Offshore sea wave parameter downscaling method and system based on generative adversarial neural network

Abstract

The invention relates to the technical field of sea wave parameter downscaling, and provides an offshore sea wave parameter downscaling method and system based on a generative adversarial neural network, and the method comprises the steps: obtaining a high-resolution sea wave parameter through a downscaling model based on a low-resolution sea wave parameter and high-resolution water depth data; wherein the downscaling model adopts a generative adversarial neural network, a generator of the generative adversarial neural network comprises a plurality of fusion convolution blocks which are sequentially connected, a convolution attention module and an up-sampling block are sequentially arranged between every two connected fusion convolution blocks, and high-resolution terrain information is embedded into the last fusion convolution block. High-resolution terrain information is embedded into the downscaling model, so that the physical adaptability of a generated result is remarkably improved, and the capturing capability of small-scale features is enhanced; fusion of multi-source data is coordinated by adopting a convolution attention mechanism, efficient integration is realized by selectively paying attention to important features, and more accurate support is provided for generating high-resolution offshore sea wave parameters.

Description

Technical Field

[0001] The present invention belongs to the technical field of ocean wave parameter downscaling, and in particular relates to an offshore ocean wave parameter downscaling method and system based on a generative adversarial neural network. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] High-resolution ocean wave parameters can improve the accuracy and efficiency of coastal protection, maritime safety, and resource development. Current offshore wave parameter downscaling methods are mainly divided into physical model-driven and data-driven models.

[0004] Physical model-driven approaches primarily rely on numerical simulation techniques, such as the third-generation wave models (WAVEWATCH III and SWAN), which simulate the spatial and temporal variations of waves using wind field data. These methods offer strong physical interpretation, but they require high precision in input data and are computationally complex, making them difficult to rapidly apply in high-resolution offshore scenarios. Furthermore, due to the complex topography of nearshore areas, physical models may lack accuracy in handling these nonlinear and inhomogeneous environments, particularly under extreme conditions such as swells and typhoons.

[0005] Data-driven models, based on statistical analysis or machine learning techniques, downscale wave parameters by mapping observational data to model outputs. While data-driven methods can extract complex nonlinear features, they lack dynamic constraints and can produce results that do not conform to physical laws when boundary conditions change. This can lead to large prediction errors, especially in extreme weather conditions. Summary of the Invention

[0006] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a near-shore wave parameter downscaling method and system based on a generative adversarial neural network. By embedding high-resolution terrain information into the downscaling model, the physical adaptability of the generated results can be significantly improved, and the ability to capture small-scale features can be enhanced. At the same time, a convolutional attention mechanism is used to coordinate the fusion of multi-source data, and efficient integration is achieved by selectively focusing on important features, providing more accurate support for the generation of high-resolution near-shore wave parameters.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A first aspect of the present invention provides a method for downscaling offshore wave parameters based on a generative adversarial neural network, comprising:

[0009] Obtain low-resolution wave parameters and high-resolution water depth data;

[0010] Based on low-resolution wave parameters and high-resolution water depth data, high-resolution wave parameters are obtained through downscaling model;

[0011] Among them, the downscaling model adopts a generative adversarial neural network, and the generator of the generative adversarial neural network contains several fused convolution blocks connected in sequence. A convolutional attention module and an upsampling block are set in sequence between every two connected fused convolution blocks, and high-resolution terrain information is embedded in the last fused convolution block.

[0012] Furthermore, the ocean wave parameters include wave height, period, and wind speed.

[0013] Furthermore, the fused convolution block uses convolution kernels of different sizes in parallel in the same layer to perform feature extraction.

[0014] Furthermore, the discriminator in the generative adversarial neural network includes an input convolution block, three feature extraction convolution blocks and an output convolution block.

[0015] Furthermore, the generator loss of the downscaling model is a weighted average of the L1 norm and the adversarial loss.

[0016] Furthermore, the discriminator loss of the downscaling model uses binary cross entropy loss to measure the difference between the generated image and the real image.

[0017] A second aspect of the present invention provides a system for downscaling offshore wave parameters based on a generative adversarial neural network, comprising:

[0018] A data acquisition module is configured to: acquire low-resolution ocean wave parameters and high-resolution water depth data;

[0019] A downscaling module is configured to obtain high-resolution wave parameters through a downscaling model based on low-resolution wave parameters and high-resolution water depth data;

[0020] Among them, the downscaling model adopts a generative adversarial neural network, and the generator of the generative adversarial neural network contains several fused convolution blocks connected in sequence. A convolutional attention module and an upsampling block are set in sequence between every two connected fused convolution blocks, and high-resolution terrain information is embedded in the last fused convolution block.

[0021] Furthermore, the ocean wave parameters include wave height, period, and wind speed.

[0022] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the offshore wave parameter downscaling method based on a generative adversarial neural network as described above.

[0023] A fourth aspect of the present invention provides a computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein when the processor executes the program, the steps of the offshore wave parameter downscaling method based on a generative adversarial neural network as described above are implemented.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention embeds high-resolution terrain information into the downscaling model, which can significantly improve the physical adaptability of the generated results and enhance the ability to capture small-scale features. Especially under extreme conditions, high-resolution terrain can effectively reduce the prediction error caused by ignoring terrain characteristics.

[0026] The present invention adopts the convolutional attention mechanism to dynamically adjust the weights in the feature map, focusing on areas with significant changes in wind and waves or areas affected by complex terrain, thereby greatly improving the downscaling model's ability to express key areas; in addition, the convolutional attention mechanism can establish a stronger connection between global features and local details, organically combining large-scale wave propagation characteristics with the refined features of local terrain; by suppressing the need to model low-correlation areas, the attention mechanism can also reduce information redundancy, optimize computational efficiency, and at the same time improve the quality of generated results under limited computing resources, providing more accurate support for generating high-resolution offshore wave parameters.

[0027] The fused convolution block of the present invention can extract features within a variety of perception fields by using convolution kernels of different sizes in parallel in the same layer, and integrate these features together, thereby achieving stronger feature extraction and expression capabilities and adapting to complex input data structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0029] Figure 1 is a structural diagram of a generative adversarial neural network according to embodiment 1 of the present invention;

[0030] Figure 2 It is a structural diagram of a computer device according to the fourth embodiment of the present invention. DETAILED DESCRIPTION

[0031] To make the objectives, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0032] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0033] Explanation of terms:

[0034] Downscaling: A technique that uses specific methods to improve low-resolution input data (such as weather model output, wave model results, etc.) to high resolution.

[0035] Example 1

[0036] This embodiment provides a method for downscaling offshore wave parameters based on a generative adversarial neural network.

[0037] The offshore wave parameter downscaling method based on a generative adversarial neural network provided in this embodiment integrates a convolutional block attention module and a conditional generative adversarial neural network model constrained by physical conditions to perform downscaling. The neural network uses numerically calculated low-resolution wave parameters to predict high-resolution wave parameters, thereby providing a new technical approach for wave parameter downscaling.

[0038] Considering that the wave dynamics in nearshore areas are affected by wind, topography, tides, and other environmental factors, the offshore wave parameter downscaling method based on a generative adversarial neural network provided in this embodiment embeds high-resolution terrain information into the downscaling model, which can significantly improve the physical adaptability of the generated results and enhance the ability to capture small-scale features. At the same time, the convolutional attention mechanism coordinates the fusion of multi-source data (such as wind fields, low-resolution wave fields, and terrain data), achieves efficient integration by selectively focusing on important features, and provides more accurate support for generating high-resolution offshore wave parameters.

[0039] The offshore wave parameter downscaling method based on a generative adversarial neural network provided in this embodiment includes the following steps:

[0040] Step 1: Establish a long-term database for wave downscaling: obtain high-resolution and low-resolution wave height and hourly data of the corresponding period, wind speed, and high-resolution water depth data through numerical hindcasting, with a data interval of 1 hour and a data length of ≥5 years; perform preprocessing steps such as data cleaning, data alignment, and format unification on the acquired data to ensure its accuracy and consistency to support the training and validation of the downscaling model.

[0041] Step 2: Divide the downscaling model into training and test sets.

[0042] Among them, the data from 2019 to 2022 are used as the training set, and the data from 2023 are used as the test set.

[0043] The input of the downscaling model includes low-resolution (12 km) wave height, period, wind speed and high-resolution (3 km) water depth data, and the output of the downscaling model is high-resolution (3 km) wave height, period and wind speed.

[0044] It should be noted that the low-resolution (12 km) wave height, period, wind speed and high-resolution (3 km) water depth data need to be processed into matrix form and visualized as graphics.

[0045] Wave height refers to the distance from the trough to the crest of a wave. It is a visual indicator of wave size. In marine weather forecasts, significant wave height (Hs) is often mentioned. It is the average height of the third largest wave height in the wave spectrum.

[0046] The wave period (period) refers to the time it takes for successive wave crests or troughs to pass through a fixed location. In this embodiment, the average wave period Tm is used. The period reflects the speed of wave motion, and different wave periods can have different impacts on maritime navigation and coastal engineering.

[0047] Wind speed (Ws) is a key factor influencing wave height and period. In open waters, wind speed, wind zone length, and wind time collectively determine wave characteristics. Generally speaking, higher wind speeds produce higher waves and longer periods.

[0048] Water depth refers to the depth of the water body, usually represented by d. Water depth has a significant impact on wave propagation and energy transfer.

[0049] Step 3: Construct a generative adversarial neural network model for wave downscaling.

[0050] (1) The downscaling model of this embodiment uses a fused convolution module as the initial module.

[0051] The main feature of the fused convolution module is the parallel application of convolution kernels of various sizes within the same layer, including 1×1, 3×3, and 5×5 kernels. The 1×1 kernel is used for feature compression or dimensionality transformation, extracting fine local features; the 3×3 kernel is used to extract mid-range features; and the 5×5 kernel is used to capture features at a wider range, focusing on global information. This design extracts multi-scale feature information at each layer, thereby improving the network's feature representation capabilities.

[0052] In this embodiment, by using convolution kernels of different sizes (1×1, 3×3, 5×5) in parallel in the same layer, features can be extracted within a variety of perception fields and integrated together, thereby achieving stronger feature extraction and expression capabilities and adapting to complex input data structures.

[0053] (2) The downscaling model of this embodiment mainly uses a deep convolutional network structure. Convolutional structures are used in the generator and discriminator. By replacing all pooling layers in the generative adversarial neural network with convolutions: using inverse convolution in the discriminator and using a fused convolution module in the generator, the entire generative adversarial neural network process is expressed by the following formula:

[0054] min G max D V(D,G)=min G max D E x～pdata(x) [logD(x)]+E z～pz(z) [log(1-D(G(z))]

[0055] Among them, x is a sample of real wave data, p data (x) represents the probability distribution of real data, D(x) represents the probability that it (the image matrix generated by the generator) comes from the real image instead of the generated image, z is the random noise input to the generation network, p z (z) represents the probability distribution of z, while G(z) represents the probability that the generated sample matches the data distribution, and D(G(z)) is the probability of judging whether G(z) comes from a real image.

[0056] (3) A convolutional attention module (CBAM) is coupled in the generative adversarial neural network and conditional variables are added, i.e., high-resolution topographic data of the study area is added to the downscaling model.

[0057] like Figure 1 As shown, the generator in the generative adversarial neural network consists of three fused convolution blocks, two CBAM blocks, and two upsampling blocks. The conditional variable (high-resolution terrain) is placed in the generator. Specifically, the first fused convolution block, the second fused convolution block, and the third fused convolution block are connected in sequence. CBAM blocks and upsampling blocks are sequentially set between the first and second fused convolution blocks, and between the second and third fused convolution blocks. The high-resolution terrain is directly input into the third fused convolution block after convolution to obtain the generated image, namely the high-resolution wave data (wave height, period, and wind speed).

[0058] like Figure 1As shown in the figure, the discriminator in the generative adversarial neural network is a deep convolutional neural network designed to evaluate the authenticity of the generator output (high-resolution wave data). Its structure consists of five convolution blocks, including an input convolution block, three feature extraction convolution blocks and an output convolution block. Specifically, the input convolution block, the first feature extraction convolution block, the second feature extraction convolution block and the third feature extraction convolution block are connected in sequence, and finally connected to the output convolution block. The high-resolution wave data (wave height, period and wind speed) generated by the generator is used as the input of the discriminator. It first undergoes initial feature extraction through the input convolution block, and then undergoes deep feature processing through three feature extraction convolution blocks in sequence. The spatial resolution is gradually compressed and the number of channels is increased between each block. Finally, the output convolution block maps the features into the probability value of a single channel to judge the authenticity of the input data.

[0059] Among them, the activation function of the last layer output of the discriminator uses sigmoid to determine whether the output 1 or 0 corresponds to right or wrong.

[0060] Among them, the channel attention in CBAM focuses on the important physical factors that affect wave changes; spatial attention allows the model to learn to effectively emphasize or suppress and refine intermediate features, thereby focusing on important areas with relatively large wave heights.

[0061] Among them, the calculation process of CBAM's convolutional attention is as follows:

[0062]

[0063] Among them, F represents the input feature map, σ is the activation function used for output normalization, and M c (F) and M s (F) represents the output of the channel attention feature map and spatial attention, W0 and W1 represent the learnable weight matrix, MLP is a multi-layer perceptron, AvgPool is the global average pooling, MaxPool is the global maximum pooling, and They represent the feature transformations achieved by the convolution operation.

[0064] Among them, the convolutional attention module uses the convolutional attention mechanism to dynamically adjust the weights in the feature map, focusing on areas with significant changes in wind and waves or areas affected by complex terrain, thereby greatly improving the downscaling model's ability to express key areas; in addition, the convolutional attention mechanism can establish a stronger connection between global features and local details, organically combining large-scale wave propagation characteristics with the refined features of local terrain; by suppressing the need to model low-correlation areas, the attention mechanism can also reduce information redundancy, optimize computational efficiency, and improve the quality of generated results under limited computing resources.

[0065] Among them, high-resolution terrain information can depict the complex terrain features of the nearshore area in detail, including shoals, harbors and reefs. These features have an important impact on the propagation, refraction and dissipation of waves. By embedding terrain information into the downscaling model, the physical adaptability of the generated results can be significantly improved, and the ability to capture small-scale features can be enhanced. Especially under extreme conditions (such as the interaction between wind waves and surges caused by typhoons), high-resolution terrain can effectively reduce the prediction errors caused by ignoring terrain characteristics.

[0066] Step 6: Define the loss function of the downscaling model.

[0067] The generator loss function is a weighted average of the L1 norm and the adversarial loss (the weight is a hyperparameter) as its loss function, which is generally defined as:

[0068]

[0069] in, It is an adversarial loss that encourages the generator to generate images that can deceive the discriminator into thinking that the generated images are real; is the L1 loss, which is used to encourage the generated image to match the target image at the pixel level; λ is a hyperparameter used to balance the weights of the adversarial loss and the L1 loss.

[0070] The discriminator loss uses binary cross entropy loss to measure the difference between the generated image and the real image, which is generally defined as:

[0071]

[0072] Where D(x) is the output probability of the discriminator for the real image x; D(G(z)) is the output probability of the discriminator for the generated image G(z); x is the probability of the discriminator outputting the generated image from the real data distribution p data (x) is the real image; z is the image from the noise distribution p z (z) is the noise vector.

[0073] Step 7: Train the generative adversarial neural network model to obtain the final downscaling model.

[0074] (1) Initialize the generator and discriminator, and define their optimizers as Adam.

[0075] (2) Generator training: a. Input a low-resolution image and generate a high-resolution image; b. Use the discriminator to determine whether the generated image is real or fake; c. Calculate the adversarial loss and content loss of the generator; d. Backpropagate and update the weights of the generator.

[0076] (3) Discriminator training: a. Input real high-resolution images and generated high-resolution images; b. Calculate the adversarial loss of the discriminator; c. Backpropagate and update the weights of the discriminator.

[0077] (3) Evaluate the performance of the downscaling model on the validation set and adjust the hyperparameters and downscaling model structure to obtain the best results.

[0078] Step 8: Evaluation and tuning of the downscaling model.

[0079] The RMSE and CC error statistics methods as well as PSNR and SSIM are used to evaluate the quality of the generated images. The hyperparameters such as learning rate and loss weight are adjusted by evaluating the downscaling model to optimize the performance of the downscaling model.

[0080] Step 9: Application of downscaling model.

[0081] Obtain low-resolution wave height, period, wind speed and high-resolution water depth data;

[0082] Based on low-resolution wave height, period, wind speed and high-resolution water depth data, high-resolution wave height, period and wind speed are obtained through a downscaling model.

[0083] The offshore wave parameter downscaling method based on a generative adversarial neural network provided in this embodiment embeds high-resolution terrain information into the downscaling model, which can significantly improve the physical adaptability of the generated results and enhance the ability to capture small-scale features. At the same time, the convolutional attention mechanism coordinates the fusion of multi-source data (such as wind fields, low-resolution wave fields, and terrain data), achieves efficient integration by selectively focusing on important features, and provides more accurate support for generating high-resolution offshore wave parameters.

[0084] Example 2

[0085] This embodiment provides a coastal wave parameter downscaling system based on a generative adversarial neural network, which specifically includes:

[0086] A data acquisition module is configured to: acquire low-resolution ocean wave parameters and high-resolution water depth data;

[0087] A downscaling module is configured to obtain high-resolution wave parameters through a downscaling model based on low-resolution wave parameters and high-resolution water depth data;

[0088] Among them, the downscaling model adopts a generative adversarial neural network, and the generator of the generative adversarial neural network contains several fused convolution blocks connected in sequence. A convolutional attention module and an upsampling block are set in sequence between every two connected fused convolution blocks, and high-resolution terrain information is embedded in the last fused convolution block.

[0089] Furthermore, the ocean wave parameters include wave height, period, and wind speed.

[0090] Furthermore, the fused convolution block uses convolution kernels of different sizes in parallel in the same layer to perform feature extraction.

[0091] Furthermore, the discriminator in the generative adversarial neural network includes an input convolution block, three feature extraction convolution blocks and an output convolution block.

[0092] Furthermore, the generator loss of the downscaling model is a weighted average of the L1 norm and the adversarial loss.

[0093] Furthermore, the discriminator loss of the downscaling model uses binary cross entropy loss to measure the difference between the generated image and the real image.

[0094] It should be noted here that the various modules in this embodiment correspond one-to-one to the various steps in Example 1, and the specific implementation processes are the same, which will not be repeated here.

[0095] Example 3

[0096] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the offshore wave parameter downscaling method based on a generative adversarial neural network as described in the first embodiment are implemented.

[0097] Example 4

[0098] This embodiment provides a computer device, such as Figure 2 As shown, the system includes a display device, an input device, a computer-readable storage medium (volatile memory and non-volatile storage medium), a processor, a communication interface (i.e., a network interface), and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor, the communication interface, and the computer-readable storage medium can be connected via a bus or other means. The communication interface is used to receive and send data, and when the processor executes the program, the steps of the offshore wave parameter downscaling method based on a generative adversarial neural network as described in the first embodiment are implemented.

[0099] Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSR DRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0100] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0101] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0102] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0103] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for downscaling offshore wave parameters based on generative adversarial neural networks, characterized by: include: Obtain low-resolution wave parameters and high-resolution water depth data; Based on low-resolution wave parameters and high-resolution water depth data, high-resolution wave parameters are obtained through downscaling model; Among them, the downscaling model adopts a generative adversarial neural network, and the generator of the generative adversarial neural network contains several fused convolution blocks connected in sequence. A convolutional attention module and an upsampling block are set in sequence between every two connected fused convolution blocks, and high-resolution terrain information is embedded in the last fused convolution block.

2. The offshore wave parameter downscaling method based on generative adversarial neural network according to claim 1, characterized in that: The ocean wave parameters include wave height, period, and wind speed.

3. The offshore wave parameter downscaling method based on generative adversarial neural network according to claim 1, characterized in that: The fused convolution block uses convolution kernels of different sizes in parallel in the same layer to perform feature extraction.

4. The offshore wave parameter downscaling method based on generative adversarial neural network according to claim 1, characterized in that: The discriminator in the generative adversarial neural network includes an input convolution block, three feature extraction convolution blocks and an output convolution block.

5. The offshore wave parameter downscaling method based on generative adversarial neural network according to claim 1, characterized in that: The generator loss of the downscaling model is a weighted average of the L1 norm and the adversarial loss.

6. The offshore wave parameter downscaling method based on generative adversarial neural network according to claim 1, characterized in that: The discriminator loss of the downscaling model uses binary cross entropy loss to measure the difference between the generated image and the real image.

7. A coastal wave parameter downscaling system based on generative adversarial neural networks, characterized by: include: A data acquisition module is configured to: acquire low-resolution ocean wave parameters and high-resolution water depth data; A downscaling module is configured to obtain high-resolution wave parameters through a downscaling model based on low-resolution wave parameters and high-resolution water depth data; Among them, the downscaling model adopts a generative adversarial neural network, and the generator of the generative adversarial neural network contains several fused convolution blocks connected in sequence. A convolutional attention module and an upsampling block are set in sequence between every two connected fused convolution blocks, and high-resolution terrain information is embedded in the last fused convolution block.

8. The offshore wave parameter downscaling system based on generative adversarial neural network according to claim 7, characterized in that: The ocean wave parameters include wave height, period, and wind speed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the offshore wave parameter downscaling method based on a generative adversarial neural network are implemented as described in any one of claims 1 to 6.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein: When the processor executes the program, the steps of the offshore wave parameter downscaling method based on a generative adversarial neural network are implemented as described in any one of claims 1 to 6.

Images