resource utilization and energy consumption.

[0039] In one embodiment of the invention, the physical layer sensing unit also integrates a power monitoring circuit, which collects and monitors the remaining battery voltage of the monitoring device's power supply in real time. When the inference computing unit runs its internal deep reinforcement learning link decision model to make decisions, it simultaneously reads the remaining battery voltage value. The inference computing unit uses the remaining battery voltage as an additional input constraint, inputting it into the decision model along with the multi-dimensional feature vector. When calculating the expected benefits of different link connection strategies, the model couples the estimated communication energy consumption corresponding to the strategy with the current remaining power for evaluation. This ensures that the final generated link connection strategy optimizes channel quality and network performance while prioritizing communication paths with lower energy consumption to extend node lifespan. Following the physical layer sensing unit, the data communication module also includes a data compression submodule. This data compression submodule performs lossless compression encoding on the raw monitoring data stream collected by the physical layer sensing unit in the time or frequency domain to reduce data volume. When performing feature extraction, the signal feature extraction unit directly reads the compressed data stream output by the data compression submodule. This data stream is first decompressed and restored to its original data format within the signal feature extraction unit. The decompressed data is then sent to subsequent wavelet denoising, time-frequency analysis, and other processing steps.

[0040] In practical implementation, the power monitoring circuit integrated into the physical layer sensing unit continuously collects the remaining battery voltage of the monitoring device's power supply. The power monitoring circuit samples the battery voltage at fixed intervals using an analog-to-digital converter, for example, once every 10 seconds, obtaining a discrete sequence of remaining battery voltage values. The original voltage values ​​obtained from the sampling are in millivolts. The power monitoring circuit converts the collected analog voltage signal into a digital value, which is then read by the inference computing unit via the device's internal bus. When the inference computing unit runs the deep reinforcement learning link decision model to make decisions, it not only receives multi-dimensional feature vectors as input states but also simultaneously reads the latest digital value of the remaining battery voltage. The inference computing unit inputs the remaining battery voltage as an independent constraint condition into the deep reinforcement learning link decision model. In some embodiments, the remaining battery voltage is normalized before input; for example, the nominal voltage of 3700 millivolts is mapped to 1.0, the cutoff voltage of 2800 millivolts is mapped to 0.0, and the currently sampled voltage of 3500 millivolts is normalized to approximately 0.737. When evaluating the expected benefit of each routing option in the candidate action set, the deep reinforcement learning link decision model couples the estimated communication energy consumption of each candidate path with the current normalized remaining battery voltage value. This means that the model's reward function includes considerations of energy efficiency. The model quantifies the energy consumption impact of choosing different paths using an energy cost formula, which is expressed as: Where: characters The representative deep reinforcement learning link decision model is the first The dimensionless energy cost factor calculated for each routing scheme. This represents the normalized value of the estimated communication energy consumption for completing one data forward using this path, derived from simulations. This value is calculated by dividing the estimated energy consumption by a reference energy consumption. Therefore This is a dimensionless ratio. (Character) This represents the remaining battery voltage value read from the power monitoring circuit and normalized; this value is a dimensionless value between 0 and 1. (Character) This represents the energy cost weighting coefficient, a dimensionless constant determined during model training. For example, the normalized value of the current remaining battery voltage. For path A, its estimated energy consumption is 0.8 times the reference energy consumption, that is... For path B, , preset Then the energy cost factor of path A Energy cost factor of path B When calculating the total expected return for path A and path B, the deep reinforcement learning link decision model subtracts the corresponding energy cost factor from the basic return. Therefore, although path A may have a slight advantage in channel quality, its higher energy cost factor is a disadvantage. This will reduce its total revenue, causing the model to ultimately generate a link connection strategy that tends to choose path B, which consumes less energy, thereby extending the node's working life.

[0041] The data compression submodule is located after the physical layer sensing unit and before the signal feature extraction unit. It performs real-time lossless compression on the raw monitoring data collected by the physical layer sensing unit. In practice, the raw monitoring data output by the physical layer sensing unit is a mixed data stream containing micrometeorological parameters, conductor vibration frequency, and tower tilt angle. The data is generated periodically in fixed-format array frames, for example, each frame is 200 bytes in size. The data compression submodule uses a lightweight lossless compression algorithm to compress each frame of raw data. It reads a 200-byte frame of raw data, compresses it, and outputs a smaller compressed data stream, for example, 120 bytes. The compressed data stream is placed in a buffer, awaiting processing by the signal feature extraction unit. The signal feature extraction unit directly reads the compressed data stream from the output buffer of the data compression submodule. The signal feature extraction unit contains corresponding decompression logic, which decompresses the compressed data stream to recover a 200-byte data frame completely identical in format and content to the original data. In some embodiments, the decompression process is a necessary step before the signal feature extraction unit performs feature extraction. Optionally, the decompression logic of the signal feature extraction unit strictly corresponds to the compression algorithm of the data compression submodule to ensure lossless data restoration. The signal feature extraction unit then performs subsequent processing procedures such as wavelet denoising and time-frequency analysis on the decompressed complete data frame. In this way, the data transmission bandwidth requirement from the physical layer sensing unit to the signal feature extraction unit is reduced, while the data ultimately used by the signal feature extraction unit to generate multidimensional feature vectors remains consistent with that used when directly processing the original data.

[0042] See Figure 5In the routing strategy verification of the data communication module of the self-organizing network transmission line monitoring equipment, the correlation between transmission success rate and routing hop count was quantified by comparing two decision schemes. Specifically, the horizontal axis represents the routing hop count during self-organizing network data transmission, ranging from 1 to 5 hops; the vertical axis represents the data transmission success rate for the corresponding hop count, ranging from 0.55 to 1.00. The figure contains two comparison curves: the dotted curve represents the traditional routing decision scheme without considering energy consumption optimization, which uses channel quality as the core decision criterion; the boxed curve represents the scheme considering energy consumption optimization (the scheme of this invention), which introduces the remaining battery voltage as a constraint in the deep reinforcement learning link decision model, coupling communication energy consumption and remaining power consumption into an energy cost factor, which is then incorporated into the total revenue evaluation. Data trends show that as the hop count increases from 1 to 5, the transmission success rate of both schemes decreases: Without considering energy optimization, the success rate is approximately 0.98% with 1 hop, decreasing to 0.60% with 5 hops, a decrease of 0.38%. Considering energy optimization, the success rate is approximately 0.98% with 1 hop, decreasing to 0.78% with 5 hops, a decrease of only 0.20%. At each hop count node, the transmission success rate of this invention is significantly higher than the traditional scheme: the difference is approximately 0.04% with 3 hops, widening to 0.10% with 4 hops, and reaching 0.18% with 5 hops, demonstrating the performance gain of the energy optimization strategy in multi-hop transmission scenarios. This result stems from the deep reinforcement learning model's weighted consideration of energy costs in the benefit evaluation, prioritizing low-energy paths to reduce the risk of link interruption due to node power depletion, thus maintaining higher transmission reliability in multi-hop transmission. At the parameter configuration level, this scheme presets the energy cost weight coefficient during the model training stage, normalizes the remaining battery voltage to the [0,1] interval, and calculates the energy cost factor by coupling it with the normalized communication energy consumption of the candidate path. This factor is directly deducted from the basic benefit, thereby achieving the coordinated optimization of channel quality and energy efficiency.

[0043] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A data communication module for self-organizing network transmission line monitoring equipment, characterized in that, include: Physical layer sensing units are deployed at monitoring nodes along the transmission line. These units are used to collect micro-meteorological parameters, conductor vibration frequency, and tower tilt angle of the transmission line. A signal feature extraction unit connected to the physical layer sensing unit extracts a multi-dimensional feature vector from the micro-meteorological parameters, conductor vibration frequency, and tower tilt angle to characterize channel quality. The inference computing unit is connected to the signal feature extraction unit. The inference computing unit has a built-in link decision model based on deep reinforcement learning. The inference computing unit infers the link connection strategy between each adjacent node in the ad hoc network topology in real time based on the multi-dimensional feature vector. A state evolution unit connected to the inference computing unit maintains a state matrix reflecting the current connectivity state of the ad hoc network. The state evolution unit updates the state matrix according to the link connection strategy so that the state matrix can reflect the network connectivity status at the next moment. A protocol adaptation unit connected to the state evolution unit reads the updated state matrix and dynamically adjusts the transmission format and routing hop count of the data frame based on the connectivity information in the state matrix.

2. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 1, characterized in that, The signal feature extraction unit includes: A wavelet denoising subunit connected to the physical layer sensing unit performs noise reduction processing on the conductor vibration frequency signal to eliminate interference caused by environmental noise. The time-frequency analysis subunit connected to the wavelet denoising subunit performs joint time-frequency analysis on the denoised conductor vibration frequency signal and the micro-meteorological parameters to extract the variation characteristics of the channel impulse response. The vector construction subunit connected to the time-frequency analysis subunit normalizes and splices together the channel impulse response variation characteristics, the tower tilt angle, and the current signal-to-noise ratio measurement value to generate the multidimensional feature vector.

3. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 2, characterized in that, The reasoning and calculation unit includes: A model loading subunit connected to the signal feature extraction unit loads the pre-trained deep reinforcement learning link decision model and uses the multi-dimensional feature vector as the input state. The action space generation subunit, which is connected to the model loading subunit, generates a set of candidate actions containing multiple routing schemes based on the current list of neighbor nodes of the ad hoc network. A benefit evaluation subunit is connected to the action space generation subunit. The benefit evaluation subunit simulates and deduces each routing scheme in the candidate action set and calculates the expected benefit value corresponding to each scheme based on the deduction results. The policy output subunit, which is connected to the benefit evaluation subunit, selects the routing scheme with the largest expected benefit value and outputs it as the link connection policy.

4. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 3, characterized in that, The state evolution unit includes: A matrix initialization subunit connected to the inference calculation unit generates an initial state matrix based on the preset node deployment positions when the monitoring device is started. The element values ​​in the state matrix represent the theoretical connection state between nodes. A state update subunit connected to the matrix initialization subunit receives the link connection strategy and maps the connection and disconnection instructions contained in the link connection strategy into modification operations on the corresponding element values ​​in the state matrix. The evolution prediction subunit, connected to the state update subunit, uses a Markov chain model to predict the network topology change trend in the next few time steps based on the modified state matrix, and feeds the prediction results back to the state matrix for secondary correction.

5. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 4, characterized in that, The protocol adaptation unit includes: A routing table reconstruction subunit is connected to the state evolution unit. The routing table reconstruction subunit parses the state matrix, identifies the valid data transmission paths in the current network, and generates a real-time dynamic routing table accordingly. A frame structure adjustment subunit connected to the routing table reconstruction subunit automatically increases or decreases the length of the routing information field in the data frame header according to the transmission path length indicated in the dynamic routing table. The transmission control subunit, which is connected to the frame structure adjustment subunit, assigns different priorities to data frames based on the link quality level reflected in the state matrix, and initiates a retransmission mechanism when the link quality is detected to be below a threshold.

6. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 5, characterized in that, The physical layer sensing unit also integrates a power monitoring circuit, which is used to collect the remaining battery voltage of the monitoring device. When generating the link connection strategy, the inference computing unit will simultaneously read the remaining battery voltage and input the remaining battery voltage as a constraint into the deep reinforcement learning link decision model, so that the link connection strategy will prioritize the communication path with lower energy consumption while considering channel quality.

7. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 6, characterized in that, The state evolution unit is also configured with abnormal state detection logic, which continuously monitors the node connection values ​​representing critical links in the state matrix. When the node connection value of the critical link changes multiple times in a row, the state evolution unit will trigger an alarm flag, and package the alarm flag and the current state matrix together, and send them out through the protocol adaptation unit.

8. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 7, characterized in that, The data communication module also includes a data compression submodule, which is located after the physical layer sensing unit. The data compression submodule performs lossless compression on the raw monitoring data collected by the physical layer sensing unit. When processing data, the signal feature extraction unit directly decompresses and extracts features from the compressed data stream processed by the data compression submodule. The decompression process of the data compression submodule outputs the decompressed data to the signal feature extraction unit for subsequent processing.

9. The data communication module of the self-organizing network transmission line monitoring equipment as described in claim 8, characterized in that, The data communication module also includes a security authentication submodule, which is connected to the protocol adaptation unit. Before each data transmission, the security authentication submodule obtains the list of active nodes in the current network from the state evolution unit. The security authentication submodule only allows data to be sent to addresses in the list of active nodes, and adds an encrypted signature based on the node identity to the data frame. The encrypted signature is used by the receiving end to verify the legitimacy of the data source.

10. A self-organizing network transmission line monitoring data feedback system, characterized in that, It includes the data communication module of the self-organizing network transmission line monitoring equipment as described in any one of claims 1 to 9, and the aggregation gateway deployed in the monitoring center; The data communication modules of the several self-organizing network transmission line monitoring equipment are distributed in different spans of the transmission line. The modules form a self-organizing network through wireless multi-hop. The modules dynamically maintain the communication links between each other through the inference calculation unit and the state evolution unit. The aggregation gateway periodically broadcasts a network-wide status request instruction to the ad hoc network. After receiving the network-wide status request instruction, the data communication modules of each ad hoc network transmission line monitoring device summarize the status matrix they maintain and send it back to the aggregation gateway through multi-hop routing. The aggregation gateway constructs a communication health status map of the entire transmission line based on all the collected status matrices, and issues new communication parameter configuration instructions to the data communication modules of the designated self-organizing network transmission line monitoring equipment according to the communication health status map. After receiving the communication parameter configuration instructions, each module adjusts its local communication behavior through the protocol adaptation unit.

Images