Method for predicting propagation of ultra-low frequency / very-low frequency electromagnetic waves in marine environment

Abstract

This application provides a method for predicting the propagation of ultra-low frequency (ULF) / extremely low frequency (ELF) electromagnetic waves in a marine environment. The method includes: acquiring initial propagation parameters of the target electromagnetic wave and initial marine environmental parameter information of the target marine area within a first target time period; determining target marine environmental parameter information of the target marine area within a second target time period based on the initial marine environmental parameter information; inputting the target marine environmental parameter information and the initial propagation parameters into a pre-trained electromagnetic wave prediction model that integrates physical law constraints and knowledge guidance mechanisms, to output a prediction result showing the propagation of the target electromagnetic wave within the target marine area, starting with the initial propagation parameters. This solution significantly improves the accuracy and robustness of electromagnetic wave propagation prediction in complex marine environments by integrating physical laws and knowledge guidance.

Description

Technical Field

[0001] This application relates to the field of electromagnetic wave propagation prediction technology, and in particular to a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment. Background Technology

[0002] In the field of marine electromagnetic wave propagation prediction technology, ultra-low frequency (SLF, 30–300 Hz) and extremely low frequency (ELF, 3–30 Hz) electromagnetic waves are of significant value in applications such as deep-sea communication, underwater sensor network deployment, marine resource exploration, and covert submarine communication due to their unique physical characteristics. These electromagnetic waves possess strong seawater penetration capabilities, enabling stable transmission at depths of hundreds to thousands of meters, and exhibit minimal attenuation in seawater, making them suitable for long-distance, large-scale underwater information transmission. Therefore, accurately predicting the propagation characteristics of ultra-low frequency / extreme low frequency electromagnetic waves in trans-domain marine environments encompassing the sea surface and underwater is a key technological support for improving the performance of underwater communication systems and the efficiency of marine exploration.

[0003] Traditional ocean electromagnetic wave propagation prediction primarily relies on model-driven methods. This involves establishing physical models based on classical electromagnetic theories such as Maxwell's equations, combined with ocean environmental parameters (e.g., conductivity, temperature, salinity, topography), and performing simulations. Specific methods include analytical methods (e.g., ray tracing models), numerical methods (e.g., finite difference time-domain method, finite element method), and hybrid algorithms. Analytical methods, centered on physical mechanisms, offer high computational efficiency but are limited in their applicability. Numerical methods can handle more complex boundary conditions and non-homogeneous media, offering higher accuracy, but typically require substantial computational resources, making real-time performance difficult. Furthermore, traditional model-driven methods often make idealized assumptions about the ocean environment (e.g., homogeneous media, flat seabed topography), failing to accurately depict the multi-scale, strongly nonlinear processes present in the actual ocean environment, such as thermohaline, internal waves, vortices, and seabed topographic undulations. This leads to discrepancies between the predicted results and the actual propagation behavior of electromagnetic waves. Data-driven approaches utilize field-measured channel data (such as channel impulse response, path loss, and delay spread) and employ machine learning (especially deep learning) methods to directly learn the mapping relationship between inputs (environment, frequency, distance, etc.) and outputs (channel characteristics). This method excels at capturing complex nonlinear relationships and can be combined with model-driven approaches to refine radio wave propagation models.

[0004] Therefore, there is an urgent need for an electromagnetic wave propagation characteristic prediction scheme that can cope with the complex and ever-changing marine environment, so as to accurately predict the propagation characteristics of ultra-low frequency / extremely low frequency electromagnetic waves in the trans-oceanic environment of the sea surface and underwater, and provide reliable technical support for key applications such as deep-sea communication and marine resource exploration. Summary of the Invention

[0005] This application provides a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, so as to accurately predict the propagation characteristics of ultra-low frequency / extremely low frequency electromagnetic waves in a trans-oceanic environment between the sea surface and underwater.

[0006] In a first aspect, embodiments of this application provide a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, the method comprising: Obtain the initial propagation parameters of the target electromagnetic wave, wherein the target electromagnetic wave is an ultra-low frequency electromagnetic wave or an extremely low frequency electromagnetic wave; Acquire initial marine environmental parameter information of the target ocean area within the first target time period. The initial marine environmental parameter information includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of electromagnetic waves of the target. Based on the initial marine environmental parameter information, the target marine environmental parameter information of the target marine area within the second target time period is determined, wherein the second target time period is the first target time period or a subsequent time period of the first target time period; The target marine environmental parameters and the initial propagation parameters are input into a pre-trained electromagnetic wave prediction model to output predicted propagation parameters. The electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. The predicted propagation parameters are used to describe the predicted propagation results of the target electromagnetic wave within the target marine area during the second target time period, with the initial propagation parameters as the starting condition.

[0007] Secondly, embodiments of this application provide a propagation prediction device for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, the device comprising: The acquisition module is used to acquire the initial propagation parameters of the target electromagnetic wave, which is an ultra-low frequency electromagnetic wave or an extremely low frequency electromagnetic wave; and to acquire the initial marine environmental parameter information of the target ocean area within the first target time period, which includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of the target electromagnetic wave. The processing module is used to determine the target marine environmental parameter information of the target marine area within a second target time period based on the initial marine environmental parameter information, wherein the second target time period is the first target time period or a subsequent time period of the first target time period; The prediction module is used to input the target marine environmental parameter information and the initial propagation parameters into a pre-trained electromagnetic wave prediction model, so as to use the electromagnetic wave prediction model to output predicted propagation parameters. The electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. The predicted propagation parameters are used to describe the predicted results of the target electromagnetic wave propagating in the target marine area with the initial propagation parameters as the starting condition during the second target time period.

[0008] Thirdly, embodiments of this application provide an electronic device, including: a memory, a processor, and a communication interface; wherein, the memory stores a computer program, and when the computer program is executed by the processor, the processor can at least implement the propagation prediction method for ultra-low frequency / extremely low frequency electromagnetic waves in the marine environment as described in the first aspect.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor of an electronic device, enables the processor to at least implement the propagation prediction method for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment as described in the first aspect.

[0010] Fifthly, embodiments of this application provide a computer program product, including: a computer program or instructions that, when executed by a processor of an electronic device, enable the processor to at least implement the propagation prediction method for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment as described in the first aspect.

[0011] The solution provided in this application is applied to the scenario of predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in the marine environment. For the target electromagnetic wave to be predicted (the target electromagnetic wave is an ultra-low frequency electromagnetic wave or an extremely low frequency electromagnetic wave), firstly, the initial propagation parameters of the target electromagnetic wave and the initial marine environmental parameter information of the target marine area within a first target time period (e.g., a period relatively close to the current time) are obtained. The initial marine environmental parameter information includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of the target electromagnetic wave. Then, based on the initial marine environmental parameter information, the target marine environmental parameter information of the target marine area within a second target time period is determined. The second target time period is either the first target time period or a subsequent time period (i.e., the second target time period includes future times after the current time). Finally, the target marine environmental parameter information and the initial propagation parameters are input into a pre-trained electromagnetic wave prediction model to output predicted propagation parameters. The electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. The predicted propagation parameters describe the predicted propagation result of the target electromagnetic wave within the target marine area during the second target time period, using the initial propagation parameters as the starting condition.

[0012] This approach addresses two key issues. First, by constructing a marine environmental field covering future periods based on initial marine environmental parameters, it provides physically reasonable and dynamically evolving environmental inputs for electromagnetic wave propagation prediction, resolving the problem of traditional methods being unable to make effective forward predictions due to a lack of future environmental information. Second, the electromagnetic wave prediction model integrates physical constraints and knowledge-guided mechanisms, effectively combining prior physical knowledge of marine electromagnetic propagation with data-driven learning capabilities to accurately predict the propagation characteristics of electromagnetic waves in future periods. This synergistic effect significantly improves prediction accuracy and the reliability of temporal extrapolation, while also enhancing the model's robustness and generalization ability in complex and dynamic marine environments, providing highly reliable electromagnetic wave propagation situation prediction support for critical applications such as underwater communication and remote sensing. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, provided as an embodiment of this application; Figure 2 A flowchart illustrating a method for obtaining initial marine environmental parameter information provided in this application embodiment; Figure 3 A flowchart illustrating a method for obtaining target marine environmental parameter information provided in this application embodiment; Figure 4 A flowchart illustrating an electromagnetic wave prediction model training method provided in this application embodiment; Figure 5 A schematic diagram of the network architecture of an electromagnetic wave prediction model provided in an embodiment of this application; Figure 6 A flowchart illustrating another electromagnetic wave prediction model training method provided in this application embodiment; Figure 7 A schematic diagram of the structure of a propagation prediction device for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, provided for an embodiment of this application; Figure 8 To and Figure 7 The illustrated embodiment provides a schematic diagram of the electronic device corresponding to the propagation prediction device for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] It should be noted that, in the cases involving user information in the embodiments of this application, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) comply with relevant laws and standards.

[0017] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0018] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0019] The propagation prediction method for ultra-low frequency / extremely low frequency electromagnetic waves in the marine environment provided in this application embodiment can be executed by an electronic device, such as a PC, laptop, or smartphone, or a server. The server can be a physical server containing an independent host, a virtual server, a cloud server, or a server cluster.

[0020] Figure 1 A flowchart illustrating a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, as provided in this application embodiment, is shown below. Figure 1 As shown, it may include the following steps: 101. Obtain the initial propagation parameters of the target electromagnetic wave, which is an ultra-low frequency electromagnetic wave or an extremely low frequency electromagnetic wave.

[0021] 102. Obtain initial marine environmental parameter information of the target ocean area within the first target time period. The initial marine environmental parameter information includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of electromagnetic waves from the target.

[0022] 103. Based on the initial marine environmental parameter information, determine the target marine environmental parameter information of the target marine area within the second target time period. The second target time period is the first target time period or a subsequent time period of the first target time period.

[0023] 104. Input the target marine environmental parameters and initial propagation parameters into the pre-trained electromagnetic wave prediction model, and use the electromagnetic wave prediction model to output the predicted propagation parameters. The electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. The predicted propagation parameters are used to describe the prediction results of the target electromagnetic wave propagating in the target marine area with the initial propagation parameters as the starting condition during the second target time period.

[0024] In summary, the propagation prediction method for ultra-low frequency (ULF) / extremely low frequency (ELF) electromagnetic waves in the marine environment provided in this embodiment, in scenarios where it is necessary to predict the propagation of ULF / ELF electromagnetic waves in a target marine area, can not only predict the propagation state of the target electromagnetic wave within a first target time period based on marine environmental parameter information observed or fused within that time period, but can also further extrapolate the marine environmental evolution in a second target time period (including future time periods after the current time). Furthermore, by combining a pre-trained electromagnetic wave prediction model that integrates physical law constraints and knowledge guidance mechanisms, it can achieve high-precision forward-looking prediction of the propagation characteristics of ULF / ELF electromagnetic waves in the future marine environment.

[0025] In practical prediction scenarios, the first step is to obtain the initial propagation parameters of the target electromagnetic wave. These initial propagation parameters are a set of parameters that characterize the state of the target electromagnetic wave at the start of propagation or at its initial location. They serve as input conditions for the electromagnetic wave propagation prediction model and are used to determine the initial boundary conditions for the propagation and evolution of the target electromagnetic wave in the ocean region.

[0026] Optionally, the type of initial propagation parameters required in the prediction scenario can be customized. For example, when the target electromagnetic wave is an ultra-low frequency electromagnetic wave, the initial propagation parameters may include: transmission frequency (e.g., 150 Hz), transmission power (e.g., 2 MW), transmission antenna type (e.g., grounded horizontal electric dipole), antenna tilt angle (e.g., 10°), initial phase (e.g., 0 rad), and transmission source depth (e.g., 50 m below sea surface). When the target electromagnetic wave is an extremely low frequency electromagnetic wave, the initial propagation parameters may include: transmission frequency (e.g., 22 Hz), transmission power (e.g., 1.5 MW), transmission antenna type (e.g., long-distance submarine cable antenna), initial polarization direction (e.g., vertical polarization), initial waveform (e.g., continuous sine wave or pulse-coded sequence), and starting position in the reference coordinate system (e.g., latitude, longitude, and depth).

[0027] Then, the initial marine environmental parameter information of the target ocean area within the first target time period is obtained. The initial marine environmental parameter information includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of electromagnetic waves of the target.

[0028] Among them, surface environmental factors are marine physical parameters located at or near the ocean surface (e.g., within a depth range of 0–50 meters) that significantly affect the propagation of ultra-low frequency / extremely low frequency electromagnetic waves. Underwater environmental factors are key ocean state parameters located within the water body below the sea surface (e.g., from a depth greater than 50 meters to the seabed) that directly affect the attenuation, phase velocity, and propagation path of electromagnetic waves in the marine medium.

[0029] Optionally, surface environmental elements and underwater environmental elements can be customized based on actual prediction needs. For example, preset surface environmental elements may include: sea surface wind speed, sea surface roughness, significant wave height, sea surface temperature, sea ice coverage, ocean currents, etc.; underwater environmental elements may include: seawater temperature, seawater salinity, vertical distribution of electrical conductivity, density stratification, mixing layer depth, and bottom sediment electrical conductivity, current velocity, etc. This application does not limit the types of surface environmental elements and underwater environmental elements.

[0030] For ease of description, surface environmental elements and underwater environmental elements are collectively referred to as marine environmental parameter information. To distinguish marine environmental parameter information from different data sources or time attributes, this application may use terms such as "initial marine environmental parameter information," "target marine environmental parameter information," or "historical marine environmental parameter information," but in essence, they all refer to the aforementioned surface environmental elements and underwater environmental elements.

[0031] In practice, the acquisition of initial marine environmental parameters is not limited to specific data sources or processing methods. For example, on-site measured data can be collected directly using satellite remote sensing, in-situ monitoring, etc., as initial marine environmental parameter information. Alternatively, high-resolution regional ocean numerical models, such as the Coastal and Regional Ocean Community model (CROCO) and the Massachusetts Institute of Technology General Circulation Model (MITgcm), can be used to generate the required environmental field through numerical simulation, serving as initial marine environmental parameter information. Furthermore, mature global or regional ocean reanalysis datasets (such as ERA5, ERA-interim, JRA-55, CFSR, etc.) can be directly accessed to obtain initial marine environmental parameter information.

[0032] Optionally, any two or all three types of data from the above-mentioned measured data, numerical simulation data, and reanalysis data can be spatiotemporally aligned and multi-source fused to construct more complete, accurate, and physically consistent initial marine environmental parameter information.

[0033] Next, we will take the three-source fusion method of "satellite remote sensing and in-situ monitoring (measurement) + regional ocean numerical model (simulation) + ocean reanalysis dataset (reanalysis)" as an example for illustration, but it is not limited to one instance.

[0034] Figure 2 A flowchart illustrating a method for obtaining initial marine environmental parameter information provided in this application embodiment is shown below. Figure 2 As shown, it includes at least the following steps: 201. Obtain measured marine environmental data for the first target time period from satellite remote sensing data and in-situ monitoring data covering the target marine area.

[0035] In ocean and climate science, satellite observation and in-situ monitoring are two important methods for acquiring observational data.

[0036] Satellite observation utilizes remote sensing technology to acquire large-scale environmental information about the Earth's surface from space. It features wide coverage, high temporal resolution, and long-term monitoring capabilities, making it suitable for observing the sea surface environment. For example, the Surface Water and Ocean Topography (SWOT) satellite has an effective spatial resolution of up to 20 kilometers in the ocean, making it suitable for deep-sea environmental exploration. It can be used to observe sea surface height, wave height, and wave direction. In practice, optionally, based on the element categories corresponding to the initial marine environmental parameters to be acquired, appropriate satellites can be selected to observe the target ocean area and obtain satellite remote sensing data.

[0037] In-situ monitoring refers to the direct environmental monitoring of the target ocean area by in-situ monitoring equipment. It features high precision, detailed parameters, localized research, and direct contact with the target, which can reduce signal deviation problems that may occur in remote sensing. Optionally, in-situ monitoring of ocean wave data, temperature, salinity, pressure, and current field profiles can be obtained through ocean vessels, underwater base probes, surface buoy probes, shore-based fixed probes, and mobile probes.

[0038] It is easy to understand that, for a target ocean area, satellite remote sensing data acquired through satellite observation and in-situ monitoring data acquired through in-situ monitoring typically cover a continuous or discrete time series (e.g., the past 30 days, 90 days, or longer). In practical applications, a subset of data matching the first target time period can be selected from the above time series according to the needs of the prediction task. This subset is then preprocessed with time alignment, outlier removal, and unit standardization to generate measured marine environmental data within the first target time period, which serves as the basic input for subsequent fusion processing.

[0039] 202. Numerical simulation of the target ocean area is carried out using a regional ocean numerical model to obtain marine environmental simulation data for the first target time period.

[0040] Optionally, the regional ocean numerical model can be CROCO or MITgcm, or other numerical models capable of numerically simulating ocean regions. This application does not impose any restrictions on this.

[0041] In the specific implementation process, firstly, based on the geographical scope of the target ocean region, such as latitude and longitude boundaries, water depth and topography, a computational domain is constructed, and corresponding horizontal and vertical grid resolutions are configured, such as a horizontal resolution of 100 meters to 1 kilometer and a vertical layer of no less than 30 layers. Subsequently, atmospheric forcing fields (such as wind speed, air pressure, heat flux, etc., which can be derived from ERA5 or JRA-55 reanalysis data), open boundary conditions (such as tides, offshore temperature, salinity and current profiles), and initial fields (such as temperature and salinity distribution from reanalysis data) are input into the selected numerical model, i.e., the regional ocean numerical model. The regional ocean numerical model outputs high spatiotemporal resolution three-dimensional ocean state variables, including but not limited to: seawater temperature, salinity, electrical conductivity, density, three-dimensional current velocity, mixing layer depth, sea surface height, and eddy kinetic energy, by performing non-hydrostatic and nonlinear numerical integration of ocean dynamics and thermodynamic processes within the first target time period (e.g., the past 7 days). Finally, after format conversion and coordinate unification, the output results are used as marine environmental simulation data for the first target time period, which will be used for subsequent multi-source fusion with measured data and reanalysis data.

[0042] Among them, the marine environment simulation data obtained by numerical simulation through regional marine numerical models has the characteristics of continuous spatial coverage and complete time series, which can effectively make up for the deficiencies of the measured marine environment data in spatially sparse areas or at time discontinuities.

[0043] 203. Extract marine environmental reanalysis data from the marine reanalysis dataset that matches the first target time period of the target marine area.

[0044] Ocean reanalysis data are marine environmental datasets generated through data assimilation techniques that dynamically fuse historical satellite remote sensing observations, in-situ monitoring data, and physical oceanographic numerical models. These datasets are global or regional in scale, spanning long time series, and exhibiting high spatiotemporal consistency. This type of data combines the realism of observations with the dynamic consistency of models, providing multidimensional environmental variables that cover the entire ocean, have no spatial gaps, and are temporally continuous.

[0045] In this embodiment, optionally, the ocean reanalysis dataset includes, but is not limited to: ERA5 and ERA-Interim released by the European Centre for Medium-Range Weather Forecasts (ECMWF), JRA-55 from the Japan Meteorological Agency (JMA), and CFSR (Climate Forecast System Reanalysis) from the National Center for Environmental Prediction (NCEP) of the United States.

[0046] Taking ERA5 as an example, its horizontal spatial resolution is approximately 31 kilometers, and its temporal resolution reaches the hourly level. It not only provides atmospheric forcing fields such as sea surface wind speed, air temperature, precipitation, and downward shortwave radiation, but also includes subsurface ocean variables (such as 0–1000 meter temperature-salinity profiles, ocean currents, and mixed layer depth) generated by coupling with the ocean module. These variables undergo rigorous quality control and physical constraints, exhibiting high spatiotemporal continuity and long-term stability.

[0047] In the specific implementation process, firstly, based on the target ocean area (e.g., latitude and longitude range) and the first target time period (e.g., January 1, 2024 to January 7, 2024) determined by the prediction task, marine environmental variables within the corresponding spatiotemporal range are extracted from the selected reanalysis dataset. Subsequently, the extracted data undergoes coordinate projection transformation, unit standardization, and vertical interpolation (e.g., interpolating the model layer to the standard depth layer) to ensure spatiotemporal grid compatibility with data from other sources (e.g., measured marine environmental data and simulated marine environmental data). Finally, the generated standardized data is used as the marine environmental reanalysis data for the first target time period for subsequent multi-source fusion processing.

[0048] 204. Integrate measured marine environmental data, simulated marine environmental data, and reanalysis data of marine environment to generate initial marine environmental parameter information aligned in both time and spatial dimensions.

[0049] Specifically, measured marine environmental data, simulated marine environmental data, and reanalysis marine environmental data are spatiotemporally aligned on a unified time grid (e.g., every 1 hour) and spatial grid (e.g., a 0.1° × 0.1° latitude and longitude grid). Subsequently, multi-source data fusion methods (e.g., optimal interpolation, co-kriging, or deep learning fusion models) are used to weightedly fuse the three types of data.

[0050] In this embodiment, measured marine environmental data is used to correct local deviations between simulated marine environmental data and reanalysis marine environmental data. Simulated and reanalysis marine environmental data are used to fill spatial gaps and temporal discontinuities in the measured marine environmental data. Based on this, fusion can generate initial marine environmental parameter information that is spatiotemporally continuous, physically reasonable, and comprehensively covered.

[0051] After obtaining the initial marine environmental parameter information of the target ocean area within the first target time period, the next step is to determine the target marine environmental parameter information of the target ocean area within the second target time period to be predicted.

[0052] The second target time period can be the first target time period itself, that is, based on the known state of the marine environment, the propagation of the target electromagnetic wave in the target marine area within the known time period is inverted or the state is assessed; the second target time period can also be a subsequent time period of the first target time period, that is, based on current or recent observations, the future evolution of the marine environment is inferred, thereby making a forward-looking prediction of the propagation characteristics of the target electromagnetic wave in the future period.

[0053] The subsequent time period of the first target time period refers to the time interval on the time axis that is completely located after the end of the first target time period, or the time interval that partially or completely overlaps with the first target time period but is not earlier than the first target time period as a whole. That is, any moment of the second target time period is not earlier than the start moment of the first target time period.

[0054] Figure 3 A flowchart illustrating a method for obtaining target marine environmental parameter information provided in this application embodiment is shown below. Figure 3 As shown, it includes at least the following steps: 301. Input the initial marine environmental parameter information into the pre-trained marine environmental field prediction model; the marine environmental field prediction model includes a recurrent neural network for temporal extrapolation and a generative deep network for spatial field reconstruction. The recurrent neural network learns the spatiotemporal evolution law of the marine environmental state and the dynamic process controlling the evolution law during the training phase.

[0055] 302. By extrapolating the initial marine environmental parameters using a recurrent neural network, the evolution trend corresponding to the second target time period is obtained.

[0056] 303. Based on the evolution trend, the spatial field is reconstructed using a generative deep network to generate a three-dimensional marine environmental field of the target marine area within the second target time period, and output as the target marine environmental parameter information.

[0057] In this embodiment, the marine environmental field prediction model adopts a two-stage architecture of "temporal extrapolation + spatial generation" to achieve high-precision and high-resolution prediction of the future marine environmental field. Specifically, the marine environmental field prediction model includes a recurrent neural network (RNN) for temporal extrapolation and a generative deep network for spatial field reconstruction.

[0058] Optionally, the recurrent neural network can employ a Long Short-Term Memory (LSTM) network. The recurrent neural network in this embodiment has already learned the nonlinear spatiotemporal evolution patterns of marine environmental elements such as sea surface temperature and salinity, and the underlying dynamic control mechanisms (such as wind stress driving, heat flux forcing, geostrophic equilibrium, etc.) during the training phase. Therefore, during the usage phase, it can perform temporal extrapolation on the marine environmental elements contained in the initial input marine environmental parameter information to obtain the evolution trend of these marine environmental elements from the first target time period to the second target time period.

[0059] Generative deep networks can employ 3D U-Net architectures or Generative Adversarial Networks (GANs). The generative deep network in this embodiment has already learned statistical priors about complex spatial structures such as ocean turbulence, thermohaline layers, and mixing layers during the training phase. Therefore, during the usage phase, it can generate a reasonable and detailed three-dimensional ocean environmental field under the constraint of the evolution trend of the recurrent network output. This three-dimensional ocean environmental field includes sea surface environmental elements (e.g., sea surface temperature, significant wave height, wind speed) and underwater environmental elements (e.g., temperature, salinity, vertical conductivity distribution, and three-dimensional current field at depths of 0–200 meters). Finally, the three-dimensional ocean environmental field is output as target ocean environmental parameter information for electromagnetic wave propagation prediction.

[0060] It should be noted that when the second target time period is completely consistent with the first target time period (i.e., when time extrapolation is not required), the marine environmental field prediction model in this embodiment can degenerate into a high-resolution spatial reconstruction module. In this case, the recurrent neural network can directly transmit the input initial marine environmental parameter information as an "evolution trend" to the generative deep network. The latter, based on this known state, uses its learned spatial priors (such as thermo-salinity strata structure, small-scale sea surface fluctuations, etc.) to perform super-resolution enhancement or physical consistency repair on the initial marine environmental parameter information, thereby outputting a more detailed and structurally complete three-dimensional marine environmental field. This scenario is suitable for high-fidelity inversion or refined evaluation of the propagation state of target electromagnetic waves within the target marine area during historical or current time periods.

[0061] After determining the target marine environmental parameters for the target ocean area within the second target time period, the target marine environmental parameters and initial propagation parameters are further input into a pre-trained electromagnetic wave prediction model to output predicted propagation parameters. These predicted propagation parameters describe the predicted propagation results of the target electromagnetic wave within the target ocean area during the second target time period, using the initial propagation parameters as the starting condition.

[0062] The prediction results describe the evolution of key electromagnetic characteristics of the target electromagnetic wave along its propagation path over time and space, providing a basis for decision-making in applications such as underwater communication link design, remote detection system scheduling, and electromagnetic environment situational awareness. Optionally, the predicted propagation parameters may include, for example, the time series of electric field strength (or magnetic field strength) at the receiving point, propagation attenuation, the time required for the signal envelope to pass through the ocean medium, phase delay, signal-to-noise ratio, and effective propagation depth. This application does not limit the types of predicted propagation parameters; in practical applications, they can be customized based on specific prediction requirements.

[0063] In summary, this approach addresses two key issues. First, by constructing a marine environmental field covering future periods based on initial marine environmental parameters, it provides physically reasonable and dynamically evolving environmental inputs for electromagnetic wave propagation prediction, resolving the problem of traditional methods being unable to make effective forward predictions due to a lack of future environmental information. Second, the electromagnetic wave prediction model integrates physical constraints and knowledge-guided mechanisms, effectively combining prior physical knowledge of marine electromagnetic propagation with data-driven learning capabilities to accurately predict the propagation characteristics of electromagnetic waves in future periods. This synergistic effect significantly improves prediction accuracy and the reliability of temporal extrapolation, while also enhancing the model's robustness and generalization ability in complex and dynamic marine environments, providing highly reliable electromagnetic wave propagation situation prediction support for key applications such as underwater communication and remote sensing.

[0064] In this embodiment, the electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. It is understood that the working principle of the model during use is the same or similar to that during training. Therefore, the working principle of the electromagnetic wave prediction model will be explained below by describing the training process of the electromagnetic wave prediction model.

[0065] Figure 4 A flowchart of an electromagnetic wave prediction model training method provided in this application embodiment is shown below. Figure 4 As shown, it may include the following steps: 401. Obtain historical marine environmental parameter information of the target marine area within a historical time period, as well as sample initial propagation parameters and sample target propagation parameters of the target electromagnetic wave propagating in the target marine area, to serve as training sample pairs.

[0066] The process of obtaining historical marine environmental parameter information can refer to the process of obtaining initial or target marine environmental parameter information in the aforementioned embodiments, and will not be repeated here.

[0067] The sample target propagation parameters are the results of the target electromagnetic wave propagating within the target ocean area under the condition that the target ocean area is in the ocean environment corresponding to the historical ocean environmental parameter information, with the initial propagation parameters of the sample as the starting condition. Optionally, the sample target propagation parameters can be actual measurement results or verified numerical simulation results. The embodiments of this application do not limit the method of obtaining the sample target propagation parameters.

[0068] 402. Based on the training sample pairs, construct a knowledge graph to describe the relationship between marine environmental elements and electromagnetic wave propagation characteristics, and a physical constraint model to describe the propagation behavior of target electromagnetic waves in the marine environment.

[0069] As an optional method for constructing a knowledge graph, a language model can be used to identify and extract marine environmental elements (e.g., sea surface temperature, sea surface height, sea surface salinity), electromagnetic wave propagation elements (e.g., signal strength, ionospheric electron concentration, conductivity), and the relationships between marine environmental elements and electromagnetic wave propagation elements (e.g., the relationship between temperature, salinity, and conductivity; the relationship between seabed topographic roughness and electromagnetic wave reflection coefficient) from training sample pairs. Then, marine environmental elements and electromagnetic wave propagation elements are treated as entities, and the relationships between them are treated as relation edges, to construct a knowledge graph describing the relationships between marine environmental elements and electromagnetic wave propagation characteristics.

[0070] The physical constraint model is used to describe the fundamental physical laws that must be satisfied for the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in the ocean medium, ensuring that the model output conforms to objective laws of nature. Its construction is based on classical electromagnetic theory and ocean electrodynamics. Optionally, the physical constraint model can be based on Maxwell's equations.

[0071] 403. Knowledge graphs and physical constraint models are introduced as prior constraints into electromagnetic wave prediction models.

[0072] Figure 5 A schematic diagram of the network architecture of an electromagnetic wave prediction model provided in an embodiment of this application is shown below. Figure 5 As shown, the electromagnetic wave prediction model includes: TrellisNet main network and PINN physical information neural network.

[0073] The TrellisNet main network is used to perform time-series modeling of the input marine environmental parameters and electromagnetic wave propagation parameters. Optionally, a knowledge graph can be embedded in the TrellisNet main network.

[0074] In this embodiment, the gating weights of the TrellisNet main network are initialized based on the correlation between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph. The gating weights are used to ensure that the TrellisNet main network retains marine environmental elements whose correlation with electromagnetic wave propagation is greater than a preset threshold (e.g., 85%) during the time-series modeling process, thereby capturing the long-range dependency characteristics of the target electromagnetic wave during propagation.

[0075] The Physical Information Neural Network (PINN) is constructed based on the partial differential equations corresponding to the physical constraint model and is used to calculate the residuals of electromagnetic wave propagation prediction results in the partial differential equations. During model training, the residuals of the electromagnetic wave propagation prediction data output by the electromagnetic wave prediction model in the partial differential equations can be used to characterize the physical loss, that is, the degree to which the electromagnetic wave propagation prediction data does not satisfy the physical constraint model.

[0076] 404. Input the training sample pairs into the electromagnetic wave prediction model so that, guided by the knowledge graph, the electromagnetic wave prediction model outputs electromagnetic wave propagation prediction data of the target electromagnetic wave within the target ocean area, using the initial propagation parameters of the samples as the starting conditions.

[0077] 405. Calculate the data loss and physical loss. The data loss is used to describe the deviation between the electromagnetic wave propagation prediction data and the sample target propagation parameters, while the physical loss is used to describe the degree to which the electromagnetic wave propagation prediction data does not meet the physical constraint model.

[0078] As mentioned above, physical loss can be represented by the residual in the partial differential equation of the electromagnetic wave propagation prediction data. If the deviation is 0, it indicates that it fully conforms to the physical laws.

[0079] 406. Construct a joint loss function based on data loss and physical loss, and optimize the model parameters of the electromagnetic wave prediction model with the goal of minimizing the joint loss function until the model converges.

[0080] In an optional embodiment, a joint loss function is constructed based on data loss and physical loss, which can be expressed as: in, Indicates data loss. Indicates physical loss. These are the weighting coefficients corresponding to the data loss. It is the weighting coefficient corresponding to the physical loss. and Used to balance the proportion of data loss and physical loss in the loss function.

[0081] The electromagnetic wave prediction model trained in this scheme embeds partial differential equations derived from Maxwell's equations as physical constraints into the neural network architecture. It also incorporates a knowledge graph, constructed from the correlation between marine environmental factors and electromagnetic propagation characteristics, to provide prior guidance to the model structure. This allows the model to not only fit the statistical laws governing marine environmental parameters and electromagnetic wave propagation during the learning process but also internalize the physical essence that electromagnetic waves must follow to propagate in conductive marine media. This fusion mechanism enables the model to effectively suppress non-physical interpretations easily generated by purely data-driven methods (i.e., machine learning methods) when facing dynamic, non-uniform, and spatiotemporally continuously evolving marine environments. It also overcomes the systematic biases caused by the oversimplification of marine environments in traditional analytical or numerical simulations. Physical constraints ensure that the output field satisfies the fundamental laws of electromagnetic fields at any spatiotemporal point, while knowledge guidance allows the network to focus on environmental variables and their evolution paths that play a dominant role in electromagnetic wave propagation from the input information. Therefore, even under complex marine conditions, it can accurately characterize the changing trends of key characteristics such as attenuation, phase shift, and propagation depth of ultra-low frequency / extremely low frequency electromagnetic waves. Thus, while maintaining the strong fitting ability of deep learning, the model achieves physical interpretability and extrapolation robustness, enabling accurate and reliable prediction of electromagnetic wave propagation behavior in the marine environment.

[0082] In one optional embodiment, during the use of the electromagnetic wave prediction model, the trained electromagnetic wave prediction model can be further optimized and trained.

[0083] Figure 6 A flowchart of another electromagnetic wave prediction model training method provided in this application embodiment, applied after obtaining the trained electromagnetic wave prediction model, such as... Figure 6 As shown, it may include the following steps: 601. Obtain newly added historical marine environmental parameter information and corresponding new sample initial propagation parameters and new sample target propagation parameters to form new training sample pairs.

[0084] 602. A bidirectional long short-term memory network is used to perform bidirectional time series modeling on the time series of newly added historical marine environmental parameters in order to extract dynamic representations of marine environmental elements.

[0085] 603. Based on dynamic characterization, new initial propagation parameters of the sample, and new target propagation parameters of the sample, calculate the causal intensity of each marine environmental element on the electromagnetic wave propagation characteristics.

[0086] 604. Update the relationship between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph based on causal strength.

[0087] 605. Adjust the gating weights of the TrellisNet main network based on the correlation between marine environmental elements and electromagnetic wave propagation characteristics in the updated knowledge graph.

[0088] 606. Using the new training sample pairs as fine-tuning samples, retrain the electromagnetic wave prediction model after adjusting the gating weights until the model converges, and obtain the optimized electromagnetic wave prediction model.

[0089] In this embodiment, continuous optimization of the electromagnetic wave prediction model is achieved by introducing incremental learning and causal-driven knowledge evolution mechanisms. Specifically, after the electromagnetic wave prediction model is deployed, newly added environmental parameter information and corresponding electromagnetic propagation parameter samples can be continuously collected to form new training sample pairs. A bidirectional long short-term memory network (BiLSTM) is used to perform forward and backward joint modeling of the marine environment time series, fully capturing the dynamic characteristics jointly contained in the historical evolution and future trends of marine environmental elements. On this basis, the actual influence intensity of each marine environmental element on the electromagnetic wave propagation characteristics is quantified by causal inference methods, and the association weights in the knowledge graph are dynamically updated accordingly. The updated knowledge graph is further used to adjust the gating weights of the TrellisNet main network, so that the model adaptively strengthens the focus on high causal intensity elements and weakens the influence of irrelevant or interfering variables during the time series modeling process. Finally, the electromagnetic wave prediction model is fine-tuned with new training sample pairs to complete the closed-loop evolution from static knowledge to dynamic cognition.

[0090] As an alternative method to calculate the causal strength of each marine environmental element on electromagnetic wave propagation characteristics, the Liang-Kleeman information flow method can be used to calculate the information transfer from each marine environmental element to electromagnetic wave propagation characteristics based on dynamic characterization, new initial propagation parameters of the sample, and new target propagation parameters of the sample. Then, based on the magnitude and sign of the information transfer, the causal strength and causal direction of each marine environmental element on electromagnetic wave propagation characteristics can be determined. Finally, the causal strength can be used as the association weight to update the directed association between the corresponding marine environmental element and electromagnetic wave propagation characteristics in the knowledge graph.

[0091] In Liang's information flow method, we first assume a d-dimensional dynamical system, which has the following form: F is the vector of drift coefficients, θ is the vector of parameters, and B is the diffusion coefficient matrix. = b ij W is the vector of the standard Wiener process ( It's white noise.

[0092] Consider the case where d=2: The flow rate of the information stream from X2 to X1 can then be expressed as: in ρ 1 X1 is the marginal probability density, and E represents the mathematical expectation.

[0093] like T 2→1 If X = 0, then the change of X1 is independent of X2, that is, X2 is unrelated to X1.

[0094] A rigorously formulated form can be obtained under the linear assumption. T 2→1 Maximum likelihood estimation: in, C ij ( i, j=1,2 )for X i and X j The sample covariance between them C i, dj for X i The new sequence obtained by forward difference The sample covariance between △t For time step.

[0095] like T 2→1 ≠0 If X2 is a factor of X1, then X2 is a factor of X1.

[0096] The multivariate causal analysis form of Liang's information flow method can be written as: in, C It is the covariance matrix. d It is the number of variables. Δ jk yes C cofactors.

[0097] This application employs the information flow calculation method of this multivariate causal analysis to provide auxiliary verification analysis of the relationship between marine environmental elements and electromagnetic wave propagation characteristics. After normalizing the calculated information flow value Tj→i, it can be used to update the directed correlation between the corresponding marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph. This, in turn, adjusts the gating weights of the TrellisNet main network, enabling the electromagnetic wave prediction model to automatically focus on marine environmental elements with high causal correlation, eliminating redundant interference, and thus improving the convergence speed and robustness of the electromagnetic wave prediction model.

[0098] The model optimization mechanism of this scheme enables the electromagnetic wave prediction model to continuously evolve with the accumulation of marine environmental parameter information. Since the knowledge graph is no longer fixed to the statistical associations of the initial training phase, but is dynamically corrected based on the real causal relationships in the newly added data, the model's understanding of the relationship between marine environmental factors and electromagnetic wave propagation characteristics is closer to physical reality. Simultaneously, the adaptive adjustment of TrellisNet's gating weights ensures that the network structure always focuses on the most discriminative propagation influencing factors in the current environment, thus maintaining high accuracy and strong robustness even under scenarios such as seasonal changes in ocean conditions, extreme event disturbances, or regional migration. Furthermore, the physical constraint model (such as partial differential equations derived from Maxwell's equations) continues to play a role throughout the retraining process, ensuring that even after knowledge updates and parameter fine-tuning, the model's output still strictly satisfies the fundamental physical laws of electromagnetic propagation. Therefore, this scheme not only achieves online improvement of model performance but also provides a physically reliable, knowledge-evolvable, and structurally adjustable intelligent prediction strategy, significantly enhancing the adaptability and reliability of electromagnetic wave propagation prediction in long-term operation and cross-scenario applications.

[0099] The following describes in detail one or more embodiments of this application a propagation prediction device for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment. Those skilled in the art will understand that these devices can all be configured using commercially available hardware components through the steps taught in this solution.

[0100] Figure 7 A schematic diagram of a propagation prediction device for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, provided as an embodiment of this application, is shown below. Figure 7 As shown, the device includes: an acquisition module 11, a processing module 12, and a prediction module 13.

[0101] The acquisition module 11 is used to acquire the initial propagation parameters of the target electromagnetic wave, wherein the target electromagnetic wave is an ultra-low frequency electromagnetic wave or an extremely low frequency electromagnetic wave; and to acquire the initial marine environmental parameter information of the target ocean area within the first target time period, wherein the initial marine environmental parameter information includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of the target electromagnetic wave.

[0102] Processing module 12 is used to determine the target marine environmental parameter information of the target marine area within a second target time period based on the initial marine environmental parameter information, wherein the second target time period is the first target time period or a subsequent time period of the first target time period.

[0103] Prediction module 13 is used to input the target marine environmental parameter information and the initial propagation parameters into a pre-trained electromagnetic wave prediction model, so as to use the electromagnetic wave prediction model to output predicted propagation parameters. The electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. The predicted propagation parameters are used to describe the predicted results of the target electromagnetic wave propagating in the target marine area with the initial propagation parameters as the starting condition during the second target time period.

[0104] Optionally, the acquisition module 11 is specifically used to: acquire measured marine environmental data within a first target time period from satellite remote sensing data and in-situ monitoring data covering the target marine area; perform numerical simulation on the target marine area using a regional marine numerical model to acquire simulated marine environmental data within the first target time period; extract marine environmental reanalysis data matching the first target time period of the target marine area from the marine reanalysis dataset; and fuse the measured marine environmental data, the simulated marine environmental data, and the reanalysis marine environmental data to generate initial marine environmental parameter information aligned in both time and spatial dimensions.

[0105] Optionally, the processing module 12 is specifically used for: inputting the initial marine environmental parameter information into a pre-trained marine environmental field prediction model; the marine environmental field prediction model includes a recurrent neural network for temporal extrapolation and a generative deep network for spatial field reconstruction, wherein the recurrent neural network learns the spatiotemporal evolution law of the marine environmental state and the dynamic process controlling the evolution law during the training phase; performing temporal extrapolation on the initial marine environmental parameter information through the recurrent neural network to obtain the evolution trend corresponding to the second target time period; and performing spatial field reconstruction based on the evolution trend through the generative deep network to generate a three-dimensional marine environmental field of the target marine area within the second target time period, and outputting it as the target marine environmental parameter information.

[0106] Optionally, the device further includes a model training module, which is configured to: acquire historical marine environmental parameter information of the target marine area within a historical time period, as well as sample initial propagation parameters and sample target propagation parameters of the target electromagnetic wave propagating in the target marine area, as training sample pairs; construct a knowledge graph describing the relationship between marine environmental elements and electromagnetic wave propagation characteristics, and a physical constraint model describing the propagation behavior of the target electromagnetic wave in the marine environment, based on the training sample pairs; introduce the knowledge graph and the physical constraint model as prior constraints into the electromagnetic wave prediction model; and input the training sample pairs into the electromagnetic wave prediction model. An electromagnetic wave prediction model is constructed to output electromagnetic wave propagation prediction data of the target electromagnetic wave within the target ocean area, guided by the knowledge graph, using the initial propagation parameters of the sample as the starting condition. Data loss and physical loss are calculated, whereby the data loss describes the deviation between the electromagnetic wave propagation prediction data and the sample target propagation parameters, and the physical loss describes the degree to which the electromagnetic wave propagation prediction data does not satisfy the physical constraint model. A joint loss function is constructed based on the data loss and the physical loss, and the model parameters of the electromagnetic wave prediction model are optimized by minimizing the joint loss function until the model converges.

[0107] Optionally, the electromagnetic wave prediction model includes: a TrellisNet main network and a Physical Information Neural Network (PINN); the TrellisNet main network is used to perform time-series modeling of the input marine environmental parameters and electromagnetic wave propagation parameters; the gating weights of the TrellisNet main network are initialized based on the correlation between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph; the gating weights are used to ensure that the TrellisNet main network retains marine environmental elements with a correlation greater than a preset threshold with electromagnetic wave propagation during the time-series modeling process, so as to capture the long-range dependency characteristics of the target electromagnetic wave during propagation; the Physical Information Neural Network (PINN) is constructed based on the partial differential equation corresponding to the physical constraint model and is used to calculate the residual of the electromagnetic wave propagation prediction data in the partial differential equation, and the residual is used to characterize the physical loss.

[0108] Optionally, the model training module is further configured to: acquire newly added historical marine environmental parameter information and corresponding new initial sample propagation parameters and new target sample propagation parameters to form new training sample pairs; utilize a bidirectional long short-term memory network to perform bidirectional time-series modeling on the time series of the newly added historical marine environmental parameter information to extract the dynamic representation of the marine environmental elements; calculate the causal strength of each marine environmental element on electromagnetic wave propagation characteristics based on the dynamic representation, the new initial sample propagation parameters, and the new target sample propagation parameters; update the association between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph based on the causal strength; adjust the gating weights of the TrellisNet main network based on the updated association between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph; and retrain the electromagnetic wave prediction model after adjusting the gating weights using the new training sample pairs as fine-tuning samples until the model converges to obtain the optimized electromagnetic wave prediction model.

[0109] Optionally, the model training module is specifically used to: calculate the information transfer amount from each marine environmental element to the electromagnetic wave propagation characteristics using the Liang's information flow method based on the dynamic representation, the new initial propagation parameters of the sample, and the new target propagation parameters of the sample; determine the causal strength and causal direction of each marine environmental element to the electromagnetic wave propagation characteristics based on the magnitude and sign of the information transfer amount; and update the directed association relationship between the corresponding marine environmental element and the electromagnetic wave propagation characteristics in the knowledge graph using the causal strength as the association weight.

[0110] Figure 7 The device shown can perform the steps described in the foregoing embodiments. For detailed execution process and technical effects, please refer to the description in the foregoing embodiments, which will not be repeated here.

[0111] In one possible design, the above Figure 7 The structure of the propagation prediction device for ultra-low frequency / extremely low frequency electromagnetic waves in the marine environment shown can be implemented as an electronic device, such as... Figure 8 As shown, the electronic device may include: a memory 21, a processor 22, and a communication interface 23. The memory 21 stores a computer program, which, when executed by the processor 22, enables the processor 22 to at least implement the propagation prediction method for ultra-low frequency / extremely low frequency electromagnetic waves in the marine environment as provided in the foregoing embodiments.

[0112] The aforementioned memory 21 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0113] Accordingly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to implement the steps in the above-described method embodiments. The computer-readable storage medium includes volatile or non-volatile components, or a combination thereof, and can be removable or non-removable. Examples of computer-readable storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technologies, CD-ROM, Digital Video Disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium. Accordingly, this application also provides a computer program product, which includes a computer program or instructions that, when executed by a processor, enable the processor to implement the steps in the above-described method embodiments. It should be understood that each step or combination of steps in the above-described method flow can be implemented by the computer program or instructions. Furthermore, these computer programs or instructions can be applied to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable device for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, enabling the processor of such a device to function as an apparatus for implementing the corresponding functions in the above-described method embodiments.

[0114] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0115] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of a necessary general-purpose hardware platform, or by a combination of hardware and software. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a computer product. This application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] Finally, it should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0117] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment, characterized in that, include: Obtain the initial propagation parameters of the target electromagnetic wave, wherein the target electromagnetic wave is an ultra-low frequency electromagnetic wave or an extremely low frequency electromagnetic wave; Acquire initial marine environmental parameter information of the target ocean area within the first target time period. The initial marine environmental parameter information includes preset sea surface environmental elements and underwater environmental elements that affect the propagation of electromagnetic waves of the target. Based on the initial marine environmental parameter information, the target marine environmental parameter information of the target marine area within the second target time period is determined, wherein the second target time period is the first target time period or a subsequent time period of the first target time period; The target marine environmental parameters and the initial propagation parameters are input into a pre-trained electromagnetic wave prediction model to output predicted propagation parameters. The electromagnetic wave prediction model integrates physical law constraints and knowledge guidance mechanisms. The predicted propagation parameters are used to describe the predicted propagation results of the target electromagnetic wave within the target marine area during the second target time period, with the initial propagation parameters as the starting condition.

2. The method according to claim 1, characterized in that, The acquisition of initial marine environmental parameter information for the target ocean area within the first target time period includes: Obtain measured marine environmental data for the first target time period from satellite remote sensing data and in-situ monitoring data covering the target marine area; Numerical simulations of the target ocean region are performed using a regional ocean numerical model to obtain ocean environment simulation data for the first target time period. Extract marine environmental reanalysis data that matches the first target time period for the target marine area from the marine reanalysis dataset; The measured marine environment data, the simulated marine environment data, and the reanalysis marine environment data are integrated to generate initial marine environment parameter information aligned in both time and spatial dimensions.

3. The method according to claim 1, characterized in that, The step of determining the target marine environmental parameter information of the target marine area within the second target time period based on the initial marine environmental parameter information includes: The initial marine environmental parameter information is input into a pre-trained marine environmental field prediction model; the marine environmental field prediction model includes a recurrent neural network for temporal extrapolation and a generative deep network for spatial field reconstruction. The recurrent neural network learns the spatiotemporal evolution law of the marine environmental state and the dynamic process controlling the evolution law during the training phase. The initial marine environmental parameter information is extrapolated over time using the recurrent neural network to obtain the evolution trend corresponding to the second target time period. The generative deep network reconstructs the spatial field based on the evolution trend, generates a three-dimensional marine environmental field of the target marine area within the second target time period, and outputs it as the target marine environmental parameter information.

4. The method according to any one of claims 1 to 3, characterized in that, The electromagnetic wave prediction model is trained in the following manner: Historical marine environmental parameter information of the target marine area within a historical time period, as well as sample initial propagation parameters and sample target propagation parameters of the target electromagnetic wave propagating in the target marine area, are obtained as training sample pairs. Based on the training sample pairs, a knowledge graph is constructed to describe the relationship between marine environmental elements and electromagnetic wave propagation characteristics, and a physical constraint model is constructed to describe the propagation behavior of the target electromagnetic wave in the marine environment. The knowledge graph and the physical constraint model are introduced as prior constraints into the electromagnetic wave prediction model. The training sample pairs are input into the electromagnetic wave prediction model so that, guided by the knowledge graph, the electromagnetic wave prediction model outputs electromagnetic wave propagation prediction data for the target electromagnetic wave within the target ocean area, using the initial propagation parameters of the samples as the starting conditions. Calculate data loss and physical loss, wherein the data loss is used to describe the deviation between the electromagnetic wave propagation prediction data and the sample target propagation parameters, and the physical loss is used to describe the degree to which the electromagnetic wave propagation prediction data does not meet the physical constraint model; A joint loss function is constructed based on the data loss and the physical loss. The model parameters of the electromagnetic wave prediction model are optimized with the goal of minimizing the joint loss function until the model converges.

5. The method according to claim 4, characterized in that, The electromagnetic wave prediction model includes: TrellisNet main network and Physical Information Neural Network (PINN); The TrellisNet main network is used to perform time-series modeling of the input marine environmental parameters and electromagnetic wave propagation parameters. The gating weights of the TrellisNet main network are initialized according to the correlation between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph. The gating weights are used to ensure that the TrellisNet main network retains marine environmental elements with a correlation greater than a preset threshold with electromagnetic wave propagation during the time-series modeling process, so as to capture the long-range dependence characteristics of the target electromagnetic wave during propagation. The physical information neural network PINN is constructed based on the partial differential equations corresponding to the physical constraint model. It is used to calculate the residuals of the electromagnetic wave propagation prediction data in the partial differential equations, and the residuals are used to characterize the physical loss.

6. The method according to claim 4, characterized in that, After obtaining the trained electromagnetic wave prediction model, the method further includes: Acquire newly added historical marine environmental parameter information and corresponding new sample initial propagation parameters and new sample target propagation parameters to form new training sample pairs; A bidirectional long short-term memory network is used to perform bidirectional time series modeling on the newly added historical marine environmental parameter information in order to extract the dynamic representation of the marine environmental elements. Based on the dynamic characterization, the new initial propagation parameters of the sample, and the new target propagation parameters of the sample, the causal intensity of each marine environmental element on the electromagnetic wave propagation characteristics is calculated. Based on the causal strength, update the correlation between marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph; Based on the correlation between marine environmental elements and electromagnetic wave propagation characteristics in the updated knowledge graph, the gating weights of the TrellisNet main network are adjusted. Using the new training sample pairs as fine-tuning samples, the electromagnetic wave prediction model with adjusted gating weights is retrained until the model converges, resulting in the optimized electromagnetic wave prediction model.

7. The method according to claim 6, characterized in that, The step of calculating the causal intensity of each marine environmental element on the electromagnetic wave propagation characteristics based on the dynamic characterization, the new initial propagation parameters of the sample, and the new target propagation parameters of the sample includes: Based on the dynamic characterization, the new initial propagation parameters of the sample, and the new target propagation parameters of the sample, the Liang's information flow method is used to calculate the amount of information transfer from various marine environmental elements to electromagnetic wave propagation characteristics. Based on the magnitude and sign of the information transfer, determine the causal intensity and causal direction of each marine environmental element on the electromagnetic wave propagation characteristics; The causal strength is used as the association weight to update the directed association between the corresponding marine environmental elements and electromagnetic wave propagation characteristics in the knowledge graph.

8. An electronic device, characterized in that, include: The system includes a memory, a processor, and a communication interface; wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform a propagation prediction method for ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor of an electronic device, causes the processor to perform a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, include: A computer program or instruction, when executed by a processor of an electronic device, causes the processor to perform a method for predicting the propagation of ultra-low frequency / extremely low frequency electromagnetic waves in a marine environment as described in any one of claims 1 to 7.

Images