A method and device for simulating ocean tide level based on hybrid neural network model

Abstract

The present disclosure relates to the technical field of ocean numerical simulation, and specifically to a method and device for simulating ocean tides based on a hybrid neural network model. The method comprises: obtaining the harmonization parameters on the open boundary of the target sea area, inputting the pre-trained physical information neural network model in the hybrid neural network model, and obtaining the simulated tide time series data of multiple tide-observing stations; constructing a training data set, inputting the training data set into the deep neural network model in the hybrid neural network model, and training the deep neural network model; inputting the measured tide time series data of multiple tide-observing stations into the trained deep neural network model, and obtaining the optimal harmonization parameters of multiple nodes from the deep neural network model. The method provided in the embodiment of the present disclosure obtains the optimal harmonization parameters by constructing and training the deep neural network model and directly inputting the measured tide data of the tide-observing stations into the hybrid neural network model, thereby reducing the manpower and time costs in the numerical simulation process and improving the efficiency of the numerical simulation.

Description

Technical Field

[0001] The present disclosure relates to the field of ocean numerical simulation, and in particular to a method and device for simulating ocean tides based on a hybrid neural network model. Background Art

[0002] In numerical simulations of meteorology, oceanography, and other fields, with the increasing demands for simulation accuracy and resolution, traditional global simulation methods are no longer able to meet the requirements for high-precision local simulations. Therefore, high-resolution local simulations are commonly used. Local simulations require "open boundaries" at regional boundaries and provide boundary-driven data that ensures the closure of the governing equations. However, these data often require correction to meet the accuracy requirements of the numerical model due to insufficient spatial resolution or large observation errors.

[0003] However, the processing of boundary-driven data in related technologies often relies on manual labor. This process is highly dependent on the operator's professional knowledge and parameter adjustment experience, resulting in high labor and time costs, and thus low numerical simulation efficiency. Therefore, how to reduce the labor and time costs of numerical simulation and improve its efficiency has become a problem that needs to be solved. Summary of the Invention

[0004] In view of this, the present disclosure provides an ocean tide simulation method and device based on a hybrid neural network model to solve the problem of how to reduce the manpower and time cost in the numerical simulation process and improve the efficiency of numerical simulation.

[0005] On the one hand, the present disclosure provides a method for simulating ocean tides based on a hybrid neural network model, the method comprising: obtaining harmonic parameters corresponding to a plurality of nodes on an open boundary of a target sea area, inputting the harmonic parameters into a pre-trained physical information neural network model in the hybrid neural network model, and obtaining simulated tide time series data corresponding to a plurality of tide-observing stations; constructing a training data set based on the harmonic parameters and the simulated tide time series data, inputting the training data set into a deep neural network model in the hybrid neural network model, and training the deep neural network model so that the deep neural network model learns the inverse mapping relationship from the simulated tide time series data to the harmonic parameters; inputting the measured tide time series data of a plurality of tide-observing stations into the trained deep neural network model, and obtaining the optimal harmonic parameters of a plurality of nodes from the deep neural network model.

[0006] On the other hand, the present disclosure also provides an ocean tide simulation device based on a hybrid neural network model, which includes: a data driving module for obtaining the harmonic parameters corresponding to multiple nodes on the open boundary of the target sea area, inputting the harmonic parameters into the pre-trained physical information neural network model in the hybrid neural network model, and obtaining the simulated tide time series data corresponding to multiple tide-watching stations; a model training module for constructing a training data set based on the harmonic parameters and the simulated tide time series data, inputting the training data set into the deep neural network model in the hybrid neural network model, and training the deep neural network model so that the deep neural network model learns the inverse mapping relationship from the simulated tide time series data to the harmonic parameters; an optimal harmonic parameter acquisition module for inputting the measured tide time series data of multiple tide-watching stations into the trained deep neural network model, and obtaining the optimal harmonic parameters of multiple nodes from the deep neural network model.

[0007] On the other hand, the present disclosure also provides an electronic device, including: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the above-mentioned ocean tide simulation method based on the hybrid neural network model by executing the computer instructions.

[0008] On the other hand, the present disclosure further provides a computer-readable storage medium having computer instructions stored thereon, and the computer instructions are used to enable a computer to implement the above-mentioned ocean tide simulation method based on the hybrid neural network model.

[0009] On the other hand, the present disclosure further provides a computer program product, including computer instructions, which are used to enable a computer to execute the above-mentioned ocean tide simulation method based on the hybrid neural network model.

[0010] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiment of the present invention, by constructing and training a deep neural network model, it has the ability to reversely infer the harmonic parameters of each node at the boundary based on the tide level data measured by multiple tide observation stations. By directly inputting the tide level data measured by the tide observation stations into the hybrid neural network model, the corresponding optimal harmonic parameters are directly obtained, thereby avoiding the traditional iterative adjustment process of multiple experimental simulations relying on manual experience, reducing the manpower and time costs in the numerical simulation process, and improving the efficiency of numerical simulation.

[0011] In addition, compared with the related art method of repeatedly running the numerical model to repeatedly verify the boundary parameters, the above-mentioned embodiment of the present disclosure obtains the optimal harmonic parameters at one time through model training and back-inference mechanism, effectively reducing the number of repeated simulations and resource consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the related technologies, the following briefly introduces the drawings required for use in the specific embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0013] Figure 1 1 is a flow chart of a method for simulating ocean tides based on a hybrid neural network model provided by an embodiment of the present disclosure;

[0014] Figure 2 This is a schematic diagram of a training data set construction process for an ocean tide simulation method based on a hybrid neural network model provided by an embodiment of the present disclosure;

[0015] Figure 3 Schematic diagram of a DNN model structure of an ocean tide simulation method based on a hybrid neural network model provided by an embodiment of the present disclosure;

[0016] Figure 4 Schematic diagram of the structure of an ocean tide simulation device based on a hybrid neural network model provided by an embodiment of the present disclosure;

[0017] Figure 5 It is a structural schematic diagram of another ocean tide simulation device based on a hybrid neural network model provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] In numerical simulation fields such as meteorology and oceanography, high-resolution models are often used to fine-tune simulations of localized key areas to balance computational resources and simulation accuracy. Such localized simulations require artificially defined boundaries within the computational region, known as open boundaries. Because open boundaries are not physical boundaries and cannot naturally close the model's governing equations, additional boundary-driven data (such as tide levels or current velocities) are required to ensure model solvability. The current mainstream approach typically relies on global ocean observations or model data (such as satellite data and global tidal databases) to obtain tidal information. These data are then interpolated at the open boundary locations of the target region to construct tidal time series data at the boundary. This process involves multiple steps, including data download, format conversion, position matching, interpolation, and accuracy correction. Due to instrumental errors and resolution limitations in the original observational data, significant errors are often introduced when projecting them onto the local area. To improve simulation accuracy, secondary processing and manual correction of the interpolated boundary data are often required. On the other hand, ocean dynamics simulations often assume shallow water, constructing tidal models based on shallow water equations as the governing equations. Open boundary conditions are often generated using astronomical tidal harmonics methods, taking the form of the amplitude and phase superposition of multiple specific frequency terms. This method requires obtaining the harmonic parameters (amplitude and phase) corresponding to each astronomical tidal component and inputting them into the model to drive the simulation. After the simulation is completed, the error between the simulation results and the measured data is compared. The simulation process is then repeated with manual adjustment of the parameters to ultimately obtain the harmonic parameter combination with the smallest error.

[0019] However, in related technologies, the ocean boundary construction method still faces the following problems:

[0020] 1. The entire boundary data preparation and correction process is highly dependent on manual operation and expertise. Harmony parameters must be manually downloaded, screened, and matched, longitude and latitude interpolation must be set based on the characteristics of the simulation area, and error correction relies on the operator's understanding and experience of the main tidal components.

[0021] 2. The boundary data optimization process requires repeated ocean numerical simulations. After each round of adjustment, the model must be fully run to verify the error. This has high computational costs and is time-consuming, making it difficult to efficiently complete the search for optimal parameters.

[0022] 3. Since nearshore simulations usually include eight astronomical tidal components, each of which contains two adjustable parameters, amplitude and phase, with a parameter dimension of up to 16 dimensions, only individual parameters of the main tidal components can be roughly adjusted during manual adjustment, and it is almost impossible to achieve global joint optimization of multiple parameters, which limits the simulation accuracy and model generalization ability.

[0023] To solve the above problems, various embodiments of the present disclosure provide an ocean tide simulation method based on a hybrid neural network model, the method comprising: obtaining harmonic parameters corresponding to multiple nodes on the open boundary of the target sea area, inputting the harmonic parameters into a pre-trained physical information neural network model in the hybrid neural network model, and obtaining simulated tide time series data corresponding to multiple tide-observing stations; constructing a training data set based on the harmonic parameters and the simulated tide time series data, inputting the training data set into a deep neural network model in the hybrid neural network model, and training the deep neural network model so that the deep neural network model learns the inverse mapping relationship from the simulated tide time series data to the harmonic parameters; inputting the measured tide time series data of multiple tide-observing stations into the trained deep neural network model, and obtaining the optimal harmonic parameters of multiple nodes from the deep neural network model.

[0024] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present disclosure.

[0025] Please refer to Figure 1 , Figure 1 : This is a flow chart of a method for simulating ocean tides based on a hybrid neural network model provided by an embodiment of the present disclosure. The method may include the following steps:

[0026] Step S101, obtain the harmonic parameters corresponding to multiple nodes on the open boundary of the target sea area, input the harmonic parameters into the pre-trained physical information neural network model in the hybrid neural network model, and obtain the simulated tide level time series data corresponding to multiple tide-observing stations.

[0027] In this embodiment, the target sea area can refer to a coastal region or local sea area with a limited spatial range selected to meet research and detection requirements. The target sea area has a clear computational boundary in the numerical simulation, and the unenclosed part is an open boundary, which requires the introduction of external tidal level driving to achieve simulation closure.

[0028] For example, the target sea area may include, but is not limited to, a nearshore area, an offshore area, or a bay area.

[0029] Furthermore, the nodes on the open boundary of the target sea area may refer to a series of discrete spatial points selected on the boundary of the simulation area for applying boundary conditions.

[0030] The harmonic parameters obtained from the multiple nodes may refer to physical quantities used to describe the components of the astronomical tide. The harmonic parameters may include the amplitude and phase of each astronomical tide.

[0031] Among them, the amplitude of the astronomical tide represents the maximum value of the tide level change caused by the tide at the node; the phase of the astronomical tide represents the angle value corresponding to the time difference when the maximum high tide occurs after the preset time.

[0032] Furthermore, the harmonic parameters corresponding to each of the multiple nodes are respectively input into the pre-trained physical information neural network model in the hybrid neural network model.

[0033] Among them, the Physics-Informed Neural Networks (PINN) model is a model that integrates physical laws and the data-driven capabilities of neural networks, and can be used to approximately solve partial differential equations of physical processes.

[0034] Specifically, the core idea of ​​the PINN model can be described as follows: construct a neural network to simulate the changes of physical quantities over time and space, introduce known physical control equations into the loss function, and fit data points through the neural network during training. At the same time, make the network output conform to the physical control equations to constrain the rationality and interpretability of the solution.

[0035] In the embodiment of the present disclosure, the purpose of pre-training the PINN model may be to enable the PINN model to learn the forward mapping relationship from the harmonic parameters to the simulated tide time series data.

[0036] Furthermore, based on the pre-trained PINN model, the simulated tide time series data corresponding to multiple tide-observing stations obtained may refer to the tide height sequence data changing with time at a specific location calculated by the PINN model.

[0037] Step S102: construct a training data set based on the harmonic parameters and the simulated tide time series data, input the training data set into the deep neural network model in the hybrid neural network model, and train the deep neural network model so that the deep neural network model learns the inverse mapping relationship from the simulated tide time series data to the harmonic parameters.

[0038] In this embodiment, the harmonic parameters are used as labels, the simulated tide time series data are used as input features, and a training data set is constructed based on the labels and the input features.

[0039] The training data set is input into the deep neural network model to train the deep neural network model.

[0040] Deep Neural Networks (DNNs) are artificial neural network structures composed of multiple layers of neurons. DNN models consist of an input layer, multiple hidden layers, and an output layer. Pre-training the DNN model allows it to learn the inverse mapping relationship from simulated tidal time series data to harmonic parameters. The trained DNN model uses this learned inverse mapping relationship to infer the corresponding optimal harmonic parameters based on the input tidal time series data.

[0041] In step S103, the measured tide level time series data of multiple tide-observing stations are input into the trained deep neural network model, and the optimal harmonic parameters of multiple nodes are obtained from the deep neural network model.

[0042] In this embodiment, measured tide level time series data of multiple tide-observing stations are obtained, and the measured tide level time series data are preprocessed so that the format of the measured tide level time series data is consistent with the format of the simulated tide level time series data.

[0043] The measured tide time series data are input into the trained DNN model, so that the DNN model outputs the optimal harmonic parameters of each node on the open boundary.

[0044] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiment of the present disclosure, by constructing and training a deep neural network model, it is able to reversely infer the harmonic parameters of each node at the boundary based on the measured tide data of multiple tide-observing stations, avoiding the traditional iterative adjustment process of multiple experimental simulations based on manual experience, reducing the manpower and time costs in the numerical simulation process, and improving the efficiency of numerical simulation. Compared with the method of repeatedly running the numerical model to repeatedly verify the boundary parameters in the related art, the above-mentioned embodiment of the present disclosure obtains the optimal harmonic parameters at one time through model training and reverse inference mechanism, effectively reducing the number of repeated simulations and resource consumption.

[0045] In a possible implementation of step S101, obtaining the harmonization parameters corresponding to the plurality of nodes on the open boundary of the target sea area includes:

[0046] Obtain N on the open boundary of the target sea area through the astronomical tide parameter random generator b Each node corresponds to N T Group harmonic parameters Each set of harmonic parameters includes the first of the eight astronomical tides. k The amplitude of the astronomical tide and phase .

[0047] In this embodiment, the astronomical tidal parameter random generator may refer to a tool for automatically generating multiple groups of astronomical tidal harmonic parameter combinations that meet statistical laws based on physical constraints and empirical distribution in a simulation environment.

[0048] Here, the astronomical tidal parameter random generator can generate physically reasonable harmonic parameter samples in batches when there is insufficient real observation data.

[0049] Furthermore, the astronomical tide parameter random generator can be The harmonic parameters are randomly generated within a range of .

[0050] in, Custom settings can be made based on the historical maximum tide level rise in the target sea area. For example, It can be set to 4 times the maximum tide level rise detected by the tide gauge station.

[0051] Furthermore, N is selected at the open boundary of the target sea area. b discrete nodes, for each node, the astronomical tide parameter random generator generates N T There are 8 sets of harmonic parameters, each of which includes the amplitudes and phases corresponding to 8 typical astronomical tides.

[0052] Here, the eight typical astronomical tidal components are S2, M2, N2, K2, K1, P1, O1, and Q1. Among them, S2 represents the fixed solar semidiurnal tidal component, M2 represents the main lunar semidiurnal tidal component, N2 represents the large elliptical lunar semidiurnal tidal component, K2 represents the lunar and solar combined semidiurnal tidal component, K1 represents the lunar and solar combined diurnal tidal component, P1 represents the fixed solar diurnal tidal component, O1 represents the main lunar diurnal tidal component, and Q1 represents the large elliptical lunar diurnal tidal component.

[0053] The ocean tide simulation method and apparatus based on a hybrid neural network model, as described in the above-mentioned embodiments of the present disclosure, expand the input space by generating multiple sets of harmonic parameter samples at multiple open-boundary nodes. This allows the PINN model to cover a wider range of harmonic parameter distributions during training, improving the model's generalization and adaptability. By generating harmonic parameters based on a preset maximum amplitude range and phase period range, non-physical simulation inputs can be effectively avoided, ensuring reliable tide simulation results.

[0054] In a possible implementation of the above embodiment, the pre-trained physical information neural network model is implemented based on the following steps:

[0055] Construct the input layer and output layer of the physical information neural network model. The input layer includes at least N b Each node corresponds to N T Group harmonic parameters The output layer includes at least N tide-watching stations for each of the M tide-watching stations. T Simulated tide level time series data ;

[0056] Based on the physical constraint term and the data-driven term, a composite loss function of the physical information neural network model is defined; wherein the physical constraint term uses the shallow water equation as the physical constraint;

[0057] The composite loss function is minimized through the gradient back-propagation algorithm. When the composite loss function is lower than the preset threshold, the training of the physical information neural network model is completed.

[0058] In this embodiment, For the j Group k Astronomical tides at nodes b The amplitude at For the j Group k Astronomical tides at nodes b The phase at . For the i Tide-viewing station locations ( ) in time Simulated tide level time series at .

[0059] Furthermore, during the pre-training process of the PINN model, the shallow water equation is introduced into the loss function of the PINN model, so that the prediction results of the PINN model not only fit the data but also satisfy the physical laws corresponding to the shallow water equation.

[0060] Here, the shallow water equations refer to a set of partial differential equations used to describe the flow behavior of water in the shallow layer of the earth's surface, which can be used to model tidal currents in shallow sea areas.

[0061] The shallow water equation is:

[0062]

[0063] The expressions of each term in the shallow water equation are as follows:

[0064] , F ,

[0065] in, is the tide level, and is the flow velocity component; Represents conserved variables, that is, physical quantities that need to evolve over time; F It represents the flux term, which describes the propagation of state quantity in space during the movement of water body; Represents the source term, which represents the internal changes of the system caused by non-conservative quantities, such as friction, Coriolis force, gravity flow caused by bottom slope, and external force. is the bottom slope elevation, is the total water depth, is the bottom friction coefficient, is the Coriolis force coefficient. At the open boundary , usually water level drive (velocity with zero gradient) is used as the open boundary condition .

[0066] Furthermore, physical constraints directly apply the governing physical equations as constraints on neural network training, ensuring that the network output not only fits the data but also satisfies the laws of physics. Data-driven terms are the loss functions of traditional supervised learning, designed to ensure that the neural network fits the existing observational data as closely as possible.

[0067] Defining a composite loss function of a physical information neural network model based on physical constraint terms and data-driven terms may include defining a composite loss function of a PINN model based on a weighted sum of physical constraint terms and data-driven terms.

[0068] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiment of the present invention, by utilizing the shallow water equation as a physical constraint term and the observation data driving term to jointly construct a composite loss function, the gradient back propagation algorithm is used to train the model, and the loss function is gradually minimized until the preset threshold is met. This can enable the PINN model to follow physical laws while ensuring the accuracy of data fitting, thereby improving the accuracy of data simulation.

[0069] In a possible implementation of the above embodiment, constructing the input layer and output layer of the physical information neural network model includes:

[0070] Construct the input layer of the physical information neural network model, which includes at least one of the following: any sampling coordinate and sampling time in the target sea area ( x, y, t ), N of each node on the open boundary T Group harmonic parameters; wherein the sampling coordinates in the target sea area include at least one of the following: nodes on the open boundary, tide-observing stations, and points inside the sea area;

[0071] Construct the output layer of the physical information neural network model, which includes at least one of the following: the simulated tide level of any sampling coordinate in the target sea area ζ ( x, y, t ), velocity component U ( x, y, t )and V ( x, y, t ), where the simulated tide level ζ ( x, y,t ), velocity component U ( x, y, t )and V ( x, y, t ) is used to construct the simulated tide time series data .

[0072] In this embodiment, the simulated tide level ζ ( x, y, t ) means that at coordinate ( x, y ) and time t The water surface height changes at . Velocity component U ( x, y, t )and V ( x, y, t ) can represent the water body in x and y The velocity component in the direction.

[0073] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiments of the present disclosure, by constructing the input and output layers of the PINN model, the model can output simulated tide level and flow velocity time series data for any point in the target sea area based on space-time coordinates and boundary reconciliation parameters, meeting the requirements of tidal physical process simulation. Based on the input of coordinates and time, the PINN model can output tidal information for any point in the target sea area, has excellent spatial and temporal interpolation capabilities, and supports high-quality simulation of unobserved points within the sea area.

[0074] In a possible implementation of the above embodiment, a composite loss function of the physical information neural network model is defined based on the physical constraint term and the data-driven term, including:

[0075] The residual term of the equation constructed based on the shallow water equation is used as a physical constraint term;

[0076] The boundary residual term constructed based on the difference between the measured tide level time series data and the simulated tide level time series data is used as the data-driven term;

[0077] The composite loss function of the physical information neural network model is defined as the weighted sum of the physical constraint terms and the data-driven terms.

[0078] In this embodiment, the preset weights in the composite loss function can prevent a certain item from dominating the training process due to its large or small value, thereby ensuring reasonable coordination between physical constraints and data fitting.

[0079] The ocean tide simulation method and apparatus based on a hybrid neural network model in the above-mentioned embodiments of the present disclosure introduce data-driven terms to constrain the error between the model output and the actual observed tide level, effectively improving the model's accuracy in fitting tide time series data and making the simulation results closer to actual observations. By weightedly fusing physical constraints and data-driven terms, the complementary nature of knowledge-driven and data-driven approaches is achieved.

[0080] In a possible implementation of the above embodiment, the residual term of the equation constructed based on the shallow water equation is:

[0081]

[0082]

[0083] in, is the vector of conserved variables of the shallow water equation; Simulating tide levels ζ ( x, y, t ) and velocity components U ( x, y, t )and V ( x, y, t ) constituted by the flux function; Simulating tide levels ζ ( x, y, t ) and velocity components U ( x, y, t )and V ( x, y, t ) is the source function composed of; is the residual of the physical information neural network model for the shallow water equation at the sampling coordinates; N is the sampling number of the sampling coordinates in the target sea area; is the set of parameters to be trained for the physical information neural network model.

[0084] In this embodiment, is the discrete residual form of the shallow water equation, which serves as the physical constraint embedded in the PNN model. Characterization at all sampling points The residual square of the shallow water equation is calculated and the average value is taken, so that the physical variables learned by the neural network in the training optimization process can satisfy the shallow water equation as much as possible.

[0085] Furthermore, the residual of the equation Indicates the degree to which the shallow water equation is not satisfied.

[0086] Furthermore, by minimizing The PINN model will be constrained to automatically satisfy the shallow water equations by generating physical prediction results, where the physical prediction results can be tide level and flow velocity.

[0087] For example, is the set of parameters to be trained for the physical information neural network model, It may include but is not limited to the weights between neurons in each layer and the bias items of each neuron.

[0088] The ocean tide simulation method and apparatus based on a hybrid neural network model, as described in the aforementioned embodiments of the present disclosure, utilizes the shallow water equation residual to constrain the neural network training process, ensuring that the tide level and flow velocity output by the model automatically adhere to physical laws, effectively improving the physical credibility of the simulation results. The equation residual term and the boundary residual term together form a composite loss function, effectively integrating physical priors with measured data characteristics, improving the overall learning effect and engineering practicality of the PINN model.

[0089] In a possible implementation of the above embodiment, the boundary residual term constructed based on the difference between the measured tide level time series data and the simulated tide level time series data is:

[0090]

[0091] in, Physical information neural network model at tide gauge station and time The simulated tide level output at At the tide gauge station and time The measured tide level at It represents the mean square error term between the simulated tide level and the measured tide level.

[0092] In this embodiment, a set of tide gauge stations in the target sea area is selected. , and multiple time points corresponding to each site , collect the measured tide level of each station at the corresponding time .

[0093] The spatial coordinates of each site and the time Input to the PINN model to obtain the simulated tide level at each station .

[0094] Determine the difference between simulated and measured tide levels , calculate the mean square error of each site as the boundary residual term .

[0095] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiment of the present invention, by constructing the boundary residual term, the error between the measured tide level and the simulated tide level is incorporated into the loss function, thereby enhancing the model's ability to fit the observation data, thereby improving the accuracy of the physical information neural network model in tide prediction.

[0096] In a possible implementation of the above embodiment, the composite loss function of the physical information neural network model is:

[0097]

[0098] in, is the preset weight coefficient.

[0099] In this embodiment, It can also refer to the open boundary learning rate, which can control the contribution rate of boundary observation error to the total loss in each round of training, thereby affecting the learning intensity of boundary information during training.

[0100] For example, when When it is too large, the model is more inclined to fit the boundary residual term to improve the accuracy of the model in the open boundary; When it is too small, the model is more inclined to fit the physical residual term to make the model more consistent with the physical equation.

[0101] In one possible implementation, during the training of the model, the bounded learning rate It can be a fixed constant or dynamically adjusted.

[0102] For example, bounding the learning rate The dynamic adjustment method may include at least one of the following: dynamic adjustment based on residual ratio, dynamic scaling adjustment, and staged adjustment.

[0103] Through the ocean tide simulation method and device based on the hybrid neural network model of the above embodiment of the present invention, Weighted control is performed to enable the model to flexibly switch between physical drive and data drive, which is particularly suitable for complex ocean modeling environments where the quality of measured data is inconsistent, regional boundaries are incomplete, or only part of the boundaries have observation data.

[0104] In a possible implementation of the above embodiment, minimizing the composite loss function by a gradient back-propagation algorithm, and completing the training of the physical information neural network model when the composite loss function is lower than a preset threshold, includes:

[0105] The physical information neural network model is trained using the Adam optimizer;

[0106] Based on the training data set, perform several random gradient iterative updates to optimize the set of parameters to be trained in the physical information neural network model. ;

[0107] During the training process, if the composite loss function When the value falls below a preset threshold or reaches the maximum number of iterations, the training of the physical information neural network model is terminated, and the pre-trained physical information neural network model is determined.

[0108] In this embodiment, the model is initialized, including initializing a feedforward neural network, which may be composed of multiple hidden layers, each of which has multiple nodes, and an activation function may be a Tanh. For example, the feedforward neural network may be composed of three hidden layers, each of which has 100 nodes.

[0109] Here, the Tanh activation function can be an S-shaped activation function with an output range of [-1, 1]. Specifically, since tidal fluctuations fluctuate around a specific equilibrium point and have periodic symmetry, the Tanh activation function is also symmetric about the origin, which helps capture periodic or symmetric fluctuations in physical phenomena, thereby facilitating the learning of representations that are more consistent with physical characteristics.

[0110] In addition, since the Tanh activation function is a smooth and continuous activation function, it can provide continuous derivatives over the entire domain, and there will be no derivative discontinuities like the ReLU activation function. This can greatly improve the numerical stability and smoothness of the model in the residual calculation of the derivative term.

[0111] Furthermore, in one possible implementation, during the model training process, an Adam optimizer can be used to perform optimization at a preset learning rate, and optimization can be performed at a preset learning rate during each training process, and each training uses a batch of a preset number of tide data.

[0112] For example, the preset learning rate can be 1*10 -3 , the preset number can be 128 groups.

[0113] Furthermore, the training dataset is divided into small blocks, and 128 groups of training samples are randomly selected from the training dataset as training inputs for each training iteration.

[0114] Furthermore, during training, if the composite loss function If the value falls below the preset threshold or the number of iterations reaches 40,000, the training of the physical information neural network model is terminated and the pre-trained physical information neural network model is determined.

[0115] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiment of the present disclosure, since the tide signal itself has periodicity, volatility and symmetry, the use of the Tanh activation function can better capture the physical characteristics of the tide change, thereby improving the calculation accuracy of the physical residual. Using a small batch consisting of 128 groups of tide data in each round of training, combined with the Adam optimizer for iterative parameter update, can effectively reduce memory overhead and improve training efficiency. In the training process, the preset composite loss function threshold and the maximum number of iterations are introduced as termination conditions, which can timely terminate the training process on the premise of ensuring that the model has achieved better physical consistency and fitting accuracy, avoid overfitting the training data, and improve the stability of the model.

[0116] In a possible implementation of step S102, a training data set is constructed based on the harmonic parameters and the simulated tide time series data, the training data set is input into a deep neural network model in the hybrid neural network model, and the deep neural network model is trained, including:

[0117] N T Each set of simulated tide time series data in the group of simulated tide time series data is used as the input tensor, and the harmonic parameters corresponding to each set of simulated tide time series data are used as the output label to construct the training data set;

[0118] The deep neural network model is trained using a training dataset, and a preset loss function is used to evaluate the deviation between the simulated harmonic parameters and the true harmonic parameters;

[0119] The parameters of the deep neural network model are optimized through the Adam optimizer. When the training loss converges or the maximum number of iterations is reached, the training of the deep neural network model is completed.

[0120] In this embodiment, N T A set of simulated tide time series data is generated, and the harmonic parameters of each set of simulated tide time series data are used as output label data to construct a training dataset, and the training dataset is used to train the DNN model. The training dataset includes multiple training pairs.

[0121] The set of parameters to be trained for the DNN model using the Adam optimizer Optimize and complete the training of the deep neural network model when the training loss converges or the maximum number of iterations is reached.

[0122] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiments of the present invention, by establishing a mapping relationship between the simulated tide time series data and its corresponding harmonic parameters, and using the DNN model to learn this nonlinear relationship, it is possible to directly predict the harmonic parameters from the tide sequence without performing computationally complex traditional harmonic analysis, thereby greatly improving the efficiency and automation of harmonic parameter extraction.

[0123] In a possible implementation of the above embodiment, each of the multiple hidden layers of the deep neural network model is connected to a batch normalization layer to standardize the output features of each hidden layer.

[0124] In this embodiment, the DNN model may include 5 hidden layers, and the number of neurons in each hidden layer may be 100, 100, 80, 50, and 30 respectively.

[0125] Batch normalization is used for each hidden layer, that is, the mean and variance of each dimension of the small batch samples are normalized.

[0126] Through the ocean tide simulation method and device based on the hybrid neural network model of the above-mentioned embodiments of the present invention, batch normalization can maintain a stable distribution of the input of each layer, so that a larger learning rate can also be used for training, which helps to accelerate model convergence, reduce the number of iterations, and improve training efficiency.

[0127] In a possible implementation of the above embodiment, the number of neurons in the input layer of the deep neural network model is M, which is equal to the number of tide-watching stations; the number of neurons in the output layer of the deep neural network model is 2*N b , the number of neurons in the output layer is equal to twice the number of nodes on the open boundary.

[0128] In this embodiment, the number of neurons in the input layer is M, which can represent that the model accepts the simulated tide level time series data of M tide observation stations at one time. The number of neurons in the output layer of the deep neural network model is 2*N b , it can be characterized that each boundary node needs to predict two harmonic parameters, namely the output amplitude and phase.

[0129] Furthermore, during the training of the DNN model, the batch size of the training data can be 215, the initial learning rate can be 0.03, and the maximum number of iterations can be 1000.

[0130] The ocean tide simulation method and apparatus based on the hybrid neural network model described in the above-mentioned embodiments of the present disclosure achieve a one-to-one correspondence between the number of input layer neurons and the number of tide-observing stations, enabling complete reception and expression of all observed tide level information, thereby improving the model's spatial coverage. The number of output layer neurons is set to twice the number of boundary nodes, allowing the model to simultaneously predict the amplitude and phase information of each boundary node, thereby enabling tidal harmonic decomposition capabilities.

[0131] In one embodiment, please refer to Figure 2 , Figure 2 The following is a flow chart of constructing a training data set for a method for simulating ocean tides based on a hybrid neural network model, provided by an embodiment of the present disclosure. The flow may include the following steps:

[0132] Step S201, obtaining the harmonic parameters of multiple nodes on the open boundary;

[0133] Step S202, inputting the harmonized parameters into the physical information neural network model;

[0134] Step S203, obtaining simulated tide time series data;

[0135] Step S204, storing the harmonized parameters and the simulated tide time series data into a training data set;

[0136] Step S205: Check whether the number of data sets is greater than the preset number; if so, end the process; if not, proceed to step S206;

[0137] Step S206, using the astronomical tidal parameter random generator to continue generating harmonic parameters;

[0138] Here, the generation of the harmonization parameters continues, and the process proceeds to step S201.

[0139] In one embodiment, please refer to Figure 3 , Figure 3 : is a schematic diagram of the DNN model structure of an ocean tide simulation method based on a hybrid neural network model provided by an embodiment of the present disclosure, wherein:

[0140] Assume that the number of tide gauge stations M = 5 and the number of nodes on the open boundary N b =3, then the number of neurons in the input layer of the DNN model is 5, and each neuron in the input layer corresponds to a simulated tide level time series data (i.e., the first tide level data to the fifth tide level data, i.e. ); The number of neurons in the output layer is 3, and each output layer corresponds to the optimal reconciliation parameter of an open boundary node (i.e., the first reconciliation parameter to the third reconciliation parameter, i.e. 、 、 ). In addition, the number of neurons in the first hidden layer of the DNN model is 6, and the number of neurons in the second hidden layer is 5.

[0141] In one embodiment, a device 400 for simulating ocean tides based on a hybrid neural network model is provided. The device 400 for simulating ocean tides based on a hybrid neural network model corresponds one-to-one to the method for simulating ocean tides based on a hybrid neural network model in the above embodiment. Figure 4 As shown, the ocean tide simulation device 400 based on the hybrid neural network model includes a data driving module 401, a model training module 402 and an optimal harmonic parameter acquisition module 403, wherein each functional module is described in detail as follows:

[0142] The data driving module 401 is used to obtain the harmonic parameters corresponding to the multiple nodes on the open boundary of the target sea area, input the harmonic parameters into the pre-trained physical information neural network model in the hybrid neural network model, and obtain the simulated tide level time series data corresponding to the multiple tide-observing stations;

[0143] A model training module 402 is configured to construct a training data set based on the harmonic parameters and the simulated tidal time series data, input the training data set into a deep neural network model in the hybrid neural network model, and train the deep neural network model so that the deep neural network model learns the inverse mapping relationship from the simulated tidal time series data to the harmonic parameters;

[0144] The optimal harmonic parameter acquisition module 403 is used to input the measured tide level time series data of multiple tide-observing stations into the trained deep neural network model, and obtain the optimal harmonic parameters of multiple nodes from the deep neural network model.

[0145] In one embodiment, the data driving module 401 is specifically used to obtain N on the open boundary of the target sea area through the astronomical tide parameter random generator. b Each node corresponds to N T Group harmonic parameters Each set of harmonic parameters includes the first of the eight astronomical tides. k The amplitude of the astronomical tide and phase .

[0146] In one embodiment, the data driving module 401 is specifically used to construct the input layer and output layer of the physical information neural network model. The input layer includes at least N b Each node corresponds to N T Group harmonic parameters The output layer includes at least N tide-watching stations for each of the M tide-watching stations. T Simulated tide level time series data ;

[0147] Based on the physical constraint term and the data-driven term, a composite loss function of the physical information neural network model is defined; wherein the physical constraint term uses the shallow water equation as the physical constraint;

[0148] The composite loss function is minimized through the gradient back-propagation algorithm. When the composite loss function is lower than the preset threshold, the training of the physical information neural network model is completed.

[0149] In one embodiment, the data driving module 401 is specifically used to construct the input layer of the physical information neural network model, and the input layer includes at least one of the following: any sampling coordinate and sampling time in the target sea area ( x, y, t ), N of each node on the open boundary T Group harmonic parameters; wherein the sampling coordinates in the target sea area include at least one of the following: nodes on the open boundary, tide-observing stations, and points inside the sea area;

[0150] Construct the output layer of the physical information neural network model, which includes at least one of the following: the simulated tide level of any sampling coordinate in the target sea area ζ ( x, y, t ), velocity component U ( x, y, t )and V ( x, y, t ), where the simulated tide level ζ ( x, y, t ), velocity component U ( x, y, t )and V ( x, y, t ) is used to construct the simulated tide time series data .

[0151] In one embodiment, the data driving module 401 is specifically configured to use the residual term of the equation constructed based on the shallow water equation as a physical constraint term;

[0152] The boundary residual term constructed based on the difference between the measured tide level time series data and the simulated tide level time series data is used as the data-driven term;

[0153] The composite loss function of the physical information neural network model is defined as the weighted sum of the physical constraint terms and the data-driven terms.

[0154] In one embodiment, the residual term of the equation constructed based on the shallow water equation is:

[0155]

[0156]

[0157] in, is the vector of conserved variables of the shallow water equation; Simulating tide levels ζ ( x, y, t ) and velocity components U ( x, y, t )and V ( x, y, t ) constituted by the flux function; Simulating tide levels ζ ( x, y, t ) and velocity components U ( x, y, t )and V ( x, y, t ) is the source function composed of; is the residual of the physical information neural network model for the shallow water equation at the sampling coordinates; N is the sampling number of the sampling coordinates in the target sea area; is the set of parameters to be trained for the physical information neural network model.

[0158] In one embodiment, the boundary residual term constructed based on the difference between the measured tide level time series data and the simulated tide level time series data is:

[0159]

[0160] in, Physical information neural network model at tide gauge station and time The simulated tide level output at At the tide gauge station and time The measured tide level at It represents the mean square error term between the simulated tide level and the measured tide level.

[0161] In one embodiment, the composite loss function of the data-driven module 401 and the physical information neural network model is:

[0162]

[0163] in, is the preset weight coefficient.

[0164] In one embodiment, the data driving module 401 is specifically configured to train the physical information neural network model using an Adam optimizer;

[0165] Based on the training data set, perform several random gradient iterative updates to optimize the set of parameters to be trained in the physical information neural network model. ;

[0166] During the training process, if the composite loss function When the value falls below a preset threshold or reaches the maximum number of iterations, the training of the physical information neural network model is terminated, and the pre-trained physical information neural network model is determined.

[0167] In one embodiment, the model training module 402 is used to T Each set of simulated tide time series data in the group of simulated tide time series data is used as the input tensor, and the harmonic parameters corresponding to each set of simulated tide time series data are used as the output label to construct the training data set;

[0168] The deep neural network model is trained using a training dataset, and a preset loss function is used to evaluate the deviation between the simulated harmonic parameters and the true harmonic parameters;

[0169] The parameters of the deep neural network model are optimized through the Adam optimizer. When the training loss converges or the maximum number of iterations is reached, the training of the deep neural network model is completed.

[0170] In one embodiment, each of the multiple hidden layers of the deep neural network model is connected to a batch normalization layer to standardize the output features of each hidden layer.

[0171] In one embodiment, the number of neurons in the input layer of the deep neural network model is M, which is equal to the number of tide-watching stations; the number of neurons in the output layer of the deep neural network model is 2*N b , the number of neurons in the output layer is equal to twice the number of nodes on the open boundary.

[0172] It should be noted that: the ocean tide simulation device based on the hybrid neural network model provided in the above embodiment only uses the division of the above program modules as an example to illustrate when implementing the corresponding ocean tide simulation method based on the hybrid neural network model. In actual application, the above processing can be assigned to different program modules as needed, that is, the internal structure of the above system can be divided into different program modules to complete all or part of the processing described above. In addition, the system provided in the above embodiment and the corresponding Figure 1 The embodiments of the method shown belong to the same concept, and their specific implementation processes are detailed in the method embodiments, which will not be repeated here.

[0173] The present disclosure also provides an electronic device having the above Figure 4 The ocean tide simulation device based on the hybrid neural network model is shown.

[0174] See also Figure 5 , Figure 5 is a structural diagram of another ocean tide simulation device based on a hybrid neural network model provided by an embodiment of the present disclosure, such as Figure 5As shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0175] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0176] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0177] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0178] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0179] The electronic device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means.Figure 5 The bus connection is taken as an example.

[0180] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the electronic device. Examples include a touch screen, keypad, mouse, trackpad, touchpad, pointer, one or more mouse buttons, trackball, joystick, etc. The output device 40 may include a display device, auxiliary lighting devices (e.g., LEDs), and tactile feedback devices (e.g., vibration motors). Such display devices include, but are not limited to, liquid crystal displays, light emitting diodes, monitors, and plasma displays. In some optional embodiments, the display device may be a touch screen.

[0181] The electronic device also includes a communication interface for the electronic device to communicate with other devices or a communication network.

[0182] The embodiments of the present disclosure also provide a computer-readable storage medium. The above-mentioned method according to the embodiments of the present disclosure can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0183] A portion of the present disclosure may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present disclosure through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes but is not limited to a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0184] Although the embodiments of the present disclosure have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for simulating ocean tides based on a hybrid neural network model, characterized in that: The method comprises: Obtaining harmonic parameters corresponding to a plurality of nodes on the open boundary of the target sea area, inputting the harmonic parameters into a pre-trained physical information neural network model in the hybrid neural network model, and obtaining simulated tide level time series data corresponding to a plurality of tide-observing stations; The method of obtaining the harmonic parameters corresponding to the multiple nodes on the open boundary of the target sea area includes: obtaining N nodes on the open boundary of the target sea area through the astronomical tide parameter random generator. b Each node corresponds to N T Group harmonic parameters Each set of harmonic parameters includes the first of the eight astronomical tides. k The amplitude of the astronomical tide and phase ; The pre-trained physical information neural network model is implemented based on the following steps: Construct the input layer and output layer of the physical information neural network model, the input layer includes at least N b Each node corresponds to N T Group harmonic parameters The output layer includes at least N tide-watching stations for each of the M tide-watching stations. T Simulated tide level time series data ; Based on the physical constraint term and the data-driven term, a composite loss function of the physical information neural network model is defined; wherein the physical constraint term uses the shallow water equation as the physical constraint; Minimizing the composite loss function by a gradient back-propagation algorithm, and completing the training of the physical information neural network model when the composite loss function is lower than a preset threshold; The input layer and output layer of the physical information neural network model are constructed as follows: Construct the input layer of the physical information neural network model, the input layer includes at least one of the following: any sampling coordinate and sampling time in the target sea area ( x,y,t ), N of each node on the open boundary T Group harmonization parameters; wherein the sampling coordinates in the target sea area include at least one of the following: nodes on the open boundary, tide-observing stations, and points inside the sea area; Construct the output layer of the physical information neural network model, the output layer includes at least one of the following: the simulated tide level of any sampling coordinate in the target sea area; ζ ( x,y,t ), velocity component U ( x,y,t )and V ( x,y,t ), wherein the simulated tide level ζ ( x,y,t ), velocity component U ( x,y,t )and V ( x,y,t ) is used to construct the simulated tide time series data ; Constructing a training data set based on the harmonic parameters and the simulated tide time series data, inputting the training data set into a deep neural network model in the hybrid neural network model, and training the deep neural network model so that the deep neural network model learns the inverse mapping relationship from the simulated tide time series data to the harmonic parameters; The measured tide level time series data of the multiple tide-observing stations are input into the trained deep neural network model, and the optimal harmonic parameters of the multiple nodes are obtained from the deep neural network model.

2. The method according to claim 1, characterized in that The composite loss function of the physical information neural network model is defined based on the physical constraint term and the data driven term, including: The residual term of the equation constructed based on the shallow water equation is used as a physical constraint term; The boundary residual term constructed based on the difference between the measured tide level time series data and the simulated tide level time series data is used as the data-driven term; The composite loss function of the physical information neural network model is defined as the weighted sum of the physical constraint term and the data-driven term.

3. The method according to claim 2, characterized in that The residual term of the equation constructed based on the shallow water equation is: in, is the vector of conserved variables of the shallow water equation; Simulating tide levels ζ ( x,y,t ) and velocity components U ( x,y,t )and V ( x,y,t ) constituted by the flux function; Simulating tide levels ζ ( x,y,t ) and velocity components U ( x,y,t )and V ( x,y,t ) is the source function composed of; is the residual of the physical information neural network model for the shallow water equation at the sampling coordinates; N is the sampling number of the sampling coordinates in the target sea area; is the set of parameters to be trained of the physical information neural network model.

4. The method according to claim 2, characterized in that The boundary residual term constructed based on the difference between the measured tide level time series data and the simulated tide level time series data is: in, Physical information neural network model at tide gauge station and time The simulated tide level output at At the tide gauge station and time The measured tide level at Characterizes the mean square error term between the simulated tide level and the measured tide level.

5. The method according to any one of claims 2 to 4, characterized in that The composite loss function of the physical information neural network model is: in, is the preset weight coefficient.

6. The method according to claim 1, characterized in that The method of minimizing the composite loss function by using a gradient back-propagation algorithm and completing the training of the physical information neural network model when the composite loss function is lower than a preset threshold value includes: The physical information neural network model is trained using the Adam optimizer; Based on the training data set, perform several random gradient iterative updates to optimize the set of parameters to be trained in the physical information neural network model. ; During the training process, if the composite loss function If the value falls below a preset threshold or reaches a maximum number of iterations, the training of the physical information neural network model is terminated, and a pre-trained physical information neural network model is determined.

7. The method according to claim 1, characterized in that The step of constructing a training data set based on the harmonic parameters and the simulated tide time series data, inputting the training data set into a deep neural network model in the hybrid neural network model, and training the deep neural network model includes: N T Each set of simulated tide time series data in the group of simulated tide time series data is used as the input tensor, and the harmonic parameters corresponding to each set of simulated tide time series data are used as the output label to construct the training data set; The deep neural network model is trained using the training data set, and the deviation between the simulated harmonic parameters and the true harmonic parameters is evaluated using a preset loss function; The parameters of the deep neural network model are optimized by the Adam optimizer, and the training of the deep neural network model is completed when the training loss converges or reaches the maximum number of iterations.

8. The method according to claim 7, characterized in that Each of the multiple hidden layers of the deep neural network model is connected to a batch normalization layer to standardize the output features of each hidden layer.

9. The method according to claim 8, characterized in that The number of neurons in the input layer of the deep neural network model is M, which is equal to the number of tide-watching stations; the number of neurons in the output layer of the deep neural network model is 2*N b , the number of neurons in the output layer is equal to twice the number of nodes on the open boundary.

10. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the ocean tide simulation method based on a hybrid neural network model as described in any one of claims 1 to 9 when executing the computer program.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the ocean tide simulation method based on the hybrid neural network model as described in any one of claims 1 to 9 are implemented.

12. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the ocean tide simulation method based on the hybrid neural network model as described in any one of claims 1 to 9 are implemented.

Images