Sea level prediction model training method, sea level prediction method and related products

Abstract

The invention discloses a training method of a sea level prediction model, a sea level prediction method and a related product, and relates to the technical field of marine environment monitoring, and the training method of the sea level prediction model comprises the steps: obtaining training data; wherein the acceleration measurement original data matrix is obtained by a plurality of MEMS acceleration sensors deployed on a plurality of submarine cables; performing variational mode decomposition on each time sequence in the acceleration measurement original data matrix to obtain a sub-sequence data matrix; and training a deep learning model by taking the subsequence data matrix, the wind speed data and the atmospheric pressure data as inputs and the sea level measurement data as an output to obtain a sea level prediction model. According to the application, the cost is greatly reduced by acquiring the data through the MEMS sensing array, and the vibration data is processed in combination with an artificial intelligence method through a laying mode attached to the submarine cable, so that large-range, high-precision and interpretable sea level state estimation and short-term trend prediction are realized.

Description

Technical Field

[0001] This application relates to the technical field of marine environmental monitoring, and in particular to a method for training a sea level prediction model, a sea level prediction method, and related products. Background Art

[0002] Traditional sea level observations mainly rely on tide gauges and satellite altimetry. However, these methods have some limitations. Although tide gauges can provide data with high time resolution, their distribution is often sparse and uneven. Due to the harsh marine environment and high deployment and maintenance costs, the spatial coverage of tide gauges is limited. This restricts their application in large-scale sea level monitoring. To solve the above problems, Microelectromechanical Systems (MEMS) sensors have received extensive attention in the field of marine observation due to their advantages such as small size, high integration, low power consumption, low cost, and high sensitivity. In addition, existing sea level prediction methods often rely on complex data analysis techniques. Although these methods have relatively high prediction accuracy, they often lack interpretability and are difficult to provide an intuitive explanation of the sea level change mechanism for decision-makers. In summary, the existing technologies still have the following problems in sea level monitoring and prediction: the traditional observation methods have limited spatial coverage and are difficult to achieve large-scale and high-precision sea level monitoring; the existing applications of MEMS sensors mainly focus on single-point measurement, and the advantages of sensor networks have not been fully utilized; the sea level prediction methods lack interpretability and are difficult to provide intuitive support for decision-making. Summary of the Invention

[0003] The purpose of this application is to provide a method for training a sea level prediction model, a sea level prediction method, and related products, which can achieve low-cost large-scale sea level monitoring and prediction.

[0004] To achieve the above purpose, this application provides the following solutions: In the first aspect, this application provides a method for training a sea level prediction model, and the method for training the sea level prediction model includes: Obtain training data; the training data includes: an original acceleration measurement data matrix, wind speed data, atmospheric pressure data, and sea level measurement data; the original acceleration measurement data matrix is obtained by a number of MEMS acceleration sensors deployed on a number of submarine cables; the original acceleration measurement data matrix includes: time series of a number of measurement points; Perform variational mode decomposition on each time series in the original acceleration measurement data matrix to obtain a subsequence data matrix; the subsequence data matrix includes k subsequences of each time series; Taking the subsequence data matrix, the wind speed data, and the atmospheric pressure data as inputs and the sea level measurement data as the output, training a deep learning model to obtain a sea level prediction model.

[0005] Optionally, taking the subsequence data matrix, the wind speed data, and the atmospheric pressure data as inputs and the sea level measurement data as the output, training a deep learning model to obtain a sea level prediction model, specifically including: Using the method of principal component analysis, extracting m principal components containing a preset percentage of variance from the subsequence data matrix to obtain an acceleration feature matrix; Taking the acceleration feature matrix, the wind speed data, and the atmospheric pressure data as inputs and the sea level measurement data as the output, training a deep learning model to obtain a sea level prediction model.

[0006] Optionally, the preset percentage is 95%.

[0007] Optionally, the deep learning model is a TFT model.

[0008] Optionally, the loss function during the training of the sea level prediction model is the root mean square error.

[0009] Optionally, the obtaining of the training data specifically includes: Obtaining wind speed data, atmospheric pressure data, and sea level measurement data; Obtaining an original acceleration measurement data matrix; the original acceleration measurement data matrix is obtained by a MEMS acceleration sensor; the MEMS acceleration sensor is deployed on the submarine cable in the offshore landing section; several MEMS acceleration sensors are deployed on each submarine cable.

[0010] In a second aspect, the present application provides a sea level prediction method, and the sea level prediction method includes: Obtaining data to be predicted; the data to be predicted includes: an original data matrix of acceleration measurement to be predicted, wind speed data to be predicted, and atmospheric pressure data to be predicted; the original data matrix of acceleration measurement to be predicted is obtained by several MEMS acceleration sensors deployed on several submarine cables; the original data matrix of acceleration measurement to be predicted includes: time series to be predicted at several measurement points; Performing variational mode decomposition on each time series to be predicted in the original data matrix of acceleration measurement to be predicted to obtain a subsequence data matrix to be predicted; the subsequence data matrix to be predicted includes q subsequences of each time series to be predicted; Input the to-be-predicted subsequence data matrix, the to-be-predicted wind speed data, and the to-be-predicted atmospheric pressure data into the sea level prediction model to obtain a sea level prediction result; the sea level prediction model is trained by the training method of the sea level prediction model described in any one of the above.

[0011] In a third aspect, the present application provides a sea level prediction device, which specifically includes: a control system and a plurality of MEMS acceleration sensors; The MEMS acceleration sensors are deployed on the undersea cables in the offshore landing section; a plurality of MEMS acceleration sensors are deployed on each undersea cable; The MEMS acceleration sensors are used to obtain an acceleration measurement raw data matrix; The control system is used for: Obtain to-be-predicted data; the to-be-predicted data includes: a to-be-predicted acceleration measurement raw data matrix, to-be-predicted wind speed data, and to-be-predicted atmospheric pressure data; the to-be-predicted acceleration measurement raw data matrix is obtained by a plurality of MEMS acceleration sensors deployed on a plurality of undersea cables; the to-be-predicted acceleration measurement raw data matrix includes to-be-predicted time series of a plurality of measurement points; Perform variational mode decomposition on each to-be-predicted time series in the to-be-predicted acceleration measurement raw data matrix to obtain a to-be-predicted subsequence data matrix; the to-be-predicted subsequence data matrix includes q subsequences of each to-be-predicted time series; Input the to-be-predicted subsequence data matrix, the to-be-predicted wind speed data, and the to-be-predicted atmospheric pressure data into the sea level prediction model to obtain a sea level prediction result; the sea level prediction model is trained by the training method of the sea level prediction model described in any one of the above.

[0012] In a fourth aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the training method or the prediction method of the sea level prediction model described in any one of the above.

[0013] In a fifth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the training method or the prediction method of the sea level prediction model described in any one of the above.

[0014] In a sixth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the training method or the prediction method of the sea level prediction model described in any one of the above.

[0015] According to the specific embodiments provided in this application, the following technical effects are disclosed in this application: This application provides a training method for a sea level prediction model, a sea level prediction method, and related products. The training method for the sea level prediction model includes: obtaining training data; the training data includes: an original acceleration measurement data matrix, wind speed data, atmospheric pressure data, and sea level measurement data; the original acceleration measurement data matrix is obtained by a number of MEMS acceleration sensors deployed on a number of submarine cables; the original acceleration measurement data matrix includes: time series of a number of measurement points; performing variational mode decomposition on each time series in the original acceleration measurement data matrix to obtain a subsequence data matrix; the subsequence data matrix includes k subsequences of each time series; using the subsequence data matrix, the wind speed data, and the atmospheric pressure data as inputs, and the sea level measurement data as an output, training a deep learning model to obtain a sea level prediction model. Compared with traditional sea level observation methods, in this application, data is obtained through a MEMS sensor array, which greatly reduces costs. And in this application, by relying on the laying method of submarine cables, the batch setting range of the monitoring sensor array can be greatly increased. At the same time, a prediction method for sea level is realized by processing vibration data through artificial intelligence methods, so as to achieve large-scale, high-precision, and interpretable sea level state estimation and short-term trend prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is an application environment diagram of a training method for a sea level prediction model in an embodiment of this application; Figure 2 It is a flowchart of a training method for a sea level prediction model provided in an embodiment of this application; Figure 3 It is a schematic diagram of offshore coastal tide monitoring of a MEMS sensor array provided in an embodiment of this application; Figure 4 It is a flowchart of a method for predicting tides using array-type acceleration measurement data provided in another embodiment of this application; Figure 5 It is an overall implementation flowchart of a sea level prediction device provided in an embodiment of this application; Figure 6 It is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed implementation manners

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0019] The present application aims to solve the problems existing in the existing sea level monitoring technologies, such as limited spatial coverage, insufficient accuracy, and lack of interpretability of prediction results. Specifically, the present application provides a system for deploying an acceleration sensor array on a submarine cable to monitor marine environmental parameters, and at the same time, a method for processing vibration data by an artificial intelligence method and realizing sea level prediction, so as to achieve large-scale, high-precision, and interpretable sea level state estimation and short-term trend prediction.

[0020] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0021] The training method of the sea level prediction model provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 Figure 14. Among them, the terminal 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set separately, integrated on the server 104, placed in the cloud or on other servers. The terminal 102 can send the training data to be processed to the server 104. After receiving the training data to be processed, for the training data to be processed, the server 104 performs variational mode decomposition on each time series in the original acceleration measurement data matrix to obtain a subsequence data matrix; the subsequence data matrix includes k subsequences of each time series; Using the subsequence data matrix, the wind speed data, and the atmospheric pressure data as inputs and the sea level measurement data as outputs, a deep learning model is trained to obtain a sea level prediction model. The server 104 can feedback the obtained sea level prediction model to the terminal 102. In addition, in some embodiments, the training method of the sea level prediction model can also be implemented separately by the server 104 or the terminal 102. For example, the terminal 102 can directly process the training data to be processed, or the server 104 can obtain the training data to be processed from the data storage system and process the training data to be processed.

[0022] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0023] In an exemplary embodiment, as Figure 2 shown, a method for training a sea level prediction model is provided. This method is executed by a computer device, and specifically can be executed alone by a computer device such as a terminal or a server, or can be jointly executed by a terminal and a server. In the embodiments of the present application, taking this method applied to Figure 1 the server 104 in as an example for illustration, it includes the following steps S1 to S3. Among them: S1. Obtain training data; the training data includes: an acceleration measurement raw data matrix, wind speed data, atmospheric pressure data, and sea level measurement data; the acceleration measurement raw data matrix is obtained by a number of MEMS acceleration sensors deployed on several submarine cables; the acceleration measurement raw data matrix includes: time series of a number of measurement points.

[0024] In this embodiment, a number of MEMS acceleration sensors are deployed on the submarine cable in the offshore landing section to form a MEMS acceleration sensing array. The data is connected to the user computer in the nearby office building through a communication cable. The user adjusts the data acquisition frequency according to actual needs, and saves the acquired data to the computer in real time for sea level estimation and prediction. For reference, see Figure 3 .

[0025] S2. Perform variational mode decomposition on each time series in the acceleration measurement raw data matrix to obtain a subsequence data matrix; the subsequence data matrix includes k subsequences of each time series.

[0026] The acceleration measurement raw data matrix obtained in this embodiment is: ; n is the number of measurement points. Among them, represents the time series of the data measured by the MEMS acceleration sensor at the i-th measurement point.

[0027] The optional decomposition process in this embodiment is as follows: For the acceleration measurement raw data matrix A in A iPerform variational mode decomposition. The main idea behind variational mode decomposition (VMD) is to solve an optimization problem by finding modes that minimize the mutual information between the decomposition components. This algorithm can decompose non-stationary acceleration measurement signals into a series of sub-modes, with advantages such as being fast, simple when dealing with data noise, and having a robust decomposition level. Decompose the original acceleration signal into multiple sub-modes , also known as intrinsic mode functions (IMFs).

[0028] ; (1) where is the original signal at time t, is the mode function, K is the number of sub-modes. Each sub-mode has a central frequency . To evaluate the bandwidth of the mode, the constrained variational problem takes the minimum sum of the frequency bandwidths of each sub-mode as the objective function, and the mathematical description is as follows: ; (2)

[0029] where is the Hilbert transform of , represents the partial derivative with respect to time t, is the delta function, j represents the imaginary unit, is the central frequency of each mode, f is the original signal. To minimize Equation (5), the augmented Lagrangian function is solved by introducing a quadratic penalty term and a Lagrange multiplier : ; (3) where is the penalty function, is the Lagrange multiplier. The saddle point of the Lagrangian function can be achieved by continuously updating the central frequency and bandwidth of each IMF through the alternating direction method of multipliers (ADMM). During the iteration process, the central frequency of each is re-estimated by the gravity of the power spectrum. The update process of the frequency and bandwidth can be expressed as (4) (5) where , and represent , and The Fourier transform of denotes has undergone update iterations, where denotes the number of update iterations. is the sum of the first mode functions, is the Fourier transform of denotes the continuous variable in the frequency domain, d denotes the integral of When the relative error is less than the convergence tolerance , the VMD process stops: ; (6) denotes has undergone update iterations, decomposed by the VMD algorithm, and k subsequences of the data at each measurement point are obtained, forming the subsequence data matrix ; denotes the data matrix formed by the j-th subsequence of each of the said measurement points.

[0030] To eliminate the correlation between data, reduce redundancy and noise, principal component analysis is performed on the said subsequence data matrix to transform it into a new acceleration vector (principal component). Principal component analysis uses orthogonal transformation to perform a linear transformation on the observed values of a series of possibly correlated variables, and thus projects them into the values of a series of linearly uncorrelated variables. The main steps include standardizing the data, calculating the covariance matrix, eigenvalue decomposition, selecting the principal components, and transforming the data.

[0031] The optional standardization formula in this embodiment is as follows: ; (6) ; (7) ; (8) In the formula, is the data of the j-th subsequence corresponding to the i-th measurement point, is the element in the standardized data matrix, is the mean of the j-th feature, is the standard deviation of the j-th feature, i = , n is the number of measurement points. The linear relationship between different features is represented by the covariance matrix: ; (9) In the formula, is the transpose matrix of the standardized data matrix, and Z is the standardized data matrix. Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues and eigenvectors: ; (10) wherein is the i-th eigenvalue of the covariance matrix C, is the i-th eigenvector of the covariance matrix C. Extract m principal components that contain 95% of the variance of the data in the subsequence data matrix, and obtain the acceleration feature matrix .

[0032] S3. Use the subsequence data matrix, the wind speed data, and the atmospheric pressure data as inputs, and the sea level measurement data as the output to train a deep learning model to obtain a sea level prediction model.

[0033] Then, verify the trained sea level prediction model according to the determination coefficient and root mean square error between the sea level prediction result and the actual sea level measurement data.

[0034] In this embodiment, obtaining data through the MEMS sensing array greatly reduces the cost. And in this application, by relying on the laying method of submarine cables, the batch setting range of the monitoring sensing array can be greatly increased. At the same time, the prediction method of the sea level is realized by processing vibration data through artificial intelligence methods to achieve large-range, high-precision, and interpretable sea level state estimation and short-term trend prediction.

[0035] In an exemplary embodiment, refer to Figure 4 , a sea level prediction method is provided. This method is executed by a computer device, and specifically can be executed alone by a computer device such as a terminal or a server, or can be jointly executed by a terminal and a server. In the embodiments of this application, taking this method applied to Figure 1 the server 104 as an example for illustration, it includes the following steps A1 to A3. Among them, it can also be called an array-type acceleration measurement data prediction tide method, and specifically includes the following steps: A1. Obtain the data to be predicted; the data to be predicted includes: the original data matrix of the acceleration measurement to be predicted, the wind speed data to be predicted, and the atmospheric pressure data to be predicted; the original data matrix of the acceleration measurement to be predicted is obtained by a number of MEMS acceleration sensors deployed on a number of submarine cables; the original data matrix of the acceleration measurement to be predicted includes: the time series to be predicted at a number of measurement points. A2. Perform variational mode decomposition on each time series to be predicted in the original data matrix of the acceleration measurement to be predicted to obtain a subsequence data matrix to be predicted; the subsequence data matrix to be predicted includes q subsequences of each time series to be predicted. A3. Input the to-be-predicted subsequence data matrix, the to-be-predicted wind speed data, and the to-be-predicted atmospheric pressure data into the sea level prediction model to obtain the sea level prediction result; the sea level prediction model is obtained by training with the training method of the sea level prediction model described above.

[0036] Among them, step A3 specifically includes: using the method of principal component analysis to extract m principal components containing 95% variance from the to-be-predicted subsequence data matrix to obtain the to-be-predicted acceleration feature matrix.

[0037] Input the to-be-predicted acceleration feature matrix, the to-be-predicted wind speed data, and the to-be-predicted atmospheric pressure data into a pre-trained time series model (Temporal Fusion Transformer, TFT) model (i.e., the sea level prediction model in the above text) to obtain the sea level prediction result.

[0038] Or, simply speaking, the training of the TFT model in this embodiment can also be briefly summarized as the following steps: (1) Prepare the training data set, including the historical acceleration feature matrix, the historical wind speed and atmospheric pressure information in the monitoring area, and the corresponding historical sea level data; (2) Input the training data set into the TFT model for training to optimize the model parameters; (3) Use the validation data set to evaluate the model performance and adjust the hyperparameters.

[0039] Based on the same inventive concept, the embodiment of the present application also provides a sea level prediction device for implementing the above-mentioned sea level prediction method. The implementation solution provided by this device to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more of the following sea level prediction device embodiments can refer to the limitations on the sea level prediction method in the above text and will not be repeated here.

[0040] In an exemplary embodiment, a sea level prediction device is provided. The overall implementation process of this device is as Figure 5 shown. The sea level prediction device specifically includes: a control system and several MEMS acceleration sensors.

[0041] The MEMS acceleration sensors are deployed on the submarine cable in the offshore landing section; several MEMS acceleration sensors are deployed on each submarine cable.

[0042] The MEMS acceleration sensors are used to obtain the raw acceleration measurement data matrix.

[0043] The control system is used for: Obtain the data to be predicted; the data to be predicted includes: the original data matrix of the acceleration measurement to be predicted, the wind speed data to be predicted, and the atmospheric pressure data to be predicted; the original data matrix of the acceleration measurement to be predicted is obtained by a number of MEMS acceleration sensors deployed on a number of submarine cables; the original data matrix of the acceleration measurement to be predicted includes: the time series to be predicted at a number of measurement points. Perform variational mode decomposition on each time series to be predicted in the original data matrix of the acceleration measurement to be predicted, and obtain the data matrix of the subsequences to be predicted; the data matrix of the subsequences to be predicted includes q subsequences of each time series to be predicted. Input the data matrix of the subsequences to be predicted, the wind speed data to be predicted, and the atmospheric pressure data to be predicted into the sea level prediction model to obtain the sea level prediction result; the sea level prediction model is trained by the training method of the sea level prediction model described above.

[0044] The sea level prediction method and system based on the acceleration sensor array provided by the present application have the following beneficial effects: 1. Improve the monitoring range: By deploying a low-cost MEMS accelerometer array on submarine cables, a large-scale and high-density sea level monitoring network is realized, significantly expanding the monitoring coverage.

[0045] 2. Improve the prediction accuracy: The variational mode decomposition and principal component analysis are used to preprocess the original acceleration data, effectively extracting key features and reducing the influence of noise. Combining the advantages of the TFT model, this method has achieved high accuracy in sea level prediction. According to the experimental data, the determination coefficient between the prediction result of this method and the actual sea level measurement value reaches 0.97, and the root mean square error is as low as 1.98 cm.

[0046] Among them, to achieve an interpretable and robust estimation and prediction of the sea level, the present application adopts Temporal fusion transformers (TFT) that combines the multi-head attention of Transformers and the recursive idea of RNN. The model architecture of TFT includes five important components, namely the gating mechanism, the variable selection network, the static covariate encoder, the temporal processing, and the prediction interval. Using the interpretable multi-head attention mechanism and the variable selection network (VSN) to evaluate the importance of the known future (weather and time data) and the observed input (the acceleration feature vector processed by PCA-VMD), the TFT model can maintain a high degree of interpretability while achieving excellent prediction performance.

[0047] The TFT can process multiple relevant variables simultaneously. This is very important in practical applications because many time series problems involve multiple factors that influence each other. For example, when predicting sea levels, multiple variables such as solar and lunar changes, wind force, and the atmosphere may need to be considered. By processing these variables simultaneously, the TFT can better capture the complex relationships between the variables, thereby improving the prediction accuracy.

[0048] The TFT can effectively capture long-term and short-term dependencies in time series. For some time series problems, both long-term trends and short-term fluctuations have important impacts on the prediction results. For example, in sea level prediction, both long-term climate patterns and short-term meteorological changes need to be considered. By using different time windows and attention mechanisms, the TFT can model long-term and short-term time dependencies simultaneously, improving the prediction accuracy.

[0049] 3. Enhanced interpretability: Through the attention mechanism and variable importance analysis of the TFT model, this method can explain the contributions of different input features to the prediction results, providing an intuitive explanation of the sea level change mechanism for decision-makers.

[0050] Among them, the specific interpretation methods include: ① Visualization of the attention mechanism: The attention mechanism in the TFT can provide a certain degree of interpretability. By visualizing the attention weights, it is possible to understand the focus of attention at different time steps and variables during the sea level prediction process. This helps managers understand the decision-making process of the model, discover key factors in the data, and thus better adjust and optimize the model.

[0051] ② Feature importance analysis: The TFT can analyze the contribution degree of different variables to the prediction results. This helps to determine which variables are the most important for the prediction results, providing a basis for decision-making. For example, in sea level tide prediction, by analyzing the influence of different signal frequencies on the sea level prediction results, the signal frequency allocation can be optimized to improve the sea level prediction accuracy and structure.

[0052] 4. Cost reduction: Compared with traditional tide gauges and satellite altimetry techniques, the cost of the MEMS accelerometers used in this method is significantly reduced. According to market research, the cost of a single MEMS accelerometer is only $1 - $100, while the cost of traditional tide gauges is between $5000 - $50000.

[0053] 5. Improved system reliability: Through array deployment, even if individual sensors fail, the system can still operate normally, improving the reliability of the overall monitoring system.

[0054] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be asFigure 6 As shown in the figure. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for training a sea level prediction model or a sea level prediction method.

[0055] Those skilled in the art can understand that Figure 6 the structure shown in the figure is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0056] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0057] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0058] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0059] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0060] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0061] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0062] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0063] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for training a sea level prediction model, characterized in that: The training method of the sea level prediction model includes: Acquire training data; the training data includes: acceleration measurement raw data matrix, wind speed data, atmospheric pressure data and sea level measurement data; the acceleration measurement raw data matrix is ​​acquired by a plurality of MEMS acceleration sensors deployed on a plurality of submarine cables; the acceleration measurement raw data matrix includes: a time series of a plurality of measurement points; Performing variational mode decomposition on each time series in the acceleration measurement original data matrix to obtain a subsequence data matrix; the subsequence data matrix includes k subsequences of each time series; The subsequence data matrix, the wind speed data and the atmospheric pressure data are used as inputs and the sea level measurement data is used as output to train a deep learning model and obtain a sea level prediction model.

2. The sea level prediction model training method according to claim 1, characterized in that: Taking the subsequence data matrix, the wind speed data and the atmospheric pressure data as input and the sea level measurement data as output, training a deep learning model to obtain a sea level prediction model specifically includes: Using the principal component analysis method, m principal components containing a preset percentage variance are extracted from the subsequence data matrix to obtain an acceleration feature matrix; The acceleration feature matrix, the wind speed data and the atmospheric pressure data are used as inputs and the sea level measurement data is used as output to train a deep learning model and obtain a sea level prediction model.

3. The sea level prediction model training method according to claim 2, characterized in that: The preset percentage is 95%.

4. The method for training a sea level prediction model according to claim 1, characterized in that: The deep learning model is a TFT model.

5. The method for training a sea level prediction model according to claim 1, characterized in that: The loss function during the sea level prediction model training is the root mean square error.

6. A sea level prediction method, characterized in that: The sea level prediction method comprises: Acquire the data to be predicted; the data to be predicted includes: a matrix of raw acceleration measurement data to be predicted, wind speed data to be predicted and atmospheric pressure data to be predicted; the matrix of raw acceleration measurement data to be predicted is obtained by a plurality of MEMS acceleration sensors deployed on a plurality of submarine cables; the matrix of raw acceleration measurement data to be predicted includes: a time series to be predicted of a plurality of measurement points; Performing variational mode decomposition on each time series to be predicted in the original data matrix of acceleration measurement to be predicted to obtain a subsequence data matrix to be predicted; the subsequence data matrix to be predicted includes q subsequences of each time series to be predicted; The subsequence data matrix to be predicted, the wind speed data to be predicted and the atmospheric pressure data to be predicted are input into a sea level prediction model to obtain a sea level prediction result; the sea level prediction model is trained by the sea level prediction model training method described in any one of claims 1 to 5.

7. A sea level prediction device, characterized in that: The sea level prediction device specifically includes: a control system and a plurality of MEMS acceleration sensors; The MEMS acceleration sensor is deployed on the submarine cable at the landing section in the offshore waters; a plurality of MEMS acceleration sensors are deployed on each submarine cable; The MEMS acceleration sensor is used to obtain the acceleration measurement raw data matrix; The control system is used for: Acquire the data to be predicted; the data to be predicted includes: a matrix of raw acceleration measurement data to be predicted, wind speed data to be predicted and atmospheric pressure data to be predicted; the matrix of raw acceleration measurement data to be predicted is obtained by a plurality of MEMS acceleration sensors deployed on a plurality of submarine cables; the matrix of raw acceleration measurement data to be predicted includes: a time series to be predicted of a plurality of measurement points; Performing variational mode decomposition on each time series to be predicted in the original data matrix of acceleration measurement to be predicted to obtain a subsequence data matrix to be predicted; the subsequence data matrix to be predicted includes q subsequences of each time series to be predicted; The subsequence data matrix to be predicted, the wind speed data to be predicted and the atmospheric pressure data to be predicted are input into a sea level prediction model to obtain a sea level prediction result; the sea level prediction model is trained by the sea level prediction model training method described in any one of claims 1 to 5.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the sea level prediction model training method described in any one of claims 1 to 5 or the sea level prediction method described in claim 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the training method of the sea level prediction model described in any one of claims 1 to 5 or the sea level prediction method described in claim 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the training method of the sea level prediction model described in any one of claims 1 to 5 or the sea level prediction method described in claim 6 is implemented.

Images