Hybrid model sea ice range prediction method based on feature key capture mechanism

Abstract

The invention provides a hybrid model sea ice range prediction method based on a feature key capture mechanism, and the method comprises the steps: obtaining monthly ERA5 reanalysis data of sea ice concentration and nine atmospheric and meteorological variables as experimental data, obtaining monthly sea ice range data of an NSIDC as reference data, and carrying out the preprocessing of all data, so as to obtain needed standard data; a spatial feature extraction module is used for extracting spatial features hidden in the two-dimensional sea ice concentration data, and the features are fused with subsequent features; sea ice spatial features extracted from the spatial feature extraction module are fused with atmosphere and weather multivariable features, the fused features are input into the bidirectional local feature extraction module, and certain bidirectional relationships and dependencies only existing in a certain specific time or region are captured; the bidirectional time sequence feature extraction module continues to process the features obtained from the bidirectional local feature extraction module, and captures a complex relation between moments before and after the features; and finally, a feature key point capturing mechanism gives a higher weight to related information.

Description

Technical Field

[0001] The present invention relates to the field of satellite remote sensing and deep learning technology, and specifically to a hybrid model sea ice range prediction method based on a feature focus capture mechanism. Background Art

[0002] The changes in the extent of Arctic sea ice are closely related to the stability of the marine ecosystem, the global carbon cycle, the survival of Arctic wildlife, and the production and life of human society. Therefore, predicting the extent of Arctic sea ice is of great significance.

[0003] Sea ice extent prediction methods include those based on statistics, dynamic numerical models, and machine learning. Statistics methods use statistics to predict Arctic SIE, and find that sea ice concentration (SIC) is a key predictor for predicting SIE. Statistics methods mainly establish statistical models of sea ice variables (such as concentration, range, ice type, and thickness) over time or characteristic quantities based on the relationship between historical observation data (such as range, density, etc.), atmospheric conditions (temperature, sea level pressure, and clouds, etc.), and ocean conditions (such as sea surface temperature) to predict future sea ice extent. Its advantage is that the calculation is simple, but its disadvantage is that it does not take into account the physical mechanism of sea ice evolution. Dynamic numerical models rely on the understanding of the physics of the climate system. Under the constraints of the model dynamic framework, initial and boundary conditions, complex numerical methods are used to solve the equations of motion, thermodynamics, and radiation to simulate the changes in sea ice. Its advantage is that its prediction results are highly interpretable, but its disadvantage is that the parameterization scheme for the sea ice change process and mechanism is not yet perfect, the prediction results still have large uncertainties, the amount of calculation is large, and it is very dependent on the initial sea ice state.

[0004] Thanks to the release of high-quality satellite remote sensing data and atmospheric ocean reanalysis data such as the National Snow and Ice Data Center (NSIDC), the method of predicting sea ice extent based on machine learning has gradually become a popular method in this field in recent years. This method is mainly divided into single-variable machine learning prediction of sea ice extent and multi-variable machine learning prediction of sea ice extent. The single-variable method only uses past sea ice extent data as a variable to predict sea ice extent. With the development of research in the field of sea ice, many scholars have gradually observed and confirmed that there is a complex relationship between sea ice and atmospheric and meteorological variables (such as wind speed, sea level pressure, temperature, long-wave and short-wave radiation, etc.). Based on the above research content, the method of predicting sea ice extent based on multi-variable machine learning has been studied by more and more scholars, which predicts by integrating atmospheric and meteorological variables. However, the existing machine learning methods are all single models and fail to fully consider the important factor of sea ice concentration; in addition, the prediction effect of the existing algorithms in the sea ice melting season (i.e., May to October each year) is not satisfactory, especially how to effectively improve the SIE prediction accuracy of the minimum sea ice extent month-September each year is a problem to be solved by the current algorithm. In view of this, the present invention proposes a hybrid model sea ice extent prediction method based on feature focus capturing mechanism. Summary of the invention

[0005] The purpose of the present invention is to propose a hybrid model sea ice range prediction method based on a feature focus capture mechanism to solve the problems of existing sea ice range prediction methods, such as a single model, low sea ice prediction accuracy in the melting season, lack of complex nonlinear relationship extraction, and ignoring the importance differences between features.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A hybrid model sea ice extent prediction method based on feature-focused capture mechanism includes the following steps:

[0008] S1. Obtain monthly EAR5 reanalysis data of sea ice concentration and atmospheric and meteorological variables as experimental data, obtain monthly sea ice extent data of the National Snow and Ice Data Center (NSIDC) as reference data, and preprocess the experimental data and reference data to obtain the required standard data;

[0009] S2, using the spatial feature extraction module to extract the spatial features hidden in the two-dimensional sea ice concentration data, and fusing the extracted features with subsequent features as important predictors to predict future sea ice extent;

[0010] S3, the spatial features extracted in S2 are fused with the atmospheric and meteorological multivariate features, and the fused tensor is input into the bidirectional local feature extraction module to capture the bidirectional relationship and dependency that only exists in a certain time or region, such as the periodicity, trend and peak of the sequence change;

[0011] S4, using the bidirectional temporal feature extraction module to continue processing the features extracted by the bidirectional local feature extraction module in S3, capturing the complex relationship between the time before and after the features;

[0012] S5. Design a feature-focused capture mechanism to give higher weight to relevant information and reduce interference from irrelevant information.

[0013] Preferably, the monthly sea ice extent data obtained from the National Snow and Ice Data Center (NSIDC) and the atmospheric, meteorological, and sea ice concentration data obtained from EAR5 may have some missing values. For the experimental data and reference data obtained in S1, cubic spline interpolation is used to replace the missing values. The cubic spline interpolation is defined as follows:

[0014] For the known n+1 data points (x0, y0)(x1, y1),…,(x n ,y n ), the cubic polynomial S between each adjacent data point i (x) is expressed as: S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3 , and meet the following conditions:

[0015] (1)S i (x) is a cubic polynomial in each small interval;

[0016] (2)S i (x) has the same first-order and second-order derivatives at the connection between two adjacent small intervals, that is,

[0017] ″″″

[0018] S i (x i )=S i (x i+1 ),S i (x i )=S i (x i+1 );

[0019] (3)S i (x) passes through all given data points, i.e. S i (x i )=y i ;

[0020] ″″

[0021] (4) The boundary conditions are S0(x0) = 0 and S n (x n )=0;

[0022] Based on the above conditions, a linear equation system is established to solve the coefficient a i ,b i ,c i ,d i ;

[0023] After solving, when interpolating at a given point x, determine the interval [x i ,x i+1 ], using the corresponding cubic polynomial S i (x) is calculated to obtain the corresponding interpolation value.

[0024] Preferably, since different data have different dimensions, if data standardization is not performed, some characteristic data may be ignored. The experimental data and reference data are preprocessed to obtain the required standard data as described in S1, specifically including the following contents:

[0025] The data is standardized using the MinMax standardization technique. The MinMax standardization formula is as follows:

[0026]

[0027] Among them, Data represents the original data; Data min Indicates the minimum value in the original data; Data max Indicates the maximum value in the original data;

[0028] After the model predicts the sea ice extent data, the data is denormalized for observation; the denormalization formula is as follows:

[0029]

[0030] Among them, Data * represents the denormalized value, Represents the model prediction value.

[0031] Preferably, the spatial feature extraction module in S2 is implemented by a convolutional neural network (CNN), and the specific design is as follows:

[0032] The convolutional neural network (CNN) of the spatial feature extraction module includes four convolutional layer block structures, each of which includes a convolutional layer, a weight normalization layer, a ReLU layer, and a maximum pooling layer, wherein the convolution kernel size of the convolutional layer is 3×3, and the convolutional layer identifies spatial features such as edges and textures in the sea ice concentration data through convolution operations; the maximum pooling layer performs a pooling operation with a 2×2 window with a step size of 2 to reduce the dimension of the features and retain the main features;

[0033] The data format of the input convolutional neural network (CNN) is (10, 1, 1440, 261), where 10 is the batch size, 1 is the number of channels, and 1440 × 261 represents the longitude and latitude of the data. The data is subjected to multi-level feature extraction and dimensionality reduction processing, and the output feature shape is (10, 64, 90, 16);

[0034] The obtained feature representation is passed to two fully connected layers for further processing to output spatial features, and the final output feature shape is (10,64); the spatial features extracted by convolutional neural network (CNN) can more accurately capture the distribution pattern and spatial change trend of sea ice. These features can be fused with subsequent features as important predictors to predict future sea ice extent.

[0035] Preferably, the bidirectional local feature extraction module in S3 is implemented by a bidirectional temporal convolutional network (BiTCN), and the specific design is as follows:

[0036] Bidirectional temporal convolutional network (BiTCN) consists of two components: forward temporal convolution and reverse temporal convolution. Forward temporal convolution and reverse temporal convolution extract features of different scales by stacking multiple convolutional layers, thereby capturing features of different time ranges in sequence data.

[0037] The above multi-scale information extraction capability makes TCN more sensitive to local dependencies in sequence data and effectively extracts local features of time series;

[0038] The TCN consists of three parts: causal convolution, dilated convolution, and residual connection. The causal convolution and dilated convolution are combined to form causal dilated convolution. The causal dilated convolution is defined as:

[0039] Filter F=(f1,f2,…,f K ), time series X = (x1, x2, …, x T ), in x tThe dilated convolution formula with a dilation rate of d is as follows:

[0040]

[0041] The residual network is used to solve the problem of gradient explosion and gradient disappearance caused by network deepening. It directly adds the input of each layer to the output. The output formula is as follows:

[0042] o=ReLU(X+F(X)) (4)

[0043] Among them, o represents the output of the residual module; X represents the input data; F(X) represents the data after nonlinear transformation;

[0044] The causal dilation convolution has a total of 4 layers, the kernel is set to 2, and the dilation coefficient d of each layer is set to 1, 2, 4, and 8 respectively; a deeper model layer is used to obtain a higher and larger field of view; the bidirectional structure of the bidirectional temporal convolutional network (BiTCN) allows the model to extract data features from two time directions, and more accurately capture the local changes and dependencies of sequence information.

[0045] Preferably, the bidirectional temporal feature extraction module in S4 is implemented by a bidirectional gated recurrent unit (BiGRU), and the specific design is as follows:

[0046] The Bidirectional Gated Recurrent Unit (BiGRU) consists of two unidirectional GRUs in opposite directions; the GRU network is trained by two gating units: the update gate and the reset gate;

[0047] The function of the update gate is expressed as follows:

[0048] z t =σ(W z ·[h t-1 ,x t ]+b z ) (5)

[0049] The function of the reset gate is expressed as follows:

[0050] r t =σ(W r ·[h t-1 ,x t ]+b r ) (6)

[0051] The new candidate status formula is as follows:

[0052]

[0053] The formula for updating the hidden state is as follows:

[0054]

[0055] Among them, x t Indicates input data; h t Represents the output of the GRU unit; r t Represents the reset gate; z t represents the update gate; σ represents the Sigmoid function; W z ,W r ,W h They represent the update gate, reset gate, and weight matrix of candidate hidden states respectively; a bidirectional gated recurrent unit (BiGRU) is selected, which mines the input features in both sequential and reverse order in the time dimension to extract the complex dependency relationship between the features of the previous and next moments.

[0056] Preferably, the basic idea of ​​the attention mechanism of the feature focus capture mechanism is to filter out useful information by certain means, find factors that are more important to the prediction target, and thus improve the efficiency and reliability of the prediction; the feature focus capture mechanism described in S5 is implemented by the attention mechanism, and the specific design is as follows:

[0057] The weight calculation formula of the feature focus capture mechanism is as follows:

[0058]

[0059] Among them, W t represents the weight matrix; b t represents the bias coefficient; h t Represents the output of the BiGRU layer; l t represents the output of the attention layer;

[0060] The feature focus capture mechanism is used to enable the model to automatically learn the key information in the input sequence. By giving higher weights to relevant information, the model pays more attention to the parts related to the sea ice range and reduces interference with irrelevant information, thereby more accurately capturing the important information at each time step.

[0061] Compared with the prior art, the present invention provides a hybrid model sea ice extent prediction method based on feature focus capture mechanism, which has the following beneficial effects:

[0062] (1) In order to overcome the limitations of the single model algorithm structure on the prediction accuracy, the present invention proposes a hybrid model prediction mechanism. The mechanism is implemented through the following modules: a spatial feature extraction module based on a convolutional neural network (CNN); a bidirectional feature extraction module based on a bidirectional temporal convolutional network (BiTCN) and a bidirectional gated recurrent unit (BiGRU); and a feature focus capture mechanism based on the attention mechanism. The hybrid model plays its own model advantages and works together to achieve an overall improvement in the accuracy of sea ice range prediction.

[0063] (2) In order to solve the problem of low prediction accuracy of sea ice extent in the melting season, the present invention proposes for the first time to explore the relationship between sea ice concentration and sea ice extent. The spatial features of sea ice concentration data are extracted through a spatial feature extraction module and integrated with atmospheric and meteorological multivariate features. Monthly sea ice concentration data provide information on the spatial distribution and changes of sea ice. Feature extraction can capture the monthly spatial correlation of sea ice, thereby enhancing the prediction model's perception of changes in sea ice extent. At the same time, atmospheric and meteorological variables provide environmental background information on the formation and changes of sea ice. Feature fusion can comprehensively reflect the dynamic change process of sea ice, thereby achieving high-precision prediction of sea ice extent in the melting season.

[0064] (3) In order to solve the problem of limited feature extraction capability of unidirectional networks and lack of extraction of complex nonlinear relationships between features, the present invention proposes a bidirectional feature extraction module, including bidirectional local feature extraction and bidirectional temporal feature extraction, which fully mines the local relationships and temporal dependencies between data in both sequential and reverse directions, thereby enhancing the modeling capability of complex nonlinear relationships and temporal dependencies in data.

[0065] (4) In order to overcome the problem that existing algorithms ignore the differences in importance between features and cannot dynamically adjust feature weights, the present invention proposes a feature focus capture mechanism, which automatically learns the key information in the time series and gives higher weights to related features, so that the model can dynamically adapt to the changes in feature importance of different data and reduce the interference of irrelevant information. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is an overall structural diagram of the hybrid model sea ice range prediction method based on the feature focus capture mechanism mentioned in Example 1 of the present invention;

[0067] Figure 2 This is a heat map of the Pearson correlation coefficients between the atmosphere, meteorological variables, and sea ice concentration and sea ice extent mentioned in Example 1 of the present invention. DETAILED DESCRIPTION

[0069] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention are described in further detail below.

[0070] Embodiment 1:

[0071] See also Figure 1 ,This paper proposes a hybrid model sea ice extent prediction algorithm ,based on the feature focus capture mechanism, such as Figure 1 As shown, the following steps are included:

[0072] Step 1: Obtain monthly ERA5 reanalysis data of sea ice concentration and nine atmospheric and meteorological variables as experimental data, and monthly sea ice extent data of NSIDC as reference data, and preprocess all data to obtain the required standard data.

[0073] The present invention uses monthly sea ice concentration data and atmospheric and meteorological data covering the Arctic Ocean and adjacent land areas from 1979 to 2021 as experimental data, and monthly sea ice range data as reference data. The monthly sea ice concentration data is derived from the monthly ERA5 reanalysis data of the European Center for Medium-Range Weather Forecasts (ECMWF). The data set integrates the sea ice concentration monitoring results provided by the Ocean and Sea Ice Satellite Application Facility (OSI SAF) of the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT), including detailed information on the edge and concentration of sea ice. The data set is continuously verified by comparing navigation ice charts to ensure the accuracy and reliability of the data. The monthly sea ice concentration data used is in NetCDF format, which organizes data in conventional latitude and longitude grids, and optimizes the geometry of the grid using bilinear interpolation technology. This data grid covers the world, with a latitude range of -90 to +90 degrees (with reference to the equator), and a longitude range of 0 to 360 degrees (with reference to the Greenwich prime meridian). The resolution of each grid cell reaches 0.25 degrees.

[0074] A large number of studies have demonstrated the causal relationship between the interaction between Arctic sea ice and atmospheric and meteorological variables such as air pressure, wind speed, and sea surface temperature. Therefore, the present invention selected 9 atmospheric and meteorological data from the monthly ERA5 reanalysis data of the European Centre for Medium-Range Weather Forecasts (ECMWF). These variables include: 10m wind speed (10m U-component of wind, 10m V-component of wind, referred to as 10m_wind); surface pressure of the atmosphere on land, ocean and inland water (Surfacepressure); air temperature 2m above the surface of land, ocean or inland water (2m Temperature, referred to as t2m); the amount of thermal (also called longwave or ground) radiation emitted by the atmosphere and clouds that reaches the horizontal plane of the earth's surface (Surfacethermal radiation downwards, referred to as LW_down); the precipitation rate on the earth's surface (Mean totalprecipitation rate, referred to as rainfall); the snowfall rate on the earth's surface (Mean snowfall rate, referred to as snowfall); sea surface temperature (Sea surface temperature, referred to as sst); the amount of solar radiation reaching the horizontal plane of the earth's surface, also called shortwave radiation (Surface solar radiation downwards, referred to as SW_down); the mass of water vapor in each kilogram of humid air (Specific humidity).

[0075] Monthly sea ice extent data are derived from the monthly average of sea ice extent from the National Snow and Ice Data Center (NSIDC) Nimbus-7SSMR&DMSP SSM / I-SSMIS passive microwave data V2. These data are generated by the NASA Team algorithm developed by the Cryospheric Science Laboratory at NASA Goddard Space Flight Center (GSFC). Monthly sea ice extent data are divided into freezing season data and melting season data. Generally speaking, March belongs to the Arctic sea ice freezing season, which is the time of the year when the Arctic sea ice extent is the maximum; September belongs to the melting season, which is the time of the year when the Arctic sea ice extent is the minimum.

[0076] Figure 2The Pearson correlation coefficient heat map between selected atmospheric and meteorological variables and sea ice concentration and sea ice extent is shown. It can be seen from the figure that there is a positive or negative correlation between the nine atmospheric and meteorological variables and sea ice concentration and sea ice extent, but their influence on sea ice extent is different. The darker the color and the larger the absolute value, the more correlation, and the lighter the color and the smaller the absolute value, the less correlation. For example, sea surface temperature (SST) has the strongest negative correlation with sea ice extent, and sea ice concentration (SIC) has the strongest positive correlation with sea ice extent.

[0077] The data set preparation process in this embodiment is as follows:

[0078] (1) Replace missing values.

[0079] The monthly sea ice extent data obtained from NSIDC and the atmospheric, meteorological, and sea ice concentration data obtained from EAR5 may have some missing values. The reason for the missing values ​​is that some data points are not observed or recorded and the missing values ​​cannot be defined. Therefore, for the integrity of the data, interpolation is used to handle the missing values. Interpolation is a method of inferring the values ​​of unknown locations by using the values ​​of known data points. It can reasonably compensate for the missing values ​​in the data. This paper uses cubic spline interpolation, which uses cubic polynomials between each adjacent data point and ensures that these polynomials have continuous function values ​​and first-order and second-order derivative values ​​at each data point. The basic idea is to approximate the data in a small interval with a low-order polynomial, and use the connection conditions of the joints of these polynomials to ensure the smoothness of the entire curve. The cubic spline interpolation is defined as follows:

[0080] For the known n+1 data points (x0,y0),(x1,y1),…,(x n ,y n ), the cubic polynomial S between each adjacent data point i (x) is represented by S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3 , and meet the following conditions: S i (x) is a cubic polynomial in each small interval; S i (x) has the same first-order and second-order derivatives at the connection between two adjacent small intervals, that is, S i '(x i )=S i '(x i+1 ),Si ”(x i )=S i ”(x i+1 );S i (x) passes through all given data points, i.e. S i (x i )=y i The boundary conditions are S″0(x0)=0 and S″ n (x n )=0. Based on these conditions, a linear equation system can be established to solve the coefficient a i ,b i ,c i ,d i .

[0081] After solving, to interpolate at a given point x, you only need to determine the interval [x i ,x i+1 ], using the corresponding cubic polynomial S i (x) is calculated to obtain the corresponding interpolation value.

[0082] (2) Standardization

[0083] Since different data have different dimensions, if data standardization is not performed, some feature data may be ignored. Therefore, this paper uses MinMax standardization technology to standardize the data. Min-Max standardization is a common method for data standardization. It is a linear transformation of the original data, which maps the original value of a data into a value in the interval [0, 1]. The MinMax standardization formula is shown in Formula 1.

[0084]

[0085] Among them, Data represents the original data, Data min Represents the minimum value in the original data, Data max Represents the maximum value in the original data.

[0086] After the model predicts the sea ice extent data, the data needs to be denormalized for observation.

[0087] The denormalization formula is shown in Formula 2.

[0088]

[0089] Among them, Data * represents the denormalized value, Represents the model prediction value.

[0090] Step 2: Use the spatial feature extraction module to extract the hidden spatial features in the two-dimensional sea ice concentration data. This feature will be fused with subsequent features as an important predictor to predict the future sea ice extent.

[0091] The spatial feature extraction module used in the present invention is used to extract the hidden spatial features in the two-dimensional sea ice concentration data, and the module is implemented by a convolutional neural network (CNN). The spatial features extracted by this module will be used as an important factor in the subsequent prediction of the sea ice range. The CNN network used in this article contains four convolutional layer blocks. The convolution kernel size of the convolution layer is 3×3, and the convolution layer identifies the spatial features such as edges and textures in the sea ice concentration data through convolution operations. The maximum pooling layer performs a pooling operation with a 2×2 window with a step size of 2 to reduce the dimension of the features and retain the main features. The data format of the input CNN is (10, 1, 1440, 261), where 10 is the batchsize, 1 is the number of channels, and 1440×261 represents the longitude and latitude of the data. After multi-level feature extraction and dimensionality reduction processing, the output feature shape is (10, 64, 90, 16). Finally, the obtained feature representation is passed to two fully connected layers for further processing to output spatial features, and the final output feature shape is (10, 64).

[0092] The spatial features extracted by CNN can more accurately capture the distribution pattern and spatial variation trend of sea ice. These features can be fused with subsequent features as important predictors to predict future sea ice extent.

[0093] Step 3: The sea ice spatial features extracted from the spatial feature extraction module will be fused with 9 atmospheric and meteorological multivariate features, and the fused tensor will be further input into the bidirectional local feature extraction module to capture certain bidirectional relationships and dependencies that only exist in a specific time or region, such as the periodicity, trend and peak of sequence changes.

[0094] The sea ice spatial features extracted from the spatial feature extraction module will be fused with 9 atmospheric and meteorological multivariate features. Since the spatial features are (10, 64) and the format of the original atmospheric and meteorological variables is (10, 9), in order to enable the features to be fused with each other, the atmospheric and meteorological variables should be mapped to (10, 64) and then connected according to the feature dimension to obtain a tensor with a shape of (10, 128) after fusion. This tensor will continue to be input into the bidirectional local feature extraction module to capture some bidirectional relationships and dependencies that may only exist in a specific time or region, such as the periodicity, trend and peak of the sequence change. The bidirectional local feature extraction module is implemented through the BiTCN network. BiTCN consists of two main components: forward temporal convolution and reverse temporal convolution. Temporal convolution can extract features of different scales by stacking multiple convolutional layers. Thus, features of different time ranges in the sequence data are captured. This multi-scale information extraction capability makes TCN more sensitive to local dependencies in sequence data and can effectively extract local features of time series. TCN is a process of dilated causal convolution, which consists of three parts: causal convolution, dilated convolution, and residual connection.

[0095] TCN combines causal convolution with dilated convolution and uses causal dilated convolution. Dilated causal convolution is defined as: filter F = (f1, f2, ····, f K ), time series X=(x1,x2,····,x T ), in x t The dilated convolution with a dilation rate of d is shown in Equation 12.

[0096]

[0097] Residual connection solves the problem of gradient explosion and gradient vanishing caused by network deepening. Residual connection directly adds the input of each layer to the output. The output formula is shown in Equation 13. o is the output of the residual module, X represents the input data, and F(X) represents the data after nonlinear transformation.

[0098] o = ReLU(X + F(X)) (13)

[0099] The causal expansion convolution of BiTCN used in the present invention is set to 4 layers, the kernel is set to 2, and the expansion coefficient d of each layer is set to 1, 2, 4, and 8 respectively, so as to obtain a higher and larger field of view with a deeper model layer. The bidirectional structure of BiTCN allows the model to extract data features from two time directions, and more accurately capture the local changes and dependencies of sequence information.

[0100] Step 4: The bidirectional temporal feature extraction module will continue to process the features obtained from the bidirectional local feature extraction module to capture the complex relationship between the previous and next moments of the feature.

[0101] The bidirectional temporal feature extraction module will continue to process the features obtained from the bidirectional local feature extraction module to capture the long-term dependencies between features. The bidirectional temporal feature extraction module is implemented through the BiGRU network, which consists of two unidirectional GRUs in opposite directions. GRU is an improved network based on LSTM. Compared with the LSTM network, this network has fewer parameters and a simpler structure. LSTM has an input gate, an output gate, and a forget gate, while the GRU network updates the gate and resets the gate for training. This makes GRU not only inherit the advantages of LSTM, but also has a higher training speed. The update gate of GRU is responsible for determining when to update the hidden state, as shown in Formula 5.

[0102] z t =σ(W z ·[h t-1 ,x t ]+b z ) (5)

[0103] The reset gate is responsible for determining when to allow past information to affect the current hidden state, as shown in Equation 6.

[0104] r t =σ(W r ·[h t-1 ,x t ]+b r ) (6)

[0105] The new candidate state formula is shown in Equation 7.

[0106]

[0107] The formula for updating the hidden state is as follows:

[0108]

[0109] Where x t represents the input data, h t is the output of the GRU unit, r t is the reset gate, z t is the update gate, z t With r t Joint control from h t-1 to h t The calculation of the hidden state, the update gate controls both the current input data and the previous memory information h t-1 , output a value z between 0 and 1 t , z t Decide to what extent h t-1 Pass to the next state. σ is the Sigmoid function, Wz , W r , W h They are the update gate, reset gate and weight matrix of candidate hidden state. The reset gate controls h t-1 The result h t The importance of h t-1 When the correlation with the new memory is high, the reset gate can be used to enhance the influence of the previous memory. According to the calculation results of the reset gate, update gate and implicit state, the output h at the current moment can be obtained by the formula t .

[0110] The neuron state of the GRU neural network is transmitted layer by layer from front to back, without considering the dependency of data from back to front. The present invention selects a bidirectional gated recurrent unit BiGRU, which mines information on the input features in both sequential and reverse order in the time dimension, extracts the feature dependency between the previous and next moments, and then adds the outputs of the GRU units at different moments, which eventually becomes the output layer of the BiGRU. The GRU output node at each moment contains the complete forward and backward time series information of the current moment in the input sequence. Moreover, the existing methods for predicting SIE only extract the influence of the forward data on the backward data, and do not consider the complex dependency between the backward data and the forward data, resulting in poor modeling ability for the time series dependency in the data. Therefore, the present invention uses BiGRU, which can simultaneously capture the complex relationship between the previous and next moments of the feature, and has a more accurate prediction of the abnormal changes in sea ice in the melting season. BiGRU consists of two unidirectional GRUs in opposite directions, the forward GRU captures the past information, and the backward GRU captures the future information. The two GRUs jointly determine the output.

[0111] Step 5: The feature focus capture mechanism reduces the interference of irrelevant information by giving higher weight to relevant information.

[0112] The nine atmospheric and meteorological variables and sea ice concentration have different degrees of influence on the sea ice extent. In order to make the strongest correlated variable have the greatest impact on the sea ice extent and the weakest correlated variable have the least impact on the sea ice extent, this paper introduces a feature focus capture mechanism. This mechanism enables the model to automatically learn the key information in the input sequence. By giving higher weights to relevant information, the model can more effectively focus on the parts that are more relevant to the sea ice extent and reduce the interference of irrelevant information. In addition, although the aforementioned CNN can perform feature extraction and dimensionality reduction operations on time series, it cannot take into account the size of the influencing factors of each feature and screen out more useful features; BiGRU mines information on input features in both sequential and reverse order in the time dimension, but it also has the common problem of RNN networks, that is, when processing longer sequences, it cannot fully capture the important information of each time step, because the transmission of hidden states is carried out through time steps, and the information gradually weakens or disappears in time. Therefore, the introduction of a feature focus capture mechanism can effectively solve this problem, dynamically assign weights to input features, and thus more accurately capture the important information of each time step.

[0113] The feature focus capture mechanism is implemented through the attention mechanism. The basic idea of ​​the attention mechanism is to filter out useful information through certain means, find factors that are more important to the prediction target, and thus improve the efficiency and reliability of the prediction. The weight calculation formulas of the attention mechanism layer are shown in Equations 9, 10, and 11.

[0114] u t =tanh(W t h t +b t ) (9)

[0115]

[0116] l t =Σa t h t (12)

[0117]

[0118] Among them, W t represents the weight matrix, b t represents the bias coefficient, h t represents the output of the BiGRU layer, a t represents the probability distribution value of the BiGRU layer output, that is, the attention weight, l t Represents the output of the attention layer.

[0119] In order to verify the effectiveness and advancement of the present invention, a comparative test was conducted between the present invention and five existing sea ice range prediction algorithms. The experiment proved that the sea ice range prediction accuracy of the algorithm of the present invention reached 99.58%, especially in the melting season, the prediction accuracy reached 0.9934, and its prediction performance of sea ice range in September and the freezing season was also excellent, which fully demonstrated the effectiveness and advancement of the algorithm.

[0120] It should be noted that, in the present invention patent, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0121] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A hybrid model sea ice extent prediction method based on feature focus capture mechanism, characterized in that: The following steps are involved: S1. Obtain monthly EAR5 reanalysis data of sea ice concentration and atmospheric and meteorological variables as experimental data, obtain monthly sea ice extent data from the National Snow and Ice Data Center as reference data, and preprocess the experimental data and reference data to obtain the required standard data; S2, using the spatial feature extraction module to extract the spatial features hidden in the two-dimensional sea ice concentration data, and fusing the extracted features with subsequent features as important predictors to predict future sea ice extent; S3, the spatial features extracted in S2 are fused with the atmospheric and meteorological multivariate features, and the fused tensor is input into the bidirectional local feature extraction module to capture the bidirectional relationship and dependency that only exists in a specific time or region; S4, using the bidirectional temporal feature extraction module to continue processing the features extracted by the bidirectional local feature extraction module in S3, capturing the complex relationship between the time before and after the features; S5. Design a feature-focused capture mechanism to give higher weight to relevant information and reduce interference from irrelevant information.

2. The hybrid model sea ice extent prediction method based on feature focus capture mechanism according to claim 1 is characterized in that: For the experimental data and reference data obtained in S1, cubic spline interpolation is used to replace missing values. The cubic spline interpolation is defined as follows: For the known n+1 data points (x0, y0)(x1, y1),…,(x n ,y n ), the cubic polynomial S between each adjacent data point i (x) is expressed as: S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3 , and meet the following conditions: (1)S i (x) is a cubic polynomial in each small interval; (2)S i (x) has the same first-order and second-order derivatives at the connection between two adjacent cells, that is, ″″″ S i (x i )=S i (x i+1 ),S i (x i )=S i (x i+1 ); (3)S i (x) passes through all given data points, i.e. S i (x i )=y i ; ″″ (4) The boundary conditions are S0(x0) = 0 and S n (x n )=0; Based on the above conditions, a linear equation system is established to solve the coefficient a i ,b i ,c i ,d i ; After solving, when interpolating at a given point x, determine the interval [x i ,x i+1 ], using the corresponding cubic polynomial S i (x) is calculated to obtain the corresponding interpolation value.

3. The hybrid model sea ice extent prediction method based on feature focus capture mechanism according to claim 2 is characterized in that: The experimental data and reference data are preprocessed as described in S1 to obtain the required standard data, which specifically includes the following contents: The data is standardized using the MinMax standardization technique. The MinMax standardization formula is as follows: Among them, Data represents the original data; Data min Indicates the minimum value in the original data; Data max Indicates the maximum value in the original data; After the model predicts the sea ice extent data, the data is denormalized. The denormalization formula is as follows: Among them, Data * represents the denormalized value, Represents the model prediction value.

4. The hybrid model sea ice extent prediction method based on feature focus capture mechanism according to claim 1 is characterized in that: The spatial feature extraction module described in S2 is implemented by a convolutional neural network, and the specific design is as follows: The convolutional neural network of the spatial feature extraction module includes four convolutional layer block structures, each of which includes a convolutional layer, a weight normalization layer, a ReLU layer, and a maximum pooling layer, wherein the convolution kernel size of the convolutional layer is 3×3, and the convolutional layer identifies the spatial features in the sea ice concentration data through convolution operations; the maximum pooling layer performs a pooling operation with a 2×2 window with a step size of 2 to reduce the dimension of the features and retain the main features; The data format of the input convolutional neural network is (10, 1, 1440, 261), where 10 is the batch size, 1 is the number of channels, and 1440 × 261 represents the longitude and latitude of the data. The data is subjected to multi-level feature extraction and dimensionality reduction processing, and the output feature shape is (10, 64, 90, 16); The obtained feature representation is passed to two fully connected layers for further processing to output spatial features, and the final output feature shape is (10, 64).

5. The hybrid model sea ice extent prediction method based on feature focus capture mechanism according to claim 1 is characterized in that: The bidirectional local feature extraction module described in S3 is implemented by a bidirectional temporal convolutional network, and the specific design is as follows: The bidirectional temporal convolution network consists of two components: forward temporal convolution and reverse temporal convolution. The forward temporal convolution and reverse temporal convolution extract features of different scales by stacking multiple convolutional layers, thereby capturing features of different time ranges in sequence data. The above multi-scale information extraction capability makes TCN more sensitive to local dependencies in sequence data and effectively extracts local features of time series; The TCN consists of three parts: causal convolution, dilated convolution, and residual connection. The causal convolution and dilated convolution are combined to form causal dilated convolution. The causal dilated convolution is defined as: Filter F=(f1,f2,…,f K ), time series X = (x1, x2, …, x T ), in x t The dilated convolution formula with a dilation rate of d is as follows: The residual network is used to solve the problem of gradient explosion and gradient disappearance caused by network deepening. It directly adds the input of each layer to the output. The output formula is as follows: o=ReLU(X+F(X)) (4) Among them, o represents the output of the residual module; X represents the input data; F(X) represents the data after nonlinear transformation; The causal dilation convolution has a total of 4 layers, the kernel is set to 2, and the dilation coefficient d of each layer is set to 1, 2, 4, and 8 respectively.

6. The hybrid model sea ice extent prediction method based on feature focus capture mechanism according to claim 1 is characterized in that: The bidirectional temporal feature extraction module in S4 is implemented by a bidirectional gated recurrent unit, and the specific design is as follows: The bidirectional gated recurrent unit is composed of two unidirectional GRUs in opposite directions; the GRU network is trained by two gating units, the update gate and the reset gate; The function of the update gate is expressed as follows: z t =σ(W z ·[h t-1 ,x t ]+b z ) (5) The function of the reset gate is expressed as follows: r t =σ(W r ·[h t-1 ,x t ]+b r ) (6) The new candidate status formula is as follows: The formula for updating the hidden state is as follows: Among them, x t Indicates input data; h t Represents the output of the GRU unit; r t Represents the reset gate; z t represents the update gate; σ represents the Sigmoid function; W z ,W r ,W h Represent the weight matrices of the update gate, reset gate, and candidate hidden states respectively.

7. The hybrid model sea ice extent prediction method based on feature focus capture mechanism according to claim 1 is characterized in that: The feature focus capture mechanism described in S5 is implemented through the attention mechanism, and the specific design is as follows: The weight calculation formula of the feature focus capture mechanism is as follows: Among them, W t represents the weight matrix; b t represents the bias coefficient; h t Represents the output of the BiGRU layer; l t represents the output of the attention layer; The feature focus capture mechanism is used to enable the model to automatically learn key information in the input sequence. By giving higher weights to relevant information, the model pays more attention to the parts related to the sea ice range and reduces interference with irrelevant information.

Images