Twin Intelligent Simulation Method and Device for the Evolution Process of Coral Reef Beach Sand

Abstract

The present invention provides a twin intelligent simulation method and device for the evolution process of coral reef beach sand, which relates to the field of high-dimensional spatio-temporal big data and artificial intelligence, and solves the technical problem of continuous and spatially refined simulation of the evolution process of coral reef beach sand. The method includes: obtaining the original data of the process elements affecting the evolution of beach sand, and the selection of process elements considers the main natural and artificial regional environmental characteristics driving the evolution of beach sand; preprocessing the original data to construct a training data set of beach sand state sequences in the form of a multi-dimensional space cube; selecting the SwinLSTM deep learning framework with spatio-temporal dependence extraction ability suitable for beach sand evolution simulation, integrating the cross-attention mechanism, designing a loss function for the beach sand morphology change characteristics, constructing a twin intelligent simulation model for the evolution process of coral reef beach sand, and using the training data set to complete the training, verification and testing of the model; applying the simulation model to realize the approximate simulation of the evolution process and terrain of coral reef beach sand in a specified historical or future time period.

Description

Technical Field

[0001] The present invention relates to the technical field of high-dimensional spatio-temporal big data and artificial intelligence, and more particularly to a twin intelligent simulation method and device for the evolution process of coral reef beach sand. Background Art

[0002] Offshore coral reef beach sand (offshore coral reef beach sand refers to the sandy tidal flats affected by natural and artificial environmental dynamic actions outside the coral reef and the sandbars as the previous geomorphic type before the stable exposed coral islands), as a key geomorphic type in the long-term evolution process of coral reefs, is deeply affected by natural and artificial environmental dynamic actions. These sandy tidal flats and the sandbars before the stable exposed coral islands are not only important components of the coral reef ecosystem, but their evolution process is also a direct reflection of the health and stability of coral reefs under the background of global climate change. Therefore, the simulation of the evolution process of coral reef beach sand is of great significance for the reasonable development, protection of coral reefs and the maintenance of the health of the marine ecosystem.

[0003] With the rapid development of remote sensing technology, researchers have begun to extract the exposed parts of coral reef beach sand from long-term remote sensing images, analyze the temporal changes in their area and perimeter, and gain insights into the dynamic evolution of beach sand by calculating the movement of the centroid and its trajectory. This process mostly uses statistical analysis methods to summarize and reveal the internal laws of beach sand evolution. However, the conclusions obtained by combining the long-term observations and professional knowledge of experts in the field of marine science have to a certain extent revealed the main factors driving the evolution of coral reef beach sand and qualitatively analyzed its evolution process. However, due to problems such as the limited availability of high-resolution images in the coral reef area and the impact of sea clouds and rain on the continuity of historical images, the continuity of the evolution process extraction is insufficient, reducing the accuracy of the evolution analysis. At the same time, for the inference of future evolution, it often relies mainly on theoretical inferences and supplemented by the calculation conclusions of shape statistical analysis, making it difficult to meet the urgent need for fine spatial simulation of beach sand evolution.

[0004] At present, although numerical simulation models such as the two-dimensional planar coastal dynamics numerical model (X-Beach) based on the structured Fortran 77 / 90 architecture and the three-dimensional hydrodynamic-water quality model system (Delft 3D) have been applied to the evolution simulation of continental coastal beaches, their application to the sandy beaches of offshore coral reefs faces many challenges. These models rely on preset model structures and parameters and need to meet strict assumptions. However, the actual evolution process of coral reef sandy beaches is extremely complex and highly nonlinear, and does not fully conform to the preset assumptions. Once the actual situation does not match the model assumptions or the data characteristics change, the numerical simulation model is difficult to automatically adjust, resulting in a deviation between the simulation results and the actual evolution situation, and the accuracy is greatly reduced. In addition, these models also need to obtain the specified variables in the calculation formula through a large number of accurate continuous field measurements, which is difficult to achieve on offshore coral reefs and the measurement cost is relatively high.

[0005] In contrast, deep learning methods have significant advantages in capturing the spatio-temporal dependence characteristics between data due to their strong data-driven and knowledge extraction capabilities. Existing research has combined the Convolutional Neural Network (CNN) and the Recurrent Neural Network (RNN) to successfully learn the spatio-temporal dependence relationships in spatio-temporal data; the Convolutional Long Short-Term Memory (ConvLSTM) further improves the simulation accuracy by extending the fully connected LSTM and replacing the linear operation with a convolutional operation. Recently, the introduction of the Vision Transformer model has provided a new perspective for capturing global dependence relationships. On this basis, the SwinLSTM deep learning framework has emerged. It combines the advantages of LSTM and Swin Transformer, can capture both global and local spatio-temporal dependence relationships, and effectively reduces the computational burden. This framework has been successfully applied in fields such as the MNIST digital dynamic change dataset (MNIST is a proper noun representing a widely used handwritten digit dataset). However, there is little research on its application in the simulation of sandy beach evolution.

[0006] In particular, the evolution process of coral reef beach sand has its uniqueness: multiple natural and anthropogenic environmental dynamic factors are intertwined and affect each other. It is necessary to use intelligent methods of geographic information system (GIS) and remote sensing to extract and calculate key spatio-temporal change information from a large amount of spatio-temporal data to characterize the evolution process of beach sand, and design modules to input the combined multi-dimensional spatio-temporal features into the model, which is an important prerequisite for model simulation; coral reef beach sand has unique geographical characteristics, and reasonable variables need to be selected according to the characteristics to characterize the evolution process for more accurate simulation and inference. For example, the periodically prevailing monsoon makes the evolution of coral reef beach sand show a certain periodic pattern. Another example is that in the future scenario, human maritime shipping activities will affect the accretion rate of coral reef development, and thus affect its growth and evolution process; the changes in the outer periphery of coral reef beach sand are relatively large, and more attention needs to be paid to the changes in the corresponding grids during simulation. Therefore, it is necessary to select a deep learning framework with strong spatio-temporal dependence extraction ability suitable for the simulation problem of beach sand evolution, and design specialized models and methods based on these characteristics to effectively solve the twin simulation problem of the evolution process of far-sea coral reef beach sand.

[0007] The expert experience inference method based on traditional statistics depends on the continuity and availability of image data, the quantification of conclusions is insufficient, and it is difficult to meet the requirements of spatial refinement simulation. The numerical simulation method requires a large amount of measured data that fits the variables of the model formula and is difficult to adapt to complex non-linear beach sand evolution conditions. For the geographical process of coral reef beach sand evolution, the existing deep learning methods have problems such as insufficient feature extraction ability for complex environmental dynamic mechanisms (which needs to be docked with remote sensing big data cloud computing and GIS methods), failure to accurately characterize unique features of geographical objects such as its periodicity and the impact of human activities, and limited ability to capture locally violently changing areas.

[0008] In view of this, we urgently need an innovative twin intelligent simulation method for the evolution process of far-sea coral reef beach sand to accurately simulate the evolution process of coral reef beach sand within a specified historical and future time period, providing strong support for the ecological protection and sustainable development of coral reefs. Summary of the Invention

[0009] In view of the above problems, the present invention provides a twin intelligent simulation method and device for the evolution process of coral reef beach sand. The method integrates the cross-attention mechanism, uses the SwinLSTM framework combining Swin Transformer and LSTM to construct a deep learning model, designs a loss function that focuses on grids with large changes, realizes the twin intelligent simulation of the evolution process of far-sea coral reef beach sand, and simulates the emergence probability and approximate terrain of coral reef beach sand within a specified time period.

[0010] Coral reef beach sand refers to the sandy tidal flat affected by natural and anthropogenic environmental dynamic actions on the outer periphery of the coral reef and the sandbar, which is the previous geomorphic type before the stable exposed coral island.

[0011] Swin Transformer refers to a visual feature extraction model based on the Transformer architecture, specifically designed for processing two-dimensional spatial data such as images. LSTM (Long Short-Term Memory) is a variant of recurrent neural network for sequence data processing, capable of modeling temporal dependencies. SwinLSTM refers to a hybrid neural network architecture that fuses the spatial modeling ability of Swin Transformer and the temporal modeling mechanism of LSTM, applicable to cross-modal data processing scenarios that contain both spatial information and temporal changes.

[0012] According to the first aspect of the present invention, there is provided a twin intelligent simulation method for the evolution process of coral reef beach sand, including: obtaining the original data of the process elements affecting the beach sand evolution, wherein the process elements include the main natural and human regional environmental characteristics driving the beach sand evolution and the exposed morphological characteristics of the beach sand itself; preprocessing the original data to construct a training data set of beach sand state sequences in the form of a multi-dimensional space cube; selecting the SwinLSTM deep learning framework with spatio-temporal dependence extraction ability suitable for beach sand evolution simulation, fusing the cross-attention mechanism, designing a loss function for the beach sand morphological change characteristics, and constructing a twin intelligent simulation model for the coral reef beach sand evolution process; using the training data set to complete the training, verification and testing of the twin intelligent simulation model for the coral reef beach sand evolution process to obtain the target twin intelligent simulation model for the coral reef beach sand evolution process; applying the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period; wherein, the twin intelligent simulation model for the coral reef beach sand evolution process includes a beach sand feature extraction module, a spatio-temporal dependence capture module and a twin simulation module, wherein the beach sand feature extraction module adopts the cross-attention mechanism; the spatio-temporal dependence capture module consists of two parts, Swin Transformer and LSTM. The Swin Transformer part is used to vertically learn the global spatial dependence; the LSTM part is used to capture long-term and short-term temporal dependencies through the horizontal update of the information of the cell state and the hidden state.

[0013] In some exemplary embodiments, applying the twin intelligent simulation model for the evolution process of target coral reef beach sand to achieve an approximate simulation of the evolution process and terrain of coral reef beach sand in the historical and future specified time periods includes: inputting the data to be simulated into the twin intelligent simulation model for the evolution process of target coral reef beach sand; using the beach sand feature extraction module in the twin intelligent simulation model for the evolution process of target coral reef beach sand and adopting a cross-attention mechanism to obtain the initial input of the coral reef beach sand state at each time slice; dividing the initial input into uniformly sized and non-overlapping data chunks, flattening the data chunks in the order from left to right and from top to bottom, and inputting them into the spatio-temporal dependence capture module; using the Swin Transformer part in the spatio-temporal dependence capture module to obtain the global spatial dependence of the initial input; using the LSTM part in the spatio-temporal dependence capture module to obtain the long-term and short-term temporal dependencies of the initial input; based on the spatial and temporal dependencies, using the twin simulation module to twin the emergence probability at the specified time of the coral reef beach sand evolution; and based on the emergence probability, achieving an approximate simulation of the evolution process and terrain of coral reef beach sand in the historical or future specified time period, where the data to be simulated has the same data type as the training data in the training dataset.

[0014] In some exemplary embodiments, if the data to be simulated input is a dataset of beach sand state elements characterizing the evolution of coral reef beach sand in a certain time period of the historical period, the result obtained from the simulation of the evolution of coral reef beach sand is the evolution process in the specified historical period; if the data to be simulated input is a dataset of beach sand state elements characterizing the evolution of coral reef beach sand in a certain recent time period, the result obtained from the simulation of the evolution of coral reef beach sand is the evolution process in the specified future period.

[0015] In some exemplary embodiments, the original data includes remote sensing images, Automatic Identification System (AIS) data of ships, natural environmental dynamic element data, and the attribute of the monsoon period to which it belongs; the dataset of beach sand state elements contains beach sand state element data organized in the form of a spatial cube under multiple time slices, covering endogenous variables and exogenous variables, where the endogenous variables include the emergence probability; and the exogenous variables include natural environmental dynamic element data, artificial environmental dynamic element data, extreme weather data such as storm surges, and the attribute of the monsoon period to which it belongs.

[0016] In some exemplary embodiments, when adopting a cross-attention mechanism to obtain the initial input of the coral reef beach sand state at each time slice, the endogenous variables are used as queries, and the exogenous variables are used as keys and values.

[0017] In some exemplary embodiments, preprocessing the original data to construct a training dataset of beach sand state sequences in the form of a multi-dimensional space cube includes: extracting the original data of indirect elements from the original remote sensing images and Automatic Identification System (AIS) data, where the indirect element data includes the beach sand exposure probability, shipping intensity and vessel density, shoreline artificialization rate, and typhoon occurrence intensity; setting a grid cell framework with a specified spatial resolution and geographic coordinate system, and unifying the indirect element data and directly obtainable natural environmental dynamic element data under this grid cell framework; according to the calculation objectives of the indirect elements, using geographic information system (GIS) spatial analysis calculation methods and remote sensing big data cloud batch processing technologies to complete batch calculations and obtain the final calculation results of the indirect elements. Using one-hot encoding, convert the attribute of the monsoon period to a vector to obtain the attribute of the monsoon period represented by the vector; fuse the final calculation results of the indirect elements, directly obtainable natural environmental dynamic elements, and the attribute of the monsoon period represented by the vector to form beach sand state element data in the form of a multi-dimensional space cube, where each time slice includes a multi-dimensional space cube; among them, the beach sand state element dataset contains beach sand state element data organized in the form of space cubes under multiple time slices.

[0018] In some exemplary embodiments, the training, validation, and testing of the twin intelligent simulation model for the coral reef beach sand evolution process using a training dataset include: dividing the training dataset into a training set, a validation set, and a test set according to a preset ratio. Each sample in the training set, validation set, and test set includes a multi-dimensional spatial cube on T+L time slices. Here, T represents the number of front time slices, used to extract spatio-temporal features characterizing the beach sand evolution process, and L represents the number of rear time slices, which is the target to be simulated through the evolution process of beach sand on T time slices. T is numbered starting from 1, that is, the numbers of the front time slices are: 1, 2, …, T−1, T; L is numbered starting from T+1; that is, the numbers of the rear time slices are: T+1, T+2, …, L−1, L; both T and L are positive integers. Based on the state element data of coral reef beach sand organized in the form of spatial cubes on T time slices in the training set, use the twin intelligent simulation model for the coral reef beach sand evolution process to extract beach sand evolution features, and obtain the simulation results on the T+1 - T+L time slices; based on the simulation results and the sample labels, use the loss function to calculate the loss value; perform backpropagation based on the loss value to update the model weights, and obtain the intermediate twin intelligent simulation model for the coral reef beach sand evolution process; adopt an early stopping mechanism, use the validation set data to calculate the loss value of the intermediate twin intelligent simulation model for the coral reef beach sand evolution process until the change in the loss value of the validation set is within a preset range, stop training, and save the model with the highest accuracy on the validation set during the training process to obtain the trained twin intelligent simulation model for the coral reef beach sand evolution process to avoid overfitting; where the preset range means setting the number of training rounds as N. When the loss value after N rounds of the validation set is still rising, stop training and save the model with the highest accuracy on the validation set, where N is a positive integer; the sample label is the actual exposure probability grid of the coral reef beach sand on the T+1 - T+L time slices.

[0019] In some exemplary embodiments, it further includes: during multiple rounds of training, automatically adjust the global hyperparameter of the learning rate through the particle swarm optimization algorithm, use the Adam algorithm to update the local weights of the network, automatically train multiple twin intelligent simulation models for the coral reef beach sand evolution process, and select the twin intelligent simulation model for the coral reef beach sand evolution process with the smallest loss value on the validation set as the target twin intelligent simulation model for the coral reef beach sand evolution process.

[0020] In some exemplary embodiments, the loss function is:

[0021]

[0022] where Loss is the loss value, is the true value on grid (i,j), is the simulated value on grid (i,j), is the weight adjustment value on the grid (i,j), and K is the total number of grids within the beach sand simulation area;

[0023] Adjust in the following manner:

[0024] Calculate the cumulative change amount C 1 , X 2 ,…, X T} for each grid over the entire front time series: ij :

[0025]

[0026] The beach sand boundary area with high-frequency changes will produce a larger cumulative value.

[0027] Normalize the cumulative value to the preset weight range [δ,γ]:

[0028]

[0029] Automatically identify the sensitive areas of beach sand morphology changes through time series dynamic analysis. The change-sensitive areas include erosion / deposition boundaries, and assign higher loss weights to the change-sensitive areas, so that the spatial positions with drastic terrain fluctuations are focused on during model training.

[0030] According to the second aspect of the present invention, there is provided a twin intelligent simulation device for the evolution process of coral reef beach sand, including: an acquisition module for acquiring the original data of the process elements affecting the beach sand evolution, where the process elements include the main natural and artificial regional environmental characteristics driving the beach sand evolution and the exposure morphological characteristics of the beach sand itself; a preprocessing module for preprocessing the original data to construct a training data set of beach sand state sequences in the form of a multi-dimensional space cube; a model construction module for selecting the SwinLSTM deep learning framework with spatio-temporal dependence extraction ability suitable for beach sand evolution simulation, integrating the cross-attention mechanism, designing a loss function for the beach sand morphological change characteristics, and constructing a twin intelligent simulation model for the coral reef beach sand evolution process; a model training module for using the training data set to complete the training, verification and testing of the twin intelligent simulation model for the coral reef beach sand evolution process to obtain the target twin intelligent simulation model for the coral reef beach sand evolution process; a model application module for applying the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period; wherein, the twin intelligent simulation model for the coral reef beach sand evolution process includes a beach sand feature extraction module, a spatio-temporal dependence capture module and a twin simulation module, wherein the beach sand feature extraction module adopts the cross-attention mechanism; the spatio-temporal dependence capture module consists of two parts, the Swin Transformer and the LSTM. The Swin Transformer part is used to vertically learn the global spatial dependence; the LSTM part is used to capture the long-term and short-term temporal dependence through the horizontal update of the cell state and the hidden state information. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Through the following description of the embodiments of the present invention with reference to the drawings, the above content and other objects, features and advantages of the present invention will become clearer. In the drawings:

[0032] Figure 1 Schematically shows a flowchart of a twin intelligent simulation method for the evolution process of coral reef beach sand according to an embodiment of the present invention.

[0033] Figure 2 Schematically shows a flowchart of preprocessing the original data according to an embodiment of the present invention.

[0034] Figure 3 Schematically shows a schematic diagram of the beach sand exposure probability under different monsoon period attributes in the same area.

[0035] Figure 4 Schematically shows a schematic diagram of the architecture overview of the twin intelligent simulation model for the coral reef beach sand evolution process according to an embodiment of the present invention.

[0036] Figure 5 Schematically shows a schematic diagram of the beach sand exposure probability on a monthly basis in a certain area.

[0037] Figure 6 Schematically shows a twin intelligent simulation device for the evolution process of coral reef beach sand according to an embodiment of the present invention. Detailed implementation manners

[0038] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, many specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present invention. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily obscuring the concepts of the present invention.

[0039] The terms used herein are merely for describing specific embodiments and are not intended to limit the present invention. The terms "including", "comprising", etc. used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0040] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0041] In the case of using expressions such as "at least one of A, B, and C", generally, it should be interpreted according to the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include, but not be limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C).

[0042] Figure 1 Schematically shows a flowchart of a twin intelligent simulation method for the evolution process of coral reef beach sand according to an embodiment of the present invention.

[0043] As Figure 1 shown, a twin intelligent simulation method for the evolution process of coral reef beach sand according to an embodiment of the present invention includes step S110 - step S150.

[0044] In step S110, original data of the process elements affecting the evolution of beach sand is acquired, where the process elements include the main natural and artificial regional environmental characteristics driving the evolution of beach sand and the exposed morphological characteristics of the beach sand itself.

[0045] Optionally, the original data includes remote sensing images, Automatic Identification System (AIS) data of ships, natural environmental dynamic element data, and the attributes of the monsoon period to which it belongs.

[0046] In step S120, the original data is preprocessed to construct a training dataset of beach sand state sequences in the form of a multi-dimensional space cube.

[0047] In some exemplary embodiments, the training dataset of beach sand state sequences includes endogenous variables and exogenous variables. Among them, the endogenous variables include the exposure probability; and the exogenous variables include natural environmental dynamic element data, artificial environmental dynamic element data, extreme weather data such as storm surges, and the attributes of the monsoon period to which it belongs.

[0048] The exposure probability characterizes the exposure state of beach sand at each spatial position on the time slice and is the basic endogenous variable. The natural environmental dynamic elements mainly include marine hydrodynamic elements, sea winds, extreme weather disturbances, and periodic monsoons. Offshore coral reefs mainly evolve under the action of seawater dynamics such as ocean currents, sea waves, and tides, as well as the drive of sea winds. They transport coral sand and affect the exposure and morphology of coral reef beach sand. Typhoons are the main extreme weather affecting the evolution of beach sand at sea; the areas where coral reefs are located often have obvious periodic monsoon characteristics, making the evolution of coral reefs also periodic. In addition, due to human activities, the physical morphology of the shoreline of coral reef beach sand has been significantly changed, and human activities such as shipping will also affect the health of the coral reef itself, change the accretion rate, and thus affect the evolution process of beach sand.

[0049] In the embodiments of the present invention, the training dataset of the state sequence of the main process of coral reef beach sand evolution selects the exposure probability data representing the actual exposure state of beach sand, the basic natural environmental dynamic element data (sea wind, ocean current, sea wave, tide level), extreme weather data such as storm surges (long-term typhoon track frequency), the attributes of the monsoon period to which it belongs (northeast monsoon period, southwest monsoon period, or transition period), and artificial environmental dynamic element data (shipping intensity, vessel density, shoreline artificialization rate) for fusion construction.

[0050] In some exemplary embodiments, step S120 includes steps S121 - S125, see Figure 2 。

[0051] In step S121, from the original remote sensing images and Automatic Identification System (AIS) data of ships, the original data of indirect elements is extracted. Among them, the indirect element data includes the beach sand exposure probability, shipping intensity, vessel density, shoreline artificialization rate, and typhoon occurrence intensity. After the original data of the indirect elements is extracted, the final calculation results of the indirect elements are obtained using Geographic Information System (GIS) spatial analysis calculation methods and remote sensing cloud computing platforms. The GIS spatial analysis calculation methods include kernel density calculation, raster operation, raster reclassification, etc.

[0052] For example, for the acquisition of exposure probability data, on a monthly basis, a multi-band remote sensing image set with a spatial resolution of 10m * 10m of the target coral reef beach sand is obtained on the cloud server, and cloud cover screening conditions and a spatial range (a rectangle containing all possible exposed positions within the long time series of the beach sand) are set. For a single image, if each pixel in the image is exposed beach sand, it is assigned a value of 0, otherwise it is assigned a value of 1. The exposure probability raster for this month is statistically obtained pixel by pixel based on all the exposed binary rasters of the current month. The specific calculation formula for the exposure probability data P is as follows:

[0053] (1)

[0054] Among them, n is the number of computable rasters that meet the cloud cover conditions within a month, and p is the raster assignment of a single image.

[0055] Using ship AIS data, calculate the number of ships within a unit raster on a monthly basis, and use the natural break method to reclassify the results of the raster to reflect the shipping intensity.

[0056] Using ship AIS data, apply the spatial analysis method of kernel density to calculate the ship density.

[0057] If the beach sand is a sandy tidal flat affected by natural and human environmental dynamic actions on the periphery of the coral reef, there may be an artificialized shore section on the island reef where it is located. Calculate the artificialization rate of the coral reef shoreline where it is located. Extract the artificial shoreline and the entire coral reef shoreline, and calculate the lengths respectively. The specific calculation formula for the shoreline artificialization rate V is as follows:

[0058] (2)

[0059] Among them, is the total length of the artificial shoreline of the coral island reef where the beach sand is located, is the total length of the shoreline of the coral island reef where the beach sand is located.

[0060] If the beach sand is a sandbank, which is the previous geomorphic type in front of a stable exposed coral island, the value of the shoreline artificialization rate is 0.

[0061] Using typhoon track data, apply the spatial analysis method of kernel density to conduct density clustering analysis on typhoon track points on a monthly basis to obtain the analysis results of the typhoon occurrence intensity in this area.

[0062] In step S122, set a grid cell framework with a specified spatial resolution and geographic coordinate system, and unify the indirect element data and the directly acquirable natural environmental dynamic element data under this grid cell framework.

[0063] In the embodiments of the present invention, the indirect element data includes the element data calculated in step S121. The spatial range of all data is consistent with the calculation range of the exposure probability, which is the specified simulation range of coral reef beach sand.

[0064] In step S123, according to the calculation objective of the indirect element, using the spatial analysis calculation method of the geographic information system and the remote sensing big data cloud batch processing technology, batch calculation is completed to obtain the final calculation result of the indirect element.

[0065] In step S124, one-hot encoding is used to convert the attribute of the monsoon period to which it belongs into a vector, obtaining the attribute of the monsoon period to which it belongs represented by the vector.

[0066] Figure 3 Schematically shows the exposure probability of beach sand in the same area under different monsoon period attributes, where Figure 3 (a) in is the transitional non-monsoon period; Figure 3 (b) in is the southwest monsoon period; Figure 3 (c) in is the transitional non-monsoon period and Figure 3 (d) in is the northeast monsoon period. By comparing Figure 3 (a), (b), (c), and (d) in, it can be seen that the exposure probability of beach sand is also different under different monsoon period attributes. Therefore, it is necessary to consider the monsoon period attribute as an important influencing factor and use one-hot encoding to convert the attribute of the monsoon period to which it belongs into a vector, obtaining the attribute of the monsoon period to which it belongs represented by the vector.

[0067] In step S125, the final calculation result of the indirect element, the natural environmental dynamic elements that can be directly obtained, and the attribute of the monsoon period to which it belongs represented by the vector are fused to form the beach sand state element data in the form of a multi-dimensional space cube. Each time slice includes a multi-dimensional space cube representing the state of the beach sand at that time. The beach sand state element data set contains the beach sand state element data organized in the form of space cubes under multiple time slices.

[0068] The state element data in the evolution process of each processed coral reef beach sand is used as a channel. If an element is a single value within the specified beach sand area of the entire simulation, the grid value within the entire spatial range takes this value. Generally speaking, each beach sand state sequence training data set includes multiple samples, and each sample consists of T + L beach sand state element data. The element data represents the self-state of the coral reef beach sand and the natural and artificial geographical environment elements driving the process evolution at this time slice, forming a multi-dimensional space cube.

[0069] Return Figure 1, in step S130, select the SwinLSTM deep learning framework with spatio-temporal dependence extraction ability applicable to beach sand evolution simulation, fuse the cross-attention mechanism, design a loss function for the characteristics of beach sand morphology changes, and construct a twin intelligent simulation model for the coral reef beach sand evolution process. The architecture of the twin intelligent simulation model for the coral reef beach sand evolution process is referred to Figure 4 , such as Figure 4 shown, where the twin intelligent simulation model for the coral reef beach sand evolution process includes a beach sand feature extraction module, a spatio-temporal dependence capture module, and a twin simulation module. Among them, the beach sand feature extraction module uses the cross-attention mechanism; the spatio-temporal dependence capture module consists of two parts, Swin Transformer and LSTM. The Swin Transformer part is used to vertically learn the global spatial dependence; the LSTM part is used to capture long-term and short-term temporal dependencies through the information horizontal update of the cell state and the hidden state.

[0070] In step S140, use the training data set to complete the training, validation, and testing of the twin intelligent simulation model for the coral reef beach sand evolution process, and obtain the target twin intelligent simulation model for the coral reef beach sand evolution process.

[0071] In some exemplary embodiments, step S140 includes steps S141 - S145.

[0072] In step S141, divide the training data set into a training set, a validation set, and a test set according to a preset ratio. Among them, each sample in the training set, the validation set, and the test set includes a multi-dimensional spatial cube on T + L time slices. Among them, T represents the number of front time slices, which is used to extract the spatio-temporal features characterizing the beach sand evolution process, and L represents the number of rear time slices, which is the target to be simulated through the evolution process of the beach sand on T time slices. T is numbered starting from 1, that is, the numbers of the front time slices are: 1, 2,..., T−1, T; L is numbered starting from T + 1; that is, the numbers of the rear time slices are: T + 1, T + 2,..., L−1, L; both T and L are positive integers.

[0073] In step S142, based on the state element data of the coral reef beach sand organized in the form of a spatial cube on T time slices in the training set, use the twin intelligent simulation model for the coral reef beach sand evolution process to extract the beach sand evolution features, and obtain the simulation results on the T + 1 - T + L time slices.

[0074] In step S143, based on the simulation results and the sample labels, use the loss function to calculate the loss value, where the sample label is the actual emergence probability grid of the coral reef beach sand on the T + 1 - T + L time slices.

[0075] For example, the loss function is expressed as:

[0076] (3)

[0077] Where Loss is the loss value, is the true value on grid (i,j), is the simulated value on grid (i,j), is the weight adjustment value on grid (i,j), and K is the total number of grids in the beach sand simulation area;

[0078] Adjust as follows:

[0079] Calculate the cumulative change amount C 1 , X 2 ,…, X T} of each grid over the entire previous time series ij :

[0080] (4)

[0081] The beach sand boundary area with high-frequency changes will produce a larger cumulative value,

[0082] Normalize the cumulative value to the preset weight range [δ,γ]:

[0083] (5)

[0084] Where, is the weight adjustment value on grid (i,j), δ is the minimum value in the preset weight range, γ is the maximum value in the preset weight range, C ij is the cumulative change amount of {X 1 , X 2 ,…, X T} of each grid over the entire previous time series, min(C) is the minimum value of C ij among all grids in the beach sand simulation area, and max(C) is the maximum value of C ij among all grids in the beach sand simulation area.

[0085] Automatically identify the sensitive areas of beach sand morphology changes through time series dynamic analysis. The change-sensitive areas include erosion / deposition boundaries, and assign higher loss weights to the change-sensitive areas, so that the model focuses on the spatial positions with drastic terrain fluctuations during training. In step S144, backpropagation is performed based on the loss value to update the model weights, and an intermediate twin intelligent simulation model for the coral reef beach sand evolution process is obtained.

[0086] In step S145, an early stopping mechanism is adopted to calculate the loss value of the twin intelligent simulation model of the intermediate coral reef beach sand evolution process using the validation set data until the change in the loss value of the validation set is within a preset range. Then, the training is stopped, and the model with the highest accuracy on the validation set during the training process is saved to obtain the trained twin intelligent simulation model of the coral reef beach sand evolution process, so as to avoid overfitting.

[0087] On the basis of the above method, to further improve the effect of the model, the training method may further include automatically adjusting the global hyperparameter of the learning rate through the particle swarm optimization algorithm in multiple rounds of training, updating the local weights of the network using the Adam algorithm, automatically training multiple twin intelligent simulation models of the coral reef beach sand evolution process, and selecting the twin intelligent simulation model of the coral reef beach sand evolution process with the smallest loss value on the validation set as the target twin intelligent simulation model of the coral reef beach sand evolution process.

[0088] Return Figure 1 , in step S150, the target twin intelligent simulation model of the coral reef beach sand evolution process is applied to approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period.

[0089] In some exemplary embodiments, step S150 includes steps S151 - S157.

[0090] In step S151, the data to be simulated is input into the target twin intelligent simulation model of the coral reef beach sand evolution process, where the data type of the data to be simulated is the same as that of the training data in the training dataset.

[0091] In step S152, using the beach sand feature extraction module in the target twin intelligent simulation model of the coral reef beach sand evolution process and adopting the cross-attention mechanism, the initial input of the coral reef beach sand state on each time slice is obtained.

[0092] In the embodiments of the present invention, the data of the exposure probability layer is an endogenous variable, and the remaining variable layers: natural environmental dynamic element data (sea breeze, ocean current, sea wave, tidal level), artificial environmental dynamic element data (shipping intensity, vessel density, shoreline artificialization rate), extreme weather data such as storm surge (long-term typhoon track frequency), and the attribute of the monsoon period to which it belongs (northeast monsoon period, southwest monsoon period or transition period) are exogenous variables. The endogenous variable is used as the query, and the exogenous variables are used as the key and value to establish the connection between the two types of variables. The cross-attention mechanism is adopted, and the initial input of the coral reef beach sand state on the t time slice is obtained based on the following formula:

[0093] (6)

[0094] Among them, is the value of the endogenous variable at time slice t, is the value of the exogenous variable at time slice t, represents the multi - head cross - attention between the endogenous variable and the exogenous variable.

[0095] Let the length of the input time series of the model be T. The dataset of beach sand state elements within the T - time length consists of T time slices, and the values on each time slice are calculated by the above formula. . The length of the output time series is L, and the output value is the exposure probability of the beach sand at time L. There are data of T moments input and values of L moments output.

[0096] In step S153, the initial input is sliced into non - overlapping small data blocks of uniform size, and the small data blocks are flattened in the order from left to right and top to bottom for input into the spatio - temporal dependence capture module.

[0097] For example, within the time series length T, on each time slice t, after being calculated by the beach sand feature extraction module, is sliced into non - overlapping small blocks, each with a size of a*a, and the data is flattened in the order from left to right and top to bottom and input into the spatio - temporal dependence capture module in an embedded manner.

[0098] In step S154, using the Swin Transformer part in the spatio - temporal dependence capture module, the global spatial dependence of the initial input is obtained.

[0099] In step S155, using the LSTM part in the spatio - temporal dependence capture module, the long - term and short - term temporal dependencies of the initial input are obtained.

[0100] In step S156, based on the spatial dependence and temporal dependence, using the twin simulation module, the exposure probability of the coral reef beach sand at a specified time of evolution is twinned.

[0101] In step S157, based on the exposure probability, an approximate simulation of the evolution process and terrain of the coral reef beach sand in a specified historical or future time period is realized.

[0102] In steps S156 and S157, due to different time periods of the input data, the simulation results are also different. The simulation results are divided into two parts. One is the evolution process within a specified period of a coral reef beach sand in a twin historical period, and the other is the prediction of the evolution process of the coral reef beach sand in the future based on the evolution of a coral reef beach sand in the recent period. The model can input the beach sand state element data representing the evolution of the coral reef beach sand within the known T time, and predict the evolution of the coral reef beach sand in the time period (T + 1~T + L), characterized by the coral reef exposure probability.

[0103] Simulate the evolution process of coral reef beach sand during a specified historical period: Input the data of beach sand state elements representing the evolution of coral reef beach sand within the historical period time T, and simulate the evolution result of beach sand within the next continuous time L (exposure probability on a monthly scale).

[0104] Predict the evolution process of coral reef beach sand within a certain period in the future: Input the data of beach sand state elements representing the evolution of coral reef beach sand within the most recent time T, and use the model to predict the evolution result of beach sand within the future time L (exposure probability on a monthly scale).

[0105] The model result is the exposure probability of coral reef beach sand for each grid and each position on a monthly basis within the twin time period L. Based on the exposure probability, equal probability lines can be obtained. Based on the relationship between the initial beach sand simulation area terrain data, the exposure probability map, and the simulated exposure probability map, the terrain between the full probability line (p = 1) and the zero probability line (p = 0) of the coral reef beach sand is approximately simulated. (The coral reef area between the full probability line (p = 1) and the zero probability line (p = 0) will be submerged at high tide and exposed at low tide, which is an approximately intertidal zone area.) Approximate terrain simulation results can be obtained for the above two types of evolution processes. The exposure probability of beach sand in a certain area on a monthly basis is as Figure 5 shown.

[0106] In the embodiment of the present invention, the main element features affecting the evolution process of coral reef beach sand are effectively extracted as the input of the deep learning model, and the characteristics of coral reef beach sand are fully considered, such as natural and human multi-element driving, periodic change characteristics caused by the monsoon period, etc.

[0107] Aiming at the problem of simulating the evolution of the entire coral reef beach sand containing spatial information at the grid scale within a specified historical continuous time period and a specified future time period, which is difficult to solve by existing tidal flat evolution analysis methods, this method integrates the cross-attention mechanism and the SwinLSTM method combining Swin Transformer and LSTM, constructs a deep learning model, designs a loss function that focuses on grids with large changes, and realizes the twin intelligent simulation of the evolution process of far-sea coral reef beach sand, simulating the exposure probability and approximate terrain of coral reef beach sand within a specified time period.

[0108] Figure 6 Schematically shows a twin intelligent simulation device for the evolution process of coral reef beach sand according to an embodiment of the present invention.

[0109] As Figure 6 shown, the twin intelligent simulation device 800 for the evolution process of coral reef beach sand in this embodiment includes an acquisition module 810, a preprocessing module 820, a model construction module 830, a model training module 840, and a model application module 850.

[0110] An acquisition module 810 is configured to acquire original data of process elements affecting beach sand evolution, where the process elements include the main natural and human regional environmental characteristics driving beach sand evolution and the exposure morphological characteristics of the beach sand itself.

[0111] A preprocessing module 820 is configured to preprocess the original data and construct a training data set of beach sand state sequences in the form of a multi-dimensional space cube.

[0112] A model construction module 830 is configured to select a SwinLSTM deep learning framework with spatio-temporal dependence extraction ability suitable for beach sand evolution simulation, fuse a cross-attention mechanism, design a loss function for beach sand morphological change characteristics, and construct a twin intelligent simulation model for the coral reef beach sand evolution process.

[0113] A model training module 840 is configured to complete the training, verification, and testing of the twin intelligent simulation model for the coral reef beach sand evolution process using the training data set, and obtain a target twin intelligent simulation model for the coral reef beach sand evolution process.

[0114] A model application module 850 is configured to apply the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period.

[0115] Among them, the twin intelligent simulation model for the coral reef beach sand evolution process includes a beach sand feature extraction module, a spatio-temporal dependence capture module, and a twin simulation module. Among them, the beach sand feature extraction module adopts a cross-attention mechanism; the spatio-temporal dependence capture module consists of two parts: a Swin Transformer and an LSTM. The Swin Transformer part is used to vertically learn global spatial dependence; the LSTM part is used to capture long-term and short-term temporal dependence through the information level update of the cell state and the hidden state.

[0116] According to an embodiment of the present invention, any multiple of the acquisition module 810, the preprocessing module 820, the model construction module 830, the model training module 840, and the model application module 850 can be combined and implemented in one module, or any one of them can be split into multiple modules. Or, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module.

[0117] Those skilled in the art can understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present invention.

Claims

1. A twin intelligent simulation method for the evolution process of coral reef beach sand, characterized in that, The method includes: Obtaining the original data of the process elements affecting the evolution of beach sand, where the process elements include the main natural and artificial regional environmental characteristics driving the evolution of beach sand and the exposure morphological characteristics of the beach sand itself; Preprocessing the original data to construct a training data set of beach sand state sequences in the form of a multi-dimensional space cube; Selecting the SwinLSTM deep learning framework with spatio-temporal dependence extraction ability applicable to beach sand evolution simulation, integrating the cross-attention mechanism, designing a loss function for the morphological change characteristics of beach sand, and constructing a twin intelligent simulation model for the coral reef beach sand evolution process; Using the training data set to complete the training, verification and testing of the twin intelligent simulation model for the coral reef beach sand evolution process, and obtaining the target twin intelligent simulation model for the coral reef beach sand evolution process; Applying the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period; Among them, the twin intelligent simulation model for the coral reef beach sand evolution process includes a beach sand feature extraction module, a spatio-temporal dependence capture module, and a twin simulation module. Among them, the beach sand feature extraction module uses the cross-attention mechanism; the spatio-temporal dependence capture module consists of two parts: Swin Transformer and LSTM. The Swin Transformer part is used to vertically learn the global spatial dependence; the LSTM part is used to capture the long-term and short-term temporal dependencies through the information level update of the cell state and the hidden state. The application of the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period includes: Inputting the data to be simulated into the target twin intelligent simulation model for the coral reef beach sand evolution process; Using the beach sand feature extraction module in the target twin intelligent simulation model for the coral reef beach sand evolution process and adopting the cross-attention mechanism to obtain the initial input of the coral reef beach sand state at each time slice; Dividing the initial input into uniformly sized and non-overlapping data chunks, and flattening the data chunks in the order from left to right and from top to bottom to input into the spatio-temporal dependence capture module; Using the Swin Transformer part in the spatio-temporal dependence capture module to obtain the global spatial dependence of the initial input; Using the LSTM part in the spatio-temporal dependence capture module to obtain the long-term and short-term temporal dependencies of the initial input; Based on the spatial dependence and the temporal dependence, using the twin simulation module to twin the exposure probability at the specified time of the coral reef beach sand evolution; and Based on the exposure probability, approximately simulate the evolution process and terrain of the coral reef beach sand in a specified historical or future time period. Among them, the data type of the data to be simulated is the same as that of the training data in the training data set.

2. The method according to claim 1, wherein If the input data to be simulated is a dataset of beach sand state elements representing the evolution of coral reef beach sand during a certain period in the historical period, the result obtained from the simulation of the evolution of coral reef beach sand is the evolution process during the specified historical period; If the input data to be simulated is a dataset of beach sand state elements representing the evolution of coral reef beach sand during a certain recent period, the result obtained from the simulation of the evolution of coral reef beach sand is the evolution process during the specified future period.

3. The method according to claim 2, characterized in that The original data includes remote sensing images, Automatic Identification System (AIS) data, natural environmental dynamic element data, and the attributes of the monsoon period to which it belongs; The dataset of beach sand state elements contains beach sand state element data organized in the form of a spatial cube under multiple time slices, covering endogenous variables and exogenous variables. Among them, the endogenous variables include the exposure probability; and the exogenous variables include natural environmental dynamic element data, artificial environmental dynamic element data, extreme weather data such as storm surges, and the attributes of the monsoon period to which it belongs.

4. The method according to claim 3, characterized in that, When using the cross-attention mechanism to obtain the initial input of the coral reef beach sand state on each time slice, the endogenous variables are used as queries, and the exogenous variables are used as keys and values.

5. The method according to claim 3, characterized in that, The preprocessing of the original data to construct a training dataset of beach sand state sequences in the form of a multi-dimensional spatial cube includes: Extract the original data of indirect elements from the original remote sensing images and Automatic Identification System (AIS) data. Among them, the indirect element data includes the beach sand exposure probability, shipping intensity and vessel density, shoreline artificialization rate, and typhoon occurrence intensity; Set a grid cell framework with a specified spatial resolution and geographic coordinate system, and unify the indirect element data and the directly obtainable natural environmental dynamic element data into this grid cell framework; According to the calculation objectives of the indirect elements, use the spatial analysis calculation method of Geographic Information System (GIS) and the cloud batch processing technology of remote sensing big data to complete batch calculations and obtain the final calculation results of the indirect elements; Use one-hot encoding to convert the attributes of the monsoon period to which it belongs into a vector to obtain the attributes of the monsoon period to which it belongs represented by the vector; Fuse the final calculation results of the indirect elements, the directly obtainable natural environmental dynamic elements, and the attributes of the monsoon period to which it belongs represented by the vector to form beach sand state element data in the form of a multi-dimensional spatial cube. Each time slice includes a multi-dimensional spatial cube; Among them, the dataset of beach sand state elements contains beach sand state element data organized in the form of a spatial cube under multiple time slices.

6. The method according to claim 1, wherein The training, verification, and testing of the twin intelligent simulation model for the coral reef beach sand evolution process using the training dataset include: Divide the training data set into a training set, a validation set, and a test set according to a preset ratio. Each sample in the training set, the validation set, and the test set includes multi-dimensional spatial cubes on T+L time slices. Here, T represents the number of front time slices, which is used to extract spatio-temporal features characterizing the evolution process of beach sand. L represents the number of rear time slices, which is the target to be simulated through the evolution process of beach sand on T time slices. T is numbered starting from 1, that is, the numbers of the front time slices are: 1, 2, …, T−1, T; L is numbered starting from T+1; that is, the numbers of the rear time slices are: T+1, T+2, …, L−1, L; both T and L are positive integers; Based on the state element data of coral reef beach sand organized in the form of spatial cubes on T time slices in the training set, use the twin intelligent simulation model of the coral reef beach sand evolution process to extract the beach sand evolution features, and obtain the simulation results on the T+1-T+L time slices; Based on the simulation results and the sample labels, use the loss function to calculate the loss value; Based on the loss value, perform backpropagation to update the model weights to obtain an intermediate twin intelligent simulation model of the coral reef beach sand evolution process; Adopt an early stopping mechanism, use the validation set data to calculate the loss value of the intermediate twin intelligent simulation model of the coral reef beach sand evolution process until the change in the loss value of the validation set is within a preset range, stop training, and save the model with the highest accuracy on the validation set during the training process to obtain a trained twin intelligent simulation model of the coral reef beach sand evolution process to avoid overfitting; Among them, the preset range means that the number of training rounds is set to N. If the loss value after N rounds of the validation set is still rising, then stop training and save the model with the highest accuracy on the validation set, where N is a positive integer; The sample label is the actual exposure probability grid of the coral reef beach sand on the T+1-T+L time slices.

7. The method according to claim 6, characterized in that The method further includes: in multiple rounds of training, automatically adjust the global hyperparameter of the learning rate through the particle swarm optimization algorithm, use the Adam algorithm to update the local weights of the network, automatically train multiple twin intelligent simulation models of the coral reef beach sand evolution process, and select the twin intelligent simulation model of the coral reef beach sand evolution process with the smallest loss value on the validation set as the target twin intelligent simulation model of the coral reef beach sand evolution process.

8. The method according to claim 6, wherein The loss function is: where Loss is the loss value, is the true value on the grid (i, j), is the simulated value on the grid (i, j), is the weight adjustment value on the grid (i, j), and K is the total number of grids within the beach sand simulation area; Adjust as follows: Calculate the cumulative change amount of {X 1 , X 2 ,…, X T} for each grid over the entire front time series: The beach sand boundary area with high-frequency changes will generate a larger cumulative value, Normalize the cumulative value to the preset weight range [δ,γ]: Automatically identify the sensitive areas of beach sand morphological changes through time-series dynamic analysis. The sensitive areas of changes include erosion / siltation boundaries, and assign higher loss weights to the sensitive areas of changes, so that the model focuses on the spatial positions with drastic terrain fluctuations during training.

9. A twin intelligent simulation device for the evolution process of coral reef beach sand, characterized in that, The device includes: An acquisition module, configured to acquire the original data of the process elements affecting the beach sand evolution. The process elements include the main natural and human regional environmental characteristics driving the beach sand evolution and the exposure morphological characteristics of the beach sand itself; A preprocessing module, configured to preprocess the original data and construct a training data set of beach sand state sequences in the form of multi-dimensional spatial cubes; The model construction module is used to select the SwinLSTM deep learning framework with spatio-temporal dependence extraction ability applicable to beach sand evolution simulation, fuse the cross-attention mechanism, design a loss function for the characteristics of beach sand morphology changes, and construct a twin intelligent simulation model for the coral reef beach sand evolution process; The model training module is used to complete the training, verification and testing of the twin intelligent simulation model for the coral reef beach sand evolution process by using the training data set, and obtain the target twin intelligent simulation model for the coral reef beach sand evolution process; The model application module is used to apply the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in the historical or future specified time period; Among them, the twin intelligent simulation model for the coral reef beach sand evolution process includes a beach sand feature extraction module, a spatio-temporal dependence capture module and a twin simulation module. Among them, the beach sand feature extraction module adopts the cross-attention mechanism; the spatio-temporal dependence capture module consists of two parts: Swin Transformer and LSTM. The Swin Transformer part is used to vertically learn the global spatial dependence; the LSTM part is used to capture the long-term and short-term temporal dependencies through the information level update of the cell state and the hidden state. Applying the target twin intelligent simulation model for the coral reef beach sand evolution process to approximately simulate the evolution process and terrain of the coral reef beach sand in the historical or future specified time period includes: Inputting the data to be simulated into the target twin intelligent simulation model for the coral reef beach sand evolution process; Using the beach sand feature extraction module in the target twin intelligent simulation model for the coral reef beach sand evolution process and adopting the cross-attention mechanism to obtain the initial input of the coral reef beach sand state at each time slice; Dividing the initial input into uniformly sized and non-overlapping data chunks, and flattening the data chunks in the order from left to right and from top to bottom to input into the spatio-temporal dependence capture module; Using the Swin Transformer part in the spatio-temporal dependence capture module to obtain the global spatial dependence of the initial input; Using the LSTM part in the spatio-temporal dependence capture module to obtain the long-term and short-term temporal dependencies of the initial input; Based on the spatial dependence and the temporal dependence, using the twin simulation module to twin the emergence probability of the coral reef beach sand evolution at the specified time; and Based on the emergence probability, approximately simulate the evolution process and terrain of the coral reef beach sand in the historical or future specified time period. Among them, the data type of the data to be simulated is the same as that of the training data in the training data set.

Images