Real-time data stable transmission method and system for ultra-far offshore distance

Abstract

The invention relates to the technical field of ultra-far offshore data transmission, and discloses an ultra-far offshore distance real-time data stable transmission method and system. The method comprises the following steps: acquiring target transmission link data, and extracting dynamic fluctuation characteristics containing signal strength distribution information and a delay jitter relationship from the target transmission link data; calling the trained channel estimation network to carry out signal positioning processing on transmission link data, and generating a signal transmission path set containing a channel type identifier and phase code mapping relation; based on the set, calling a waveform analysis network to perform multi-modal feature fusion on the signal intensity distribution information to obtain a fusion result containing time-frequency domain features and modulation mode features; and inputting a fusion result into a transmission strategy model to perform dynamic path selection, and outputting a real-time data transmission element containing a control instruction generated based on a signal transmission sequence and a phase self-calibration parameter. According to the method, stable transmission of real-time data under the ultra-far offshore distance can be realized.

Description

Technical Field

[0001] The present invention relates to the field of offshore wind power technology, and in particular to a method and system for stable transmission of real-time data at an ultra-far-from-shore distance. Background Art

[0002] With the rapid development of marine resource development, ocean shipping, and deep-sea exploration, the demand for real-time data transmission over vast distances from shore is becoming increasingly urgent. In these scenarios, data transmission faces extremely complex environmental challenges, and transmission links are often subject to interference from various factors, making it difficult to ensure stable and real-time data transmission. In traditional offshore data transmission, commonly used satellite communication methods have problems such as limited bandwidth and high transmission delay. Especially in ultra-long-distance situations, the signal attenuation is severe and it is easily affected by natural phenomena such as clouds and ionospheric fluctuations, which increases the data packet loss rate and cannot meet application scenarios with high real-time requirements, such as real-time monitoring data feedback in deep-sea scientific research and dynamic navigation information exchange of ocean-going ships.

[0003] Although ground-based microwave relay technology can achieve high-quality transmission within a certain distance, in ultra-far-shore scenarios, due to the influence of the earth's curvature, a large number of relay stations need to be set up. Not only is the construction cost high, but in the marine environment, the relay equipment is susceptible to salt spray corrosion, wave impact, etc., making maintenance extremely difficult. At the same time, the transmission distance is also significantly limited, making it difficult to cover vast offshore areas. Existing wireless self-organizing network technologies face challenges with limited node energy and dynamic topology changes when handling ultra-long-distance transmission. Signals travel through various channels during transmission, such as ocean reflection and atmospheric scattering. These channels have significantly different transmission characteristics, resulting in varying signal attenuation rates and delay jitter. This makes traditional fixed-path transmission methods incapable of adapting to dynamic channel changes. Furthermore, during signal transmission, multipath effects can cause phase shifts and interference. Traditional signal processing methods struggle to accurately extract and effectively fuse multimodal features, causing transmission strategy adjustments to lag behind channel variations, impacting data transmission stability. Furthermore, existing technologies often select signal transmission paths based on static parameters, failing to dynamically adjust based on real-time signal strength distribution and delay jitter. This makes data transmission prone to interruptions or errors in complex environments. As data transmission volumes continue to increase, the demands for transmission efficiency and reliability are becoming increasingly stringent. Existing data transmission methods, when used in complex environments at ultra-farshore distances, have gradually exposed problems such as poor adaptability and weak anti-interference capabilities, failing to meet the diverse needs of practical applications. Therefore, developing a method that can adapt to the complex environments of ultra-farshore distances and achieve stable real-time data transmission has become a pressing issue in this field. Summary of the Invention

[0004] The object of the present invention is to provide a method and system for stable real-time data transmission over ultra-far-shore distances to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides a method for stable real-time data transmission over an ultra-far-from-shore distance, the method comprising: Acquire target transmission link data and extract dynamic fluctuation characteristics from the transmission link data, wherein the transmission link data includes signal strength distribution information and delay jitter relationship; Calling the trained channel estimation network to perform signal positioning processing on the transmission link data to generate a signal transmission path set, wherein the signal transmission path set includes a mapping relationship between a channel type identifier and a phase code; Based on the signal transmission path set, calling a waveform analysis network to perform multimodal feature fusion on the signal strength distribution information to generate a feature fusion result including time-frequency domain features and modulation mode features; The feature fusion result is input into a transmission strategy model for dynamic path selection processing, and a real-time data transmission element is output. The real-time data transmission element includes a control instruction and a phase self-calibration parameter generated based on a signal transmission sequence.

[0006] Preferably, the calling of the trained channel estimation network to perform signal positioning processing on the transmission link data to generate a signal transmission path set includes: Performing signal framing processing on the transmission link data to generate multiple signal frame areas and corresponding time-frequency window initial parameters; Extracting a spectrum feature map of each signal frame region, and performing energy attention weighting processing on the spectrum feature map to generate a weighted time domain feature vector; Calculating a signal-to-noise ratio between the time domain feature vector and a preset path delay to select candidate paths that meet a reliability threshold; Performing multipath interference suppression processing on the candidate paths to generate a set of signal transmission paths including the channel type identifier, each signal transmission path including a normalized phase code relative to the transmission link data; The method of performing multimodal feature fusion on the signal strength distribution information based on the signal transmission path set and calling a waveform analysis network to generate a feature fusion result including time-frequency domain features and modulation mode features includes: performing a region interception process on the transmission link data according to the normalized phase coding in the signal transmission path set to generate a plurality of signal segments; Filter and denoise each signal segment to generate optimized waveform analysis input data; Calling the bidirectional gated recurrent subnetwork in the waveform analysis network to extract sequence features from the waveform analysis input data to generate a time-frequency domain feature vector; In parallel, a modulation classification model is called to perform carrier feature matching on the signal segments to generate a modulation mode probability distribution; The time-frequency domain feature vector and the modulation mode probability distribution are tensor-concatenated to generate a multimodal feature fusion result.

[0007] Preferably, the step of inputting the feature fusion result into a transmission strategy model for dynamic path selection processing and outputting a real-time data transmission element includes: parsing the context dependency of the current transmission scenario and generating a transmission priority weight associated with the multimodal feature fusion result; generating a state transition matrix based on the transmission priority weights, and traversing and sorting the channel type identifiers in the signal transmission path set; Using a reinforcement learning strategy to select the optimal transmission sequence from the state transition matrix, generating a basic instruction set including sending, relaying, and retransmitting operations; The basic instruction set and the phase self-calibration parameters are integrated into a protocol structure to generate an instruction code segment that complies with a target transmission protocol.

[0008] Preferably, the method further comprises: Capture link response data in real time during data transmission and generate a feedback log containing signal phase offset and state change information; Extracting abnormal event features from the feedback log, and performing similarity matching between the abnormal event features and a historical transmission case library to generate adaptive adjustment instructions; Dynamically updating a path delay parameter of the channel estimation network based on the adaptive adjustment instruction; The updated path delay parameter is injected into the transmission strategy model, and the transmission priority weight in the state transfer matrix is ​​recalculated.

[0009] Preferably, extracting abnormal event features from the feedback log, performing similarity matching between the abnormal event features and a historical transmission case library, and generating adaptive adjustment instructions includes: Dividing the abnormal event characteristics into time windows to generate multiple event fragments and their corresponding signal screenshot sequences; The trained anomaly classification model is used to perform root cause analysis on each event segment, generating classification labels including signal recognition errors, phase drift, and transmission timeouts. Retrieving a solution template matching the classification label from the historical transmission case library to generate a candidate adjustment strategy set; Based on the matching degree ranking between the signal screenshot sequence and the candidate adjustment strategy set, the strategy with the highest confidence is selected to generate an adaptive adjustment instruction; The trained anomaly classification model is called to perform root cause analysis on each event segment, generating classification labels including signal recognition error, phase drift, and transmission timeout, including: Performing time series segmentation processing on the event fragments to generate signal screenshot sequence slices including the start trigger time and the end time; Performing key frame sampling processing on the signal screenshot sequence slices to generate a signal state change key frame set and a corresponding timestamp index; Extracting the signal phase offset between adjacent key frames in the signal state change key frame set to generate a signal displacement trajectory vector and a time interval sequence; Performing frequency offset calculation processing on the signal displacement trajectory vector to generate an abnormal fluctuation pattern feature vector; Inputting the abnormal fluctuation pattern feature vector into the spatiotemporal convolution subnet in the abnormal classification model to perform local feature extraction and generate a spatial abnormality activation map; Performing a sliding window mean filtering process on the time interval sequence to generate a smoothed response delay time series; Inputting the response delay time series into the gated recurrent subnet in the anomaly classification model for periodic pattern matching to generate a time dimension anomaly score; performing channel maximum pooling processing on the spatial anomaly activation map to generate a spatial dimension anomaly score; performing feature cross-fusion processing on the spatial dimension anomaly score and the time dimension anomaly score to generate a spatiotemporal joint anomaly probability distribution; Extracting peak detection results of signal recognition error probability, phase drift probability, and transmission timeout probability from the spatiotemporal joint anomaly probability distribution; performing dynamic threshold comparison processing on a preset probability threshold based on the peak detection results to generate a candidate set of classification labels including probability sorting; The classification label candidate set is subjected to timestamp index alignment verification processing to generate a target classification label that is consistent with the phase of the abnormal event in the signal screenshot sequence slice; the target classification label and the signal displacement trajectory vector are subjected to anomaly type reverse verification processing to generate a final classification label set including confidence weights.

[0010] Preferably, the method further comprises: Injecting a virtual phase disturbance parameter before data transmission, wherein the virtual phase disturbance parameter is used to simulate a random offset scenario of the signal phase; Monitoring the fault-tolerant processing result of the transmission strategy model on the post-disturbance phase and generating a stability evaluation index; When the stability evaluation index is lower than a preset threshold, triggering an online learning mode of the channel estimation network; Performing gradient backpropagation update on the spectral feature parameters of the channel estimation network based on the phase difference data before and after the disturbance; The monitoring of the fault-tolerant processing result of the transmission strategy model on the post-disturbance phase and generating a stability evaluation index includes: Count the number of positioning failures and abnormal capture success rate caused by phase errors during transmission; Calculating a ratio of the number of positioning failures to the total number of transmission steps to generate a first evaluation coefficient; Extracting the proportion of events that match the preset fault tolerance rule in the exception capture success rate to generate a second evaluation coefficient; A weighted sum is performed on the first evaluation coefficient and the second evaluation coefficient to generate a comprehensive stability evaluation index.

[0011] Preferably, performing gradient back propagation update on the spectral characteristic parameters of the channel estimation network based on the phase difference data before and after the disturbance includes: Performing timestamp alignment processing on the phase difference data before and after the disturbance to generate a matching pair sequence including a phase set before the disturbance and a phase set after the disturbance; Extracting the phase offset in the matching pair sequence to generate a horizontal offset vector and a vertical offset vector for each signal segment; Performing Euclidean distance calculation on the horizontal offset vector and the vertical offset vector to generate a phase difference vector set for each signal segment; Constructing a position regression loss function according to the phase difference vector set, wherein the position regression loss function includes a mean square error between the phase predicted by the channel estimation network and the actual phase after disturbance; Performing a differentiable transformation on the position regression loss function to generate a loss value tensor required for gradient backpropagation; Traversing the spectrum feature parameters of the channel estimation network, calculating the partial derivative of the loss value tensor with respect to each spectrum feature, and generating a spectrum feature gradient matrix; Performing momentum smoothing on the spectrum feature gradient matrix to generate an attenuated parameter update direction vector; Iteratively incrementally adjusting the spectral characteristic parameters of the channel estimation network based on the parameter update direction vector and a preset learning rate parameter; During the incremental adjustment process, the rate of change of the position regression loss function is monitored in real time to generate a loss convergence state indicator; When the loss convergence state indicator reaches a stable threshold, terminating the gradient back propagation update and freezing the spectral feature parameters; The updated spectrum characteristic parameters are injected into the forward propagation path of the channel estimation network, and the signal framing processing and signal positioning processing are re-executed to verify the results of the phase positioning accuracy improvement.

[0012] Preferably, the method further comprises: Building a cross-protocol transmission adapter, and parsing the instruction syntax differences of different transmission protocols through the cross-protocol transmission adapter; Converting the real-time data transmission elements into underlying driving instructions supported by the target protocol; Preserving the timing dependencies and exception handling context of the control instructions during the conversion process; Inject performance optimization parameters matched by the target protocol to generate transmission instruction bytecodes that meet cross-link execution conditions.

[0013] Preferably, the step of constructing a cross-protocol transmission adapter and parsing the differences in instruction syntax of different transmission protocols through the cross-protocol transmission adapter includes: Establish a transmission protocol grammar rule base to store keyword mapping tables and parameter transfer paths for each protocol; Performing abstract syntax tree analysis on the real-time data transmission element to generate intermediate presentation layer data; Traversing and querying the keyword mapping table based on the intermediate representation layer data to generate a protocol-compatible instruction replacement solution; Perform dependency injection on conflicting parameter transfer paths to generate unambiguous instruction conversion results; The dependency injection processing is performed on the conflicting parameter transfer paths to generate an unambiguous instruction conversion result, including: Identifying conflicting nodes with type ambiguity or scope overlap in the parameter transfer path, and generating a conflicting path identifier set; Extracting context metadata of each node in the conflict path identifier set, wherein the context metadata includes parameter type declaration, life cycle tag and call stack fingerprint; Generate an interface proxy class name corresponding to the conflicting node according to the context metadata, wherein the interface proxy class name inherits the native interface of the target protocol and injects a type forced conversion constraint; Parsing the dependency chain of the interface proxy class name in the target protocol to generate a dynamic binding strategy including parameter scope isolation boundaries; Mapping conflicting nodes in the parameter transfer path to protocol-compatible interface method instances based on the dynamic binding strategy to generate a dynamic binding queue of parameter instances; Traversing the interface method instances in the dynamic binding queue, verifying the compatibility status of the interface method instances with the target protocol instruction syntax, and generating a parameter binding validity status code; Filtering out interface method instances without type conflicts according to the parameter binding validity status code, and generating a protocol-independent intermediate instruction sequence; The intermediate instruction sequence is filled with an instruction template according to the grammatical rules of the target protocol to generate an unambiguous instruction conversion result.

[0014] Preferably, the present invention also includes a real-time data stable transmission system at ultra-far-shore distances, which includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above-mentioned real-time data stable transmission method at ultra-far-shore distances.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This method for stable, real-time data transmission over ultra-long-distance to shore transmission demonstrates multiple advantages in complex scenarios involving ultra-long-distance to shore data transmission. First, by acquiring target transmission link data and extracting dynamic fluctuation characteristics, it accurately captures signal variations during transmission, including signal strength distribution and delay jitter relationships. This enables a more comprehensive understanding of the transmission link, moving beyond a single signal parameter and providing rich foundational information for subsequent processing. The trained channel estimation network is called to generate a set of signal transmission paths, including the mapping relationship between channel type identification and phase coding. This process can clearly distinguish different types of channels and clarify the correspondence between phase coding and channel type. It avoids the transmission path selection errors caused by the vague judgment of channel type in traditional methods, makes the determination of signal transmission paths more targeted, and can select the appropriate transmission path according to different channel characteristics. Based on the set of signal transmission paths, a waveform analysis network is used to perform multimodal feature fusion on the signal strength distribution information, generating a fusion result that includes both time-frequency domain features and modulation pattern features. This provides a deeper understanding of the signal. This fusion of time-frequency domain features and modulation pattern features overcomes the limitations of single feature analysis and comprehensively reflects the signal's characteristics in different dimensions, enabling a more accurate understanding of the signal's transmission status and providing a more comprehensive basis for subsequent transmission strategy formulation. The feature fusion results are fed into the transmission strategy model for dynamic path selection, outputting real-time data transmission elements, including control instructions and phase self-calibration parameters generated based on the signal transmission sequence. This process enables dynamic adjustment of the transmission strategy. The control instructions enable instantaneous regulation of the transmission process based on real-time signal conditions, while the phase self-calibration parameters automatically correct for phase deviations during transmission, reducing signal interference and data errors caused by phase offsets. This allows data transmission to adapt to dynamic channel changes and maintain stability in complex ultra-offshore environments. Compared to traditional transmission methods, this approach forms a complete dynamic processing flow, from link data acquisition and analysis to path selection and transmission strategy formulation. Each link is tightly integrated, enabling collaborative responses to various interference factors at ultra-farshore distances. Its inherent processing mechanism effectively addresses the influence of different channels, such as sea surface reflections and atmospheric scattering, as well as fluctuations in signal strength and changes in latency jitter. This improves the stability and reliability of real-time data transmission at ultra-farshore distances, meeting the high data transmission requirements of fields such as marine development and ocean shipping. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a working principle diagram of the method for stable real-time data transmission over ultra-far-shore distances according to the present invention; Figure 2 Flowchart of signal localization and feature fusion; Figure 3 Flowcharts for dynamic path selection and instruction generation; Figure 4 Flowchart of link adaptive adjustment. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] See also Figure 1-Figure 4The present invention provides a method for stable real-time data transmission at an ultra-far-shore distance, the method comprising:

[0019] The target transmission link data is acquired and dynamic fluctuation characteristics are extracted from this transmission link data. This transmission link data includes signal strength distribution information and delay jitter relationships. Transmission link data can be collected through a sensor network deployed between offshore equipment and shore-based terminals. Signal strength distribution information reflects signal attenuation at different transmission nodes, while delay jitter relationships reflect the changing pattern of time delay during data transmission. Dynamic fluctuation characteristics are extracted based on these two factors, including the frequency of instantaneous peak and valley changes in signal strength and the amplitude range of delay jitter.

[0020] The trained channel estimation network is invoked to perform signal location processing on the transmission link data, generating a set of signal transmission paths containing a mapping between channel type identifiers and phase codes. The channel estimation network utilizes a deep neural network architecture and, after training with extensive historical transmission data, possesses signal location capabilities. By analyzing the transmission link data, it identifies possible signal transmission paths and assigns each path a corresponding channel type identifier (e.g., wireless channel, satellite channel, etc.) and phase code. The phase code is used to distinguish signal characteristics across different paths.

[0021] Based on the set of signal transmission paths, a waveform parsing network is invoked to perform multimodal feature fusion on the signal strength distribution information, generating a feature fusion result that includes both time-frequency domain features and modulation mode features. The waveform parsing network is composed of multiple subnetworks, capable of extracting the signal's time-frequency domain features and modulation mode features, respectively. Time-frequency domain features reflect the signal's distribution characteristics in the time and frequency domains, while modulation mode features reflect the signal's modulation method (such as amplitude modulation, frequency modulation, and phase modulation). Multimodal feature fusion integrates these two types of features to form a more comprehensive signal description.

[0022] The feature fusion results are input into the transmission strategy model for dynamic path selection, which then outputs a real-time data transmission element containing control instructions and phase self-calibration parameters generated based on the signal transmission sequence. Based on the feature fusion results and the actual transmission environment, the transmission strategy model selects the optimal signal transmission path and generates corresponding control instructions to guide operations such as data transmission and reception. The phase self-calibration parameters are used to compensate for phase offsets during signal transmission, ensuring data transmission accuracy.

[0023] In Example 1, when a trained channel estimation network is called to perform signal location processing on transmission link data to generate a set of signal transmission paths, the transmission link data is first subjected to signal framing. The transmission link data is a continuous data stream, which is divided into multiple independent signal frame regions by setting a fixed time interval or data length as the framing standard. Each signal frame region corresponds to a set of initial time-frequency window parameters, including the start and end times of the time window and the coverage range of the frequency window. These parameters are determined based on the carrier frequency and bandwidth characteristics of the transmitted signal.

[0024] The spectral feature map of each signal frame region is extracted. By performing a fast Fourier transform on the signal frame region, the time domain signal is converted to a frequency domain representation, resulting in a spectral feature map with frequency as the horizontal axis and energy as the vertical axis. Energy attention weighting is performed on the spectral feature map, and weight coefficients are calculated based on the energy values ​​of different frequency components. Frequency components with higher energy values ​​are assigned greater weights. The weight coefficients are then multiplied by the corresponding frequency components in the spectral feature map, and then converted back to the time domain via an inverse Fourier transform to generate a weighted time domain feature vector.

[0025] The signal-to-noise ratio (SNR) is calculated by comparing the time-domain feature vector with the preset path delay. The path delay is a pre-estimated range of possible delays based on transmission distance and signal propagation speed. The SNR is calculated by comparing the signal energy to the noise energy in the time-domain feature vector. A reliability threshold is set to select candidate paths with SNRs above this threshold. These candidate paths are considered to have basic transmission reliability.

[0026] Multipath interference suppression is performed on the candidate paths, and an adaptive equalization algorithm is used to eliminate mutual interference between signals on different paths, resulting in multiple independent signal transmission paths. A corresponding channel type identifier is added to each path, including information such as the channel's transmission medium and frequency band. A normalized phase code relative to the transmission link data is calculated for each path. Normalization is achieved by mapping the phase value to a range of 0 to 2π, ultimately forming a set of signal transmission paths that contains the mapping between channel type identifiers and phase codes.

[0027] When calling the waveform analysis network based on the signal transmission path set to perform multimodal feature fusion on the signal strength distribution information, the time position and frequency range of each path signal in the transmission link data are determined according to the normalized phase coding in the signal transmission path set. Based on this, the transmission link data is regionally intercepted to obtain multiple signal segments corresponding to different paths.

[0028] Each signal segment is filtered and denoised. Filtering uses a bandpass filter whose passband is set according to the signal's carrier frequency and bandwidth, allowing only signals within this range to pass. Denoising uses a wavelet threshold denoising algorithm, which performs a multi-scale decomposition of the signal segment. High-frequency coefficients below the threshold are set to zero, and then reconstructed to obtain the optimized waveform parsing input data.

[0029] The bidirectional gated recurrent subnetwork in the waveform parsing network is called to extract sequence features from the waveform parsing input data. The bidirectional gated recurrent subnetwork contains two gated recurrent units: the forward unit processes data from the beginning of the signal segment backward, and the backward unit processes data from the end of the signal segment forward. The outputs of these two units are concatenated to form a time-frequency domain feature vector, which contains information about the dynamic changes of the signal in the time and frequency dimensions.

[0030] The modulation classification model is invoked in parallel to match the signal segments' carrier characteristics. The modulation classification model stores carrier feature templates for various common modulation modes (such as amplitude keying, frequency keying, and phase shift keying). By calculating the similarity between the signal segment's carrier frequency, phase variation patterns, and other characteristics and each template, a modulation mode probability distribution is generated. Each value in this distribution corresponds to the probability that the signal segment belongs to a particular modulation mode.

[0031] The time-frequency domain feature vector and the modulation pattern probability distribution are tensor-concatenated. The time-frequency domain feature vector is a three-dimensional tensor whose dimensions correspond to the number of time steps, feature dimensions, and channels of the signal segment, respectively. The modulation pattern probability distribution is a one-dimensional vector whose length equals the number of modulation patterns. By concatenating the two along the feature dimension, they are combined into a new three-dimensional tensor, the multimodal feature fusion result, which contains both the time-frequency domain features and the modulation pattern characteristics of the signal.

[0032] Throughout the entire process, the parameters of the channel estimation network and waveform analysis network are trained based on a large amount of historical transmission data. The channel estimation network's training data covers transmission links at varying offshore distances and in varying weather conditions. By continuously adjusting the network weights, the deviation between the generated signal transmission path set and the actual path is gradually reduced. The waveform analysis network is trained using signal segments labeled with time-frequency domain features and modulation patterns. A backpropagation algorithm is used to optimize network parameters and improve the accuracy of the multimodal feature fusion results.

[0033] Furthermore, during signal framing, if the tail portion of the transmission link data is less than a full signal frame length, it is padded with zeros to ensure that all signal frame regions are of the same length. In multipath interference mitigation, for candidate paths with significant overlap, the signal strength and stability are compared, retaining the more optimal path and eliminating the path with greater interference. In the calculation of the modulation mode probability distribution, when the maximum probability value falls below the set confidence level, the corresponding modulation mode is marked as unknown for further processing.

[0034] In Example 2, the feature fusion results are input into the transmission strategy model for dynamic path selection processing. When outputting real-time data transmission elements, the context dependency of the current transmission scenario is first parsed. The context information of the transmission scenario includes the geographical location of the offshore equipment, real-time meteorological data (such as wind speed, precipitation intensity), the distribution of electromagnetic interference sources, and link stability data in past transmission records. By performing structured processing on this information, an association mapping with the multimodal feature fusion results is established. For example, when strong electromagnetic interference is detected, the transmission priority weight corresponding to the channel type with strong anti-interference ability is increased; when the offshore distance exceeds a preset threshold, the proportion of satellite channels in the priority weight is increased. The transmission priority weight is expressed in the form of a vector, and each element corresponds to the priority value of a signal transmission path.

[0035] A state transition matrix is ​​generated based on transmission priority weights. The rows and columns of the state transition matrix correspond to different signal transmission paths. Each element in the matrix represents the probability of switching from one path to another. This probability is calculated based on the transmission priority weights; greater priority differences correspond to smaller transition probabilities. Simultaneously, the channel type identifiers in the set of signal transmission paths are traversed and sorted based on the channel's transmission rate, historical bit error rate, and current load. The sorting results are used to determine the initial traversal order for path selection.

[0036] A reinforcement learning strategy is used to select the optimal transmission sequence from the state transition matrix. The agent in this reinforcement learning strategy aims to maximize the cumulative transmission benefit, which takes into account factors such as transmission delay, data integrity, and energy consumption. The agent continuously updates its assessment of the transmission performance of each path through interaction with the transmission environment. When the transmission benefit of a particular path falls below expectations, a path switching mechanism is triggered. Through multiple iterative learning cycles, a basic instruction set encompassing send, relay, and retransmit operations is generated. Send instructions specify the starting time and power for data transmission; relay instructions specify the sequence of nodes involved in signal forwarding and the forwarding time for each node; and retransmit instructions include the retransmission trigger condition (e.g., failure of the receiver to receive an acknowledgment signal) and the maximum number of retransmissions.

[0037] The basic instruction set and phase self-calibration parameters are integrated into the protocol structure. These parameters include the phase compensation value, calibration period, and calibration trigger threshold, which are determined based on the phase characteristics of the signal transmission path. The integration process encapsulates the basic instruction set and phase self-calibration parameters into a protocol data unit (PDU) according to the frame structure requirements of the target transmission protocol (such as TCP / IP or satellite communication protocols). The basic instruction set is used as the data field, and the phase self-calibration parameters are used as the header option field. This encapsulation generates an instruction code segment that complies with the target transmission protocol and can be directly parsed and executed by the transmitting device.

[0038] During data transmission, monitoring modules deployed on transmission nodes capture link response data in real time. This data includes signal strength, phase deviation, data verification results, and transmission status codes reported by the receiving end. Feedback logs are generated based on this data and stored in a time-series database. Each record includes a timestamp, signal phase deviation (in radians), and status change information (e.g., from "normal transmission" to "transmission interrupted").

[0039] Extract abnormal event features from feedback logs. Abnormal events include sudden drops in signal strength, phase offsets exceeding the allowable range, and transmission timeouts. Features include the time of occurrence, duration, impact range, and associated path identifiers. These features are then matched against a historical transmission case library, which stores features of past abnormal events and corresponding solutions. The matching process uses a cosine similarity algorithm to calculate the similarity between the current abnormal event features and feature vectors in the case library. The top N cases with the highest similarity are selected as references.

[0040] Based on the matching results, adaptive adjustment instructions are generated. These include corrections to channel estimation network parameters and recommendations for adjusting the transmission strategy. For example, when a "phase drift" case is matched, the adjustment instructions may include increasing the phase calibration frequency. Based on these adaptive adjustment instructions, the path delay parameters of the channel estimation network are dynamically updated. This update process is implemented through an online parameter adjustment interface without interrupting current data transmission.

[0041] The updated path delay parameters are injected into the transmission policy model, and the transmission priority weights in the state transition matrix are recalculated. During the recalculation, paths with significantly improved performance after the parameter updates are prioritized, enabling the transmission policy model to quickly adapt to changes in link conditions. This entire dynamic adjustment process forms a closed loop, ensuring stable data transmission despite fluctuations in the link environment.

[0042] Furthermore, a sliding time window technique is used to extract features from abnormal events in feedback logs. The window size is set based on the average duration of the abnormal event. Statistical analysis of the data within the window is performed to filter out false anomalies caused by transient interference. An exploration factor is introduced during the training of reinforcement learning strategies, allowing the agent to explore underutilized paths with a certain probability while ensuring transmission stability, in order to discover more optimal transmission sequences.

[0043] In Example 3, the abnormal event features in the feedback log are extracted, and the abnormal event features are matched with the historical transmission case library for similarity. When generating adaptive adjustment instructions, the abnormal event features are first divided into time windows. The length of the time window is set according to the average duration of the abnormal event, and usually covers the complete cycle from the occurrence of the abnormality to the restoration of normality. By sliding the time window, the continuous abnormal event features are divided into multiple non-overlapping event segments. Each event segment corresponds to a set of continuous signal screenshot sequences. The signal screenshot sequence consists of signal waveform diagrams, spectrum diagrams and phase distribution diagrams at different times, which can intuitively reflect the development process of the abnormal event.

[0044] The trained anomaly classification model is used to perform root cause analysis on each event segment. This model utilizes a deep neural network architecture. The input layer receives the feature vector of the event segment. The hidden layer extracts features through multiple layers of convolution and pooling. The output layer outputs the probabilities of classification labels, including signal recognition errors, phase drift, and transmission timeouts. Classification labels are determined based on the output layer's probability values. When a label has a higher probability than other labels, it is considered the primary anomaly type for that event segment.

[0045] Retrieve solution templates that match the classification labels from the historical transmission case library. The historical transmission case library is stored by anomaly type, and each category contains multiple solution templates. The template content includes parameter adjustment suggestions, operation steps, and applicable scenario descriptions. Based on the degree of matching of the classification labels, select the templates with the highest relevance to form a set of candidate adjustment strategies. Based on the matching degree of the signal screenshot sequence and the candidate adjustment strategy set, calculate the overlap between the anomaly features in the signal screenshot sequence and the preset features of each template. The higher the overlap, the higher the confidence of the corresponding candidate strategy. The strategy with the highest confidence is selected to generate the adaptive adjustment instruction.

[0046] The trained anomaly classification model is used to perform root cause analysis on each event segment. To generate classification labels, the event segments are first segmented into time series. Based on the inflection points of the abnormal event's characteristic changes (such as a sudden drop in signal strength), the event segments are divided into multiple signal screenshot sequence slices. Each slice is labeled with the start and end times, ensuring that each slice encompasses a complete phase of the abnormal characteristic change.

[0047] Keyframe sampling is performed on the signal screenshot sequence slices. Keyframes are selected based on the rate of change of signal characteristics. When the rate of change exceeds a set threshold, the signal screenshot at the current moment is recorded as a keyframe. After sampling, a set of signal state change keyframes and corresponding timestamp indexes are generated. The timestamp index is accurate to the millisecond level and is used to mark the time position of the keyframe within the entire event segment.

[0048] Extract the signal phase offset between adjacent keyframes in a set of signal state change keyframes. By comparing the phase distribution diagrams of adjacent keyframes, calculate the phase difference at each frequency point, and integrate these differences into a signal displacement trajectory vector with the same dimension as the number of signal frequency points. Simultaneously, calculate the timestamp difference between adjacent keyframes to generate a time interval sequence that reflects the time interval changes between keyframes.

[0049] The frequency offset of the signal displacement trajectory vector is calculated and processed. The displacement trajectory vector is converted to the frequency domain through Fourier transform. The offset amplitude and phase changes of each frequency component are analyzed to generate the abnormal fluctuation pattern feature vector. This vector contains information such as the main frequency, harmonic components, and attenuation coefficient of the abnormal fluctuation.

[0050] The abnormal fluctuation pattern feature vector is input into the spatiotemporal convolutional subnetwork in the anomaly classification model for local feature extraction. The spatiotemporal convolutional subnetwork consists of a temporal convolutional layer and a spatial convolutional layer. The temporal convolutional layer captures how features change over time, while the spatial convolutional layer extracts correlation features between different frequency points. The outputs of both layers are processed by an activation function to generate a spatial anomaly activation map. Regions with higher values ​​in the activation map correspond to frequency ranges with significant abnormal fluctuations.

[0051] Apply sliding window mean filtering to the time interval series. The size of the sliding window is set based on the average fluctuation period of the time interval. By calculating the average value of the time interval within the window, high-frequency noise in the series is smoothed, generating a smoothed response delay time series that more clearly reflects the changing trend of the time interval of abnormal events.

[0052] The response delay time series is fed into the gated recurrent subnetwork in the anomaly classification model for periodic pattern matching. The gated recurrent subnetwork controls the flow of information through a gating mechanism, effectively capturing long-term dependencies in the time series. By comparing the results with the preset anomaly periodic pattern, it outputs a time-dimensional anomaly score. A higher score indicates a more significant anomaly periodicity within that time period.

[0053] The spatial anomaly activation map is processed with channel maximum pooling. The maximum value is selected on the feature map of each channel, the spatial dimension is compressed, and the spatial dimension anomaly score is generated. The score reflects the strength of the anomaly features corresponding to different channels.

[0054] The spatial dimension anomaly score and the temporal dimension anomaly score are cross-fused using the following formula:

[0055] in, represents the comprehensive score of the spatiotemporal joint anomaly probability distribution, represents the spatial dimension anomaly score, represents the abnormal score of the time dimension, It is a weight coefficient ranging from 0 to 1, and is set according to the importance of spatial and temporal features in anomaly recognition.

[0056] The spatiotemporal joint anomaly probability distribution is generated based on the comprehensive score, and each element in the distribution corresponds to the anomaly probability within a certain frequency range at a certain moment.

[0057] Peak detection results for signal recognition error probability, phase drift probability, and transmission timeout probability are extracted from the spatiotemporal joint anomaly probability distribution. Peak detection results are determined by traversing the probability distribution to find the maximum probability and location corresponding to each anomaly type.

[0058] Based on the peak detection results, dynamic threshold comparison is performed against the preset probability threshold. The preset probability threshold is set for each anomaly type. When the peak probability of a certain anomaly type exceeds its corresponding threshold, the type is included in the classification label candidate set and sorted from high to low by probability value, generating a classification label candidate set containing the probability ranking.

[0059] Perform timestamp index alignment verification on the candidate set of classification labels to check whether the timestamps of the candidate labels are consistent with the actual occurrence time of the abnormal event in the signal screenshot sequence slices. Remove labels with mismatched times and generate target classification labels that are consistent with the phase of the abnormal event.

[0060] The target classification label and the signal displacement trajectory vector are reverse-verified for anomaly type. The anomaly type indicated by the target classification label is compared with the characteristics reflected by the signal displacement trajectory vector. For example, the trajectory vector corresponding to phase drift should show a continuous phase shift trend. Based on the verification results, a confidence weight is assigned to each label, and a final classification label set containing the confidence weights is generated.

[0061] In Example 4, virtual phase perturbation parameters are injected before data transmission. These parameters are random values ​​generated based on the phase offset range commonly encountered in ultra-far-shore transmission scenarios. For example, in an offshore wind power data transmission scenario 1000 kilometers offshore, considering that factors such as atmospheric refraction and wave reflection may cause the signal phase to shift by ±0.5π, virtual phase perturbation parameters are randomly generated within this range. By superimposing these parameters on the phase data of the original signal, various phase offsets that may be encountered in actual transmission are simulated.

[0062] Monitor the transmission strategy model's fault-tolerant processing results for the post-disturbance phase and generate stability evaluation indicators. Specifically, after injecting virtual phase disturbances, the transmission strategy model will process the disturbed signal according to the preset algorithm, including phase compensation, path adjustment and other operations. In this process, the positioning failures caused by phase errors are recorded. For example, the signal should have been transmitted to the shore-based terminal A, but was mistakenly located to terminal B due to the phase error. Such situations are counted as the number of positioning failures. At the same time, the number of successful anomaly captures is counted, that is, the number of times the transmission strategy model successfully identifies the phase disturbance and initiates the corresponding processing mechanism. The ratio of the two constitutes the basic data for the anomaly capture success rate.

[0063] After counting the number of positioning failures and the success rate of abnormal capture due to phase errors during the transmission process, the ratio of the number of positioning failures to the total number of transmission steps is calculated to generate the first evaluation coefficient. The total number of transmission steps refers to the operation steps included in the complete process from data transmission to reception confirmation. For example, a transmission process including four steps of signal encoding, transmission, relaying, and reception decoding has a total number of transmission steps of 4. The proportion of events in the abnormal capture success rate that match the preset fault tolerance rules is extracted to generate the second evaluation coefficient. The preset fault tolerance rules may stipulate that when the phase offset is within ±0.2π, the transmission strategy model should complete the abnormal capture within 100ms. The proportion of events that meet this rule in all abnormal events is the basis for calculating the second evaluation coefficient.

[0064] The first and second evaluation coefficients are weighted and summed to generate a comprehensive stability evaluation index. Weighting is determined based on actual transmission requirements. If the transmission scenario requires high positioning accuracy, the first evaluation coefficient is given a higher weight; if timely handling of anomalies is more important, the second evaluation coefficient is given a higher weight. When the comprehensive stability evaluation index falls below a preset threshold, the channel estimation network's online learning mode is triggered, allowing the network to adjust its parameters based on new disturbance data.

[0065] When performing gradient backpropagation to update the spectral characteristic parameters of the channel estimation network based on the phase difference data before and after the perturbation, the phase difference data before and after the perturbation is first timestamp aligned. For example, the phase data at 10:00:00.000 before the perturbation is matched with the phase data at the same time point after the perturbation to generate a sequence of matching pairs containing the phase set before the perturbation and the phase set after the perturbation, ensuring that each pair of data corresponds one-to-one in time.

[0066] The phase offsets in the matching pair sequence are extracted to generate horizontal and vertical offset vectors for each signal segment. The horizontal offset vector reflects the phase offset on the frequency axis, while the vertical offset vector reflects the phase offset on the time axis. For example, if a signal segment has a phase offset of 0.1π at a frequency of 1 GHz and a phase offset of 0.05π between time t1 and t2, the corresponding horizontal and vertical offset vectors will record these specific values.

[0067] The Euclidean distance between the horizontal and vertical offset vectors is calculated. This Euclidean distance quantifies the phase offset of each signal segment, generating a set of phase difference vectors for each signal segment. Each vector in this set corresponds to a signal segment, and the magnitude of the vector represents the total phase offset for that segment.

[0068] A position regression loss function is constructed based on the set of phase difference vectors. This function takes the phase data predicted by the channel estimation network and the actual measured phase data after perturbation as input and calculates the mean squared error between the two. The smaller the error, the more accurate the network prediction. A differentiable transformation is performed on the position regression loss function to convert it into a form suitable for gradient calculation. This generates the loss value tensor required for gradient backpropagation. The dimension of this tensor matches the dimension of the network output layer.

[0069] The spectral feature parameters of the channel estimation network, including convolution kernel weights and bias terms, are traversed. The partial derivatives of the loss tensor with respect to each spectral feature are calculated to generate a spectral feature gradient matrix. Each element in the gradient matrix represents the degree of influence of the corresponding parameter on the loss value. A positive value indicates an increase in the loss value when the parameter increases, while a negative value indicates a decrease in the loss value when the parameter increases.

[0070] Momentum smoothing is performed on the spectral feature gradient matrix, introducing a momentum factor. This allows the current parameter update direction to not only consider the current gradient but also retain part of the previous update direction, reducing gradient fluctuations and generating an attenuated parameter update direction vector. For example, if the previous update direction vector is v1, the direction vector obtained by the current gradient calculation is v2, and the momentum factor is 0.9, the new update direction vector might be 0.9v1 + 0.1v2.

[0071] Based on the parameter update direction vector and the preset learning rate parameter, the spectral characteristic parameters of the channel estimation network are iteratively and incrementally adjusted. The learning rate parameter determines the magnitude of each parameter adjustment. For example, when the learning rate is 0.001, the parameter adjustment amount is the product of the update direction vector and 0.001. During the incremental adjustment process, the rate of change of the position regression loss function is monitored in real time. That is, the ratio of the difference between the loss values ​​of two consecutive iterations to the difference in the number of iterations is monitored. When the absolute value of this rate of change is less than the preset stability threshold, a loss convergence status indicator is generated.

[0072] When the loss convergence indicator reaches a stable threshold, the gradient backpropagation update is terminated and the spectral characteristic parameters are frozen. These parameters are then considered to be the current optimal values. The updated spectral characteristic parameters are then injected into the forward propagation path of the channel estimation network, and the signal framing and positioning processes are re-executed. By comparing the positioning results of the network for the same signal before and after the update, the change in phase positioning accuracy is verified. This entire process is completed automatically by the system without the need for human intervention.

[0073] In Example 5, when building a cross-protocol transmission adapter and parsing the instruction syntax differences of different transmission protocols through the adapter, a transmission protocol syntax rule base is first established. The rule base adopts a distributed database storage structure and divides sub-libraries according to protocol type. Each sub-library contains a keyword mapping table and parameter transfer path for the corresponding protocol. The keyword mapping table records the correspondence between keywords with the same or similar functions in different protocols. For example, the keyword "SOF" that represents "start of data frame" in the satellite communication protocol may be represented by "StartFrame" in the wireless communication protocol. The mapping table will store these two keywords in association. The parameter transfer path records the complete flow process of parameters from generation to execution in the form of a flowchart, including the encapsulation format and parsing order of parameters at each layer of the protocol stack.

[0074] Perform abstract syntax tree parsing on real-time data transmission elements. Real-time data transmission elements contain control instructions and phase self-calibration parameters. The parsing process converts these elements into an abstract syntax tree through a syntax analyzer. The nodes of the tree represent grammatical units (such as instruction type, parameter name, and value), and the connections between nodes represent grammatical relationships (such as inclusion and subordination). After the abstract syntax tree is generated, the core semantic information is extracted and converted into intermediate representation layer data. The intermediate representation layer data uses a protocol-independent structured format, such as JSON format, retaining only core information such as the instruction operation type, parameter value, and execution order, stripping away the grammatical details of the specific protocol.

[0075] The keyword mapping table is traversed and queried based on the intermediate presentation layer data. Based on the operation type and parameter name in the intermediate presentation layer data, the keyword mapping table is searched for the keyword corresponding to the target protocol, and a protocol-compatible instruction replacement solution is generated. For example, the "retransmission count" parameter in the intermediate presentation layer data must be represented as "RetransCount" in the target protocol. The replacement solution records this mapping relationship and ensures that the parameter value range meets the requirements of the target protocol.

[0076] Dependency injection is performed on conflicting parameter paths. When type ambiguity or scope overlap occurs at a certain point in the parameter paths of different protocols, the conflicting nodes are first identified and a set of conflict path identifiers is generated. A conflict path identifier consists of the protocol type, parameter name, and conflict location, for example, "satellite protocol - power parameter - transport layer encapsulation."

[0077] Extract contextual metadata for each node in the conflict path identifier set. This contextual metadata includes parameter type declarations (e.g., integer, floating-point), lifetime markers (e.g., temporary variables, global variables), and call stack fingerprints (recording the sequence of functions called with the parameter). Based on this metadata, generate the interface proxy class name corresponding to the conflicting node. The interface proxy class name uses the "protocol name + parameter name + proxy identifier" naming convention, such as "SatellitePowerProxy." This interface proxy class inherits the native interface of the target protocol and injects type coercion constraints into the class definition, for example, forcing floating-point parameters to the integer type required by the target protocol.

[0078] Parse the dependency chain of the interface proxy class name within the target protocol. By analyzing the target protocol's class inheritance and function call order, determine the interface proxy class's position within the protocol stack and generate a dynamic binding strategy that includes parameter scope isolation boundaries. Scope isolation boundaries are implemented through namespace partitioning. For example, assigning a separate namespace, "ProxySpace," to the proxy class prevents naming conflicts with the target protocol's native classes.

[0079] Based on a dynamic binding strategy, conflicting nodes in the parameter passing path are mapped to protocol-compatible interface method instances. The generation of interface method instances must adhere to the target protocol's function definition specifications, including the number of parameters, return type, and exception handling mechanism. Generated instances are arranged in execution order, forming a dynamic binding queue for parameter instances.

[0080] The interface method instances in the dynamic binding queue are traversed to verify their compatibility with the target protocol's instruction syntax. Verification includes whether the method name conforms to the target protocol's naming conventions, whether the parameter types are consistent with the protocol requirements, and whether the calling method matches the protocol's function call rules. After verification, a parameter binding validity status code is generated. Status code "00" indicates full compatibility, "01" indicates minor adjustments are required (such as parameter reordering), and "10" indicates a serious conflict.

[0081] Based on the parameter binding validity status code, interface method instances without type conflicts are filtered out to generate a protocol-independent intermediate instruction sequence. The intermediate instruction sequence is arranged in execution order and only contains the method call and parameter passing logic, without involving the specific protocol syntax format.

[0082] The intermediate instruction sequence is filled into the instruction template according to the grammatical rules of the target protocol. The target protocol's instruction template predefines the instruction structure (such as header, data segment, check digit) and grammatical format (such as separators and keyword case). The filling process embeds the contents of the intermediate instruction sequence into the corresponding positions in the template, generating an unambiguous instruction conversion result.

[0083] When converting real-time data transmission elements into low-level driver instructions supported by the target protocol, the control instructions' timing dependencies and exception handling context are preserved. Timing dependencies are implemented through timestamps, with each instruction accompanied by the execution time, ensuring that instructions are executed in their original order. The exception handling context records retry mechanisms and fallback options in the event of instruction execution failures. For example, if a "send instruction" fails, the "switching channels and retransmitting" exception handling process is automatically triggered.

[0084] Inject performance optimization parameters that match the target protocol. These parameters include transmission rate adjustment, channel coding, and power control thresholds. These parameters are determined based on the characteristics of the target protocol and the transmission environment. For example, in a high-interference environment, the optimization parameter "reducing the frequency hopping interval to 50ms" can be injected into the wireless communication protocol. The resulting transmission instruction bytecode conforms to the binary format requirements of the target protocol and can be directly parsed and executed by the underlying driver of the transmission device.

[0085] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0086] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for stable real-time data transmission over an ultra-far-from-shore distance, characterized in that: include: Acquire target transmission link data and extract dynamic fluctuation characteristics from the transmission link data, wherein the transmission link data includes signal strength distribution information and delay jitter relationship; Calling the trained channel estimation network to perform signal positioning processing on the transmission link data to generate a signal transmission path set, wherein the signal transmission path set includes a mapping relationship between a channel type identifier and a phase code; Based on the signal transmission path set, calling a waveform analysis network to perform multimodal feature fusion on the signal strength distribution information to generate a feature fusion result including time-frequency domain features and modulation mode features; The feature fusion result is input into a transmission strategy model for dynamic path selection processing, and a real-time data transmission element is output. The real-time data transmission element includes a control instruction and a phase self-calibration parameter generated based on a signal transmission sequence.

2. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 1, wherein: The calling of the trained channel estimation network to perform signal positioning processing on the transmission link data to generate a signal transmission path set includes: Performing signal framing processing on the transmission link data to generate a plurality of signal frame regions and corresponding time-frequency window initial parameters; Extracting a spectrum feature map of each signal frame region, and performing energy attention weighting processing on the spectrum feature map to generate a weighted time domain feature vector; Calculating a signal-to-noise ratio between the time domain feature vector and a preset path delay to select candidate paths that meet a reliability threshold; Performing multipath interference suppression processing on the candidate paths to generate a set of signal transmission paths including the channel type identifier, each signal transmission path including a normalized phase code relative to the transmission link data; The method of performing multimodal feature fusion on the signal strength distribution information based on the signal transmission path set and calling a waveform analysis network to generate a feature fusion result including time-frequency domain features and modulation mode features includes: performing a region interception process on the transmission link data according to the normalized phase coding in the signal transmission path set to generate a plurality of signal segments; Filter and denoise each signal segment to generate optimized waveform analysis input data; Calling the bidirectional gated recurrent subnetwork in the waveform analysis network to extract sequence features from the waveform analysis input data to generate a time-frequency domain feature vector; In parallel, a modulation classification model is called to perform carrier feature matching on the signal segments to generate a modulation mode probability distribution; The time-frequency domain feature vector and the modulation mode probability distribution are tensor-concatenated to generate a multimodal feature fusion result.

3. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 2, wherein: Inputting the feature fusion result into a transmission strategy model for dynamic path selection processing and outputting a real-time data transmission element includes: parsing the context dependency of the current transmission scenario and generating a transmission priority weight associated with the multimodal feature fusion result; generating a state transition matrix based on the transmission priority weights, and traversing and sorting the channel type identifiers in the signal transmission path set; Using a reinforcement learning strategy to select the optimal transmission sequence from the state transition matrix, generating a basic instruction set including sending, relaying, and retransmitting operations; The basic instruction set and the phase self-calibration parameters are integrated into a protocol structure to generate an instruction code segment that complies with a target transmission protocol.

4. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 3, wherein: The method further comprises: Capture link response data in real time during data transmission and generate a feedback log containing signal phase offset and state change information; Extracting abnormal event features from the feedback log, and performing similarity matching between the abnormal event features and a historical transmission case library to generate adaptive adjustment instructions; Dynamically updating a path delay parameter of the channel estimation network based on the adaptive adjustment instruction; The updated path delay parameter is injected into the transmission strategy model, and the transmission priority weight in the state transfer matrix is ​​recalculated.

5. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 4, wherein: The extracting abnormal event features from the feedback log, performing similarity matching on the abnormal event features with a historical transmission case library, and generating an adaptive adjustment instruction includes: Dividing the abnormal event characteristics into time windows to generate multiple event fragments and their corresponding signal screenshot sequences; The trained anomaly classification model is used to perform root cause analysis on each event segment, generating classification labels including signal recognition errors, phase drift, and transmission timeouts. Retrieving a solution template matching the classification label from the historical transmission case library to generate a candidate adjustment strategy set; Based on the matching degree ranking between the signal screenshot sequence and the candidate adjustment strategy set, the strategy with the highest confidence is selected to generate an adaptive adjustment instruction; The trained anomaly classification model is called to perform root cause analysis on each event segment, generating classification labels including signal recognition error, phase drift, and transmission timeout, including: Performing time series segmentation processing on the event fragments to generate signal screenshot sequence slices including the start trigger time and the end time; Performing key frame sampling processing on the signal screenshot sequence slices to generate a signal state change key frame set and a corresponding timestamp index; Extracting the signal phase offset between adjacent key frames in the signal state change key frame set to generate a signal displacement trajectory vector and a time interval sequence; Performing frequency offset calculation processing on the signal displacement trajectory vector to generate an abnormal fluctuation pattern feature vector; Inputting the abnormal fluctuation pattern feature vector into the spatiotemporal convolution subnet in the abnormal classification model to perform local feature extraction and generate a spatial abnormality activation map; Performing a sliding window mean filtering process on the time interval sequence to generate a smoothed response delay time series; Inputting the response delay time series into the gated recurrent subnet in the anomaly classification model for periodic pattern matching to generate a time dimension anomaly score; performing channel maximum pooling processing on the spatial anomaly activation map to generate a spatial dimension anomaly score; performing feature cross-fusion processing on the spatial dimension anomaly score and the time dimension anomaly score to generate a spatiotemporal joint anomaly probability distribution; Extracting peak detection results of signal recognition error probability, phase drift probability, and transmission timeout probability from the spatiotemporal joint anomaly probability distribution; performing dynamic threshold comparison processing on a preset probability threshold based on the peak detection results to generate a candidate set of classification labels including probability sorting; The classification label candidate set is subjected to timestamp index alignment verification processing to generate a target classification label that is consistent with the phase of the abnormal event in the signal screenshot sequence slice; the target classification label and the signal displacement trajectory vector are subjected to anomaly type reverse verification processing to generate a final classification label set including confidence weights.

6. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 1, wherein: The method further comprises: Injecting a virtual phase disturbance parameter before data transmission, wherein the virtual phase disturbance parameter is used to simulate a random offset scenario of the signal phase; Monitoring the fault-tolerant processing result of the transmission strategy model on the post-disturbance phase and generating a stability evaluation index; When the stability evaluation index is lower than a preset threshold, triggering an online learning mode of the channel estimation network; Performing gradient backpropagation update on the spectral feature parameters of the channel estimation network based on the phase difference data before and after the disturbance; The monitoring of the fault-tolerant processing result of the transmission strategy model on the post-disturbance phase and generating a stability evaluation index includes: Count the number of positioning failures and abnormal capture success rate caused by phase errors during transmission; Calculating a ratio of the number of positioning failures to the total number of transmission steps to generate a first evaluation coefficient; Extracting the proportion of events that match the preset fault tolerance rule in the exception capture success rate to generate a second evaluation coefficient; A weighted sum is performed on the first evaluation coefficient and the second evaluation coefficient to generate a comprehensive stability evaluation index.

7. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 6, wherein: The updating of the spectral characteristic parameters of the channel estimation network by gradient back propagation based on the phase difference data before and after the disturbance includes: Performing timestamp alignment processing on the phase difference data before and after the disturbance to generate a matching pair sequence including a phase set before the disturbance and a phase set after the disturbance; Extracting the phase offset in the matching pair sequence to generate a horizontal offset vector and a vertical offset vector for each signal segment; Performing Euclidean distance calculation on the horizontal offset vector and the vertical offset vector to generate a phase difference vector set for each signal segment; Constructing a position regression loss function according to the phase difference vector set, wherein the position regression loss function includes a mean square error between the phase predicted by the channel estimation network and the actual phase after disturbance; Performing a differentiable transformation on the position regression loss function to generate a loss value tensor required for gradient backpropagation; Traversing the spectrum feature parameters of the channel estimation network, calculating the partial derivative of the loss value tensor with respect to each spectrum feature, and generating a spectrum feature gradient matrix; Performing momentum smoothing on the spectral feature gradient matrix to generate an attenuated parameter update direction vector; Iteratively incrementally adjusting the spectral characteristic parameters of the channel estimation network based on the parameter update direction vector and a preset learning rate parameter; During the incremental adjustment process, the rate of change of the position regression loss function is monitored in real time to generate a loss convergence state indicator; When the loss convergence state indicator reaches a stable threshold, terminating the gradient back propagation update and freezing the spectral feature parameters; The updated spectrum characteristic parameters are injected into the forward propagation path of the channel estimation network, and the signal framing processing and signal positioning processing are re-executed to verify the results of the phase positioning accuracy improvement.

8. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 1, wherein: The method further comprises: Building a cross-protocol transmission adapter, and parsing the instruction syntax differences of different transmission protocols through the cross-protocol transmission adapter; Converting the real-time data transmission elements into underlying driving instructions supported by the target protocol; Preserving the timing dependencies and exception handling context of the control instructions during the conversion process; Inject performance optimization parameters matched by the target protocol to generate transmission instruction bytecodes that meet cross-link execution conditions.

9. The method for stable real-time data transmission over an ultra-far-shore distance as claimed in claim 8, wherein: The step of constructing a cross-protocol transmission adapter and parsing instruction syntax differences between different transmission protocols through the cross-protocol transmission adapter includes: Establish a transmission protocol grammar rule base to store keyword mapping tables and parameter transfer paths for each protocol; Performing abstract syntax tree analysis on the real-time data transmission element to generate intermediate presentation layer data; Traversing and querying the keyword mapping table based on the intermediate representation layer data to generate a protocol-compatible instruction replacement solution; Perform dependency injection on conflicting parameter transfer paths to generate unambiguous instruction conversion results; The dependency injection processing is performed on the conflicting parameter transfer paths to generate an unambiguous instruction conversion result, including: Identifying conflicting nodes with type ambiguity or scope overlap in the parameter transfer path, and generating a conflicting path identifier set; Extracting context metadata of each node in the conflict path identifier set, wherein the context metadata includes parameter type declaration, life cycle tag and call stack fingerprint; Generate an interface proxy class name corresponding to the conflicting node according to the context metadata, wherein the interface proxy class name inherits the native interface of the target protocol and injects a type forced conversion constraint; Parsing the dependency chain of the interface proxy class name in the target protocol to generate a dynamic binding strategy including parameter scope isolation boundaries; Mapping conflicting nodes in the parameter transfer path to protocol-compatible interface method instances based on the dynamic binding strategy to generate a dynamic binding queue of parameter instances; Traversing the interface method instances in the dynamic binding queue, verifying the compatibility status of the interface method instances with the target protocol instruction syntax, and generating a parameter binding validity status code; Filtering out interface method instances without type conflicts according to the parameter binding validity status code, and generating a protocol-independent intermediate instruction sequence; The intermediate instruction sequence is filled with an instruction template according to the grammatical rules of the target protocol to generate an unambiguous instruction conversion result.

10. A real-time data stable transmission system with ultra-far-from-shore distance, characterized in that: It includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of a method for stable transmission of real-time data at ultra-far-shore distances as described in any one of claims 1 to 9.

Images