Sea ice area prediction method and device
By combining long and short-term memory networks and attention mechanism models, the problem of difficult computing resource consumption and nonlinear relationship capture in Arctic sea ice area prediction is solved, and efficient and accurate sea ice area prediction is achieved, providing an important role in climate change. support.
Patent Information
- Application Number
- CN202411179087.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-06
AI Technical Summary
The prior art requires a large amount of computing resources in the prediction of Arctic sea ice area, and empirically based statistical models cannot capture the nonlinear relationship between parameters, resulting in insufficient prediction accuracy and efficiency.
Using a combination of long and short-term memory network (LSTM) and attention mechanism model, the feature extraction and attention processing are obtained by obtaining historical sea ice area time series data to predict the current sea ice area, and the powerful time series processing capability and attention mechanism of LSTM are used. The ability to highlight key information and improve prediction accuracy and efficiency.
It realizes long-term dependence capture and adaptive learning of sea ice changes, improves the accuracy and efficiency of prediction, can more comprehensively understand the dynamic process of sea ice changes, has strong adaptability and robustness, and provides A powerful tool for climate change research.
Smart Images

Figure CN119089937A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of sea ice area prediction, and in particular to a sea ice area prediction method and device. Background Art
[0002] The Arctic region is a key link in the global ecosystem. The stability of the marine ecology, the global carbon cycle, and the living environment of marine life are closely linked to the state of Arctic sea ice. As the most active element in the Arctic region, Arctic sea ice plays a vital role in the process of sea-air heat exchange. The existence of sea ice limits the heat transfer between the atmosphere and the ocean, thereby affecting the thermal balance of seawater. In addition, Arctic sea ice affects the surface buoyancy flux in the North Atlantic deep water formation zone, which in turn acts on the Atlantic thermohaline circulation / Atlantic meridional overturning circulation. As global warming becomes increasingly obvious and global temperatures rise, the sea ice in the Arctic region shows a decreasing trend every year. Changes in Arctic sea ice are one of the key indicators to measure the extent of ongoing climate change. It has a far-reaching impact on polar ecosystems, Arctic shipping, and the global carbon cycle.
[0003] Accurately predicting the changes in Arctic sea ice extent (SIE) helps to understand the impact of the reduction in sea ice extent on local and global ecosystems. Many scholars have conducted research and analysis using three methods: numerical models, numerical models based on data assimilation, and statistical models. Numerical models integrate the ice-ocean-atmosphere interactions by controlling the physical equations of system dynamics and other model dynamics, and are widely used. Numerical models based on data assimilation assimilate satellite data to improve the prediction accuracy of numerical models. However, numerical models based on numerical and data assimilation require a lot of computing resources for Arctic sea ice prediction and rely on high-performance computing based on central processing unit (CPU) clusters, which cannot achieve lightweight prediction. Statistical models make predictions by establishing statistical relationships between sea ice extent, atmospheric conditions, ocean conditions, and sea ice variables. Statistical methods provide a lightweight method for predicting SIC, but they are empirical and cannot capture nonlinear relationships between parameters. Summary of the invention
[0004] The present invention aims to solve at least one of the technical problems existing in the prior art, and proposes a sea ice area prediction method and device.
[0005] In a first aspect, the present invention provides a method for predicting sea ice area, the method comprising the following steps:
[0006] Obtain historical sea ice area time series data for the target ocean area;
[0007] Inputting the historical sea ice area time series data into the trained long short-term memory network LSTM model for feature extraction to obtain sea ice area features in the time dimension;
[0008] The sea ice area characteristics of the time dimension are input into the attention mechanism model for attention processing to obtain the sea ice area attention characteristics of the time dimension, and the current sea ice area is predicted based on the sea ice area attention characteristics of the time dimension.
[0009] In some possible embodiments, the LSTM model includes a forget gate, an input gate, and an output gate;
[0010] The forget gate is used to forget the historical sea ice area data to be discarded from the cell state according to the currently input historical sea ice area data and the state at the previous moment;
[0011] The input gate is used to add the historical sea ice area data to be added from the cell state according to the historical sea ice area data input at the current moment and the state at the previous moment;
[0012] The output gate is used to determine the target historical sea ice area data to be output according to the cell state, and obtain the sea ice area characteristics of the time dimension.
[0013] In some possible embodiments, inputting the sea ice area feature in the time dimension into the attention mechanism model for attention processing to obtain the sea ice area attention feature in the time dimension, and predicting the current sea ice area based on the sea ice area attention feature in the time dimension includes:
[0014] Calculate the attention weight for each input position;
[0015] Multiply the representation of each input position by the corresponding attention weight and sum all weighted results;
[0016] The output is calculated based on the weighted input representation and other model parameters to predict the current sea ice extent.
[0017] In some possible embodiments, MAE is used as the loss function to optimize the LSTM model.
[0018] In some possible embodiments, the LSTM model includes a hidden layer of 128 neurons.
[0019] In some possible embodiments, the number of neurons in the first fully connected layer and the second fully connected layer of the attention mechanism model are 32 and 16 respectively.
[0020] In some possible embodiments, Adam is used as the optimizer for optimizing weights, and the learning rate and batch size are set to 0.001 and 128, respectively.
[0021] In a second aspect, the present invention provides a sea ice area prediction device, the device comprising:
[0022] An acquisition module is used to obtain the historical sea ice area time series data of the target ocean area;
[0023] An extraction module is used to input the historical sea ice area time series data into a trained long short-term memory network LSTM model for feature extraction to obtain sea ice area features in the time dimension;
[0024] The prediction module is used to input the sea ice area characteristics of the time dimension into the attention mechanism model for attention processing, obtain the sea ice area attention characteristics of the time dimension, and predict the current sea ice area based on the sea ice area attention characteristics of the time dimension.
[0025] In a third aspect, the present invention provides an electronic device, characterized in that it includes:
[0026] one or more processors;
[0027] A storage unit is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors can implement the method described above.
[0028] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, can implement the method described above.
[0029] The beneficial effects of the present invention are:
[0030] The sea ice area prediction method and device of the embodiment of the present invention are combined with the LSTM model and the attention mechanism model. The advantages of LSTM in sea ice prediction are mainly reflected in its powerful processing ability for time series data, which can capture the long-term dependence of sea ice changes, effectively solve the gradient vanishing problem, and have adaptability, which enables LSTM to learn the law of sea ice changes from complex historical data and provide support for accurate prediction. The attention mechanism further improves the accuracy and efficiency of prediction by highlighting key information, improving model interpretability, optimizing long sequence processing, and realizing partial parallel computing, so that the model can focus more on the important period of sea ice changes, thereby playing an important role in the sea ice prediction task. Combining LSTM and attention mechanism, the advantage of predicting sea ice is that it realizes the enhancement of learning ability and the improvement of prediction accuracy, so that the model can understand the dynamic process of sea ice changes more comprehensively and deeply, and has strong adaptability and robustness. This fusion not only improves the accuracy of prediction, but also provides a powerful tool for climate change research, helps to better cope with the challenges brought by sea ice changes, and provides important support for marine environmental protection and sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 A schematic diagram of the structure of an exemplary electronic device for a method for predicting sea ice area according to an embodiment of the present invention;
[0032] Figure 2 A flow chart of a method for predicting sea ice area according to another embodiment of the present invention;
[0033] Figure 3 A diagram showing a network model structure used in a sea ice area prediction method according to another embodiment of the present invention;
[0034] Figure 4 The figure is a schematic diagram of the structure of a sea ice area prediction device according to another embodiment of the present invention. DETAILED DESCRIPTION
[0035] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0036] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.
[0037] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0038] Figure 1 FIG. 1 is a schematic diagram of the structure of an exemplary electronic device for implementing a method for predicting sea ice area according to an embodiment of the present invention. Figure 1 As shown, the electronic device 100 includes one or more processors 110, one or more storage devices 120, one or more input devices 130, one or more output devices 140, etc. These components are interconnected via a bus system 150 and / or other forms of connection mechanisms. It should be noted that Figure 1 The components and structures of the electronic device shown are merely exemplary and non-limiting. The electronic device may also have other components and structures as required.
[0039] The processor 110 may be a central processing unit (CPU), or may be composed of multiple processing cores, or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 100 to perform desired functions.
[0040] The storage device 120 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may run the program instructions to implement the client functions (implemented by the processor) and / or other desired functions in the embodiments of the present disclosure described below. Various applications and various data may also be stored in the computer-readable storage medium, such as various data used and / or generated by the application.
[0041] The input device 130 may be a device used by a user to input instructions, and may include one or more of a keyboard, a mouse, a microphone, a touch screen, and the like.
[0042] The output device 140 may output various information (eg, images or sounds) to the outside (eg, a user), and may include one or more of a display, a speaker, and the like.
[0043] Figure 2 is a flow chart of a method for predicting sea ice area according to another embodiment of the present invention. Figure 3 FIG. 1 is a diagram of a network model structure used in a method for predicting sea ice area according to another embodiment of the present invention. Figure 2 and Figure 3 As shown, a method for predicting sea ice area includes the following steps S201 to S203:
[0044] Step S201: Acquire historical sea ice area time series data of the target ocean area.
[0045] Specifically, in this step, the daily SIE data obtained by passive sensors can be used as the historical sea ice area time series data of the target ocean area.
[0046] Step S202: input the historical sea ice area time series data into a trained long short-term memory network (LSTM) model for feature extraction to obtain sea ice area features in the time dimension.
[0047] Specifically, in this step, the LSTM model includes a forget gate, an input gate, and an output gate, which control the retention or discarding of information through specific mathematical operations (such as the sigmoid function). Specifically:
[0048] Forget gate: responsible for deciding which information needs to be discarded from the cell state. It determines which information is no longer needed by looking at the current input and the state of the previous moment, which helps the LSTM network forget old information and avoid information accumulation when processing sequence data.
[0049] Input gate: responsible for deciding which new information is added to the cell state. It determines which information is important based on the current input and the state at the previous moment, and adds it to the cell state. This helps the LSTM network remember important new information when processing sequence data.
[0050] Output gate: responsible for deciding which information is output to the current state. It determines which information is output based on the cell state, which helps the LSTM network to accurately output relevant information when processing sequence data.
[0051] Step S203: input the sea ice area characteristics in the time dimension into the attention mechanism model for attention processing to obtain the sea ice area attention characteristics in the time dimension, and predict the current sea ice area based on the sea ice area attention characteristics in the time dimension.
[0052] Specifically, in this step, the principle of the attention mechanism mainly involves three steps: calculating the attention weight, weighted summation of input representations, and calculating the output.
[0053] Calculating attention weights: This is the first step of the attention mechanism and involves calculating the attention weights for each input position. These weights can be weighted based on different parts of the input data, i.e. different weights are given to different parts. The calculation of weights is usually based on a function of the input data and model parameters, and can be calculated using different methods, such as dot product attention, additive attention, self-attention, etc.
[0054] Weighted sum input representation: After calculating the attention weights, the next step is to multiply the representation of each input position by the corresponding attention weight and sum all weighted results. This will result in a weighted input representation that better reflects the important parts of the input data.
[0055] Calculate output: The last step of the attention mechanism is to calculate the output result based on the weighted input representation and other model parameters as the final output, that is, the predicted current sea ice area.
[0056] The sea ice area prediction method of the embodiment of the present invention combines the LSTM model and the attention mechanism model. The advantages of LSTM in sea ice prediction are mainly reflected in its powerful processing ability for time series data, which can capture the long-term dependence of sea ice changes, effectively solve the gradient vanishing problem, and have adaptability, which enables LSTM to learn the law of sea ice changes from complex historical data and provide support for accurate prediction. The attention mechanism further improves the accuracy and efficiency of prediction by highlighting key information, improving model interpretability, optimizing long sequence processing, and realizing partial parallel computing, so that the model can focus more on the important period of sea ice changes, thereby playing an important role in the sea ice prediction task. The sea ice area prediction method of the embodiment of the present invention combines LSTM and the attention mechanism. The advantage of predicting sea ice is that it realizes the enhancement of learning ability and the improvement of prediction accuracy, so that the model can understand the dynamic process of sea ice changes more comprehensively and deeply, and has strong adaptability and robustness. This fusion not only improves the accuracy of prediction, but also provides a powerful tool for climate change research, helps to better cope with the challenges brought by sea ice changes, and provides important support for marine environmental protection and sustainable development.
[0057] The sea ice area prediction method of the embodiment of the present invention combines the attention mechanism with the LSTM model (LSTM_Att) ( Figure 3 ), the introduction of the attention mechanism enables the model to better capture the key features in the input data and improve the generalization and robustness of the model. The model takes 15 days of data as input units, and the output data is the predicted 1-day SIE. The module is built by two fully connected layers, and the last hidden layer outputs the weight within 15 days after undergoing the attention mechanism module. The hidden state is then multiplied by the corresponding weight to obtain the final output data. In the training phase, MAE is used as the loss function to better optimize the proposed network. For the LSTM in the network, there is a hidden layer containing 128 neurons. For the attention module, the number of neurons in the first and second fully connected layers are 32 and 16, respectively. Adam is selected as the optimizer for optimizing the weights, and the learning rate and batch size are set to 0.001 and 128, respectively.
[0058] Based on the same inventive concept, the present invention provides a sea ice area prediction device, which can be applied to the method described above. For details, please refer to the relevant records in the above text and will not be repeated here. Figure 4 FIG. 2 is a schematic diagram of the structure of a sea ice area prediction device according to another embodiment of the present invention. Figure 4 As shown, the device comprises:
[0059] The acquisition module 401 is used to obtain the historical sea ice area time series data of the target ocean area. The extraction module 402 is used to input the historical sea ice area time series data into the trained long short-term memory network LSTM model for feature extraction to obtain the sea ice area features in the time dimension. The prediction module 403 is used to input the sea ice area features in the time dimension into the attention mechanism model for attention processing to obtain the sea ice area attention features in the time dimension, and predict the current sea ice area based on the sea ice area attention features in the time dimension.
[0060] The sea ice area prediction device of the embodiment of the present invention combines the LSTM model and the attention mechanism model. The advantages of LSTM in sea ice prediction are mainly reflected in its powerful processing ability for time series data, which can capture the long-term dependence of sea ice changes, effectively solve the gradient vanishing problem, and have adaptability, which enables LSTM to learn the law of sea ice changes from complex historical data and provide support for accurate prediction. The attention mechanism further improves the accuracy and efficiency of prediction by highlighting key information, improving model interpretability, optimizing long sequence processing, and realizing partial parallel computing, so that the model can focus more on the important period of sea ice changes, thereby playing an important role in the sea ice prediction task. The sea ice area prediction device of the embodiment of the present invention combines LSTM and the attention mechanism. The advantage of predicting sea ice is that it realizes the enhancement of learning ability and the improvement of prediction accuracy, so that the model can understand the dynamic process of sea ice changes more comprehensively and deeply, and has strong adaptability and robustness. This fusion not only improves the accuracy of prediction, but also provides a powerful tool for climate change research, helps to better cope with the challenges brought by sea ice changes, and provides important support for marine environmental protection and sustainable development.
[0061] Another aspect of an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program can implement the method described above when executed by a processor.
[0062] The computer-readable medium may be included in the apparatus, device, or system of the present disclosure, or may exist independently.
[0063] Among them, computer-readable storage media can be any tangible media that contains or stores a program, which can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment. More specific examples include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, an optical fiber, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.
[0064] The computer-readable storage medium may also include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code, specific examples of which include but are not limited to electromagnetic signals, optical signals, or any suitable combination thereof.
[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting sea ice area, characterized in that: The method comprises the following steps: Obtain historical sea ice area time series data for the target ocean area; Inputting the historical sea ice area time series data into the trained long short-term memory network LSTM model for feature extraction to obtain sea ice area features in the time dimension; The sea ice area characteristics of the time dimension are input into the attention mechanism model for attention processing to obtain the sea ice area attention characteristics of the time dimension, and the current sea ice area is predicted based on the sea ice area attention characteristics of the time dimension.
2. The method according to claim 1, characterized in that The LSTM model includes a forget gate, an input gate and an output gate; The forget gate is used to forget the historical sea ice area data to be discarded from the cell state according to the currently input historical sea ice area data and the state at the previous moment; The input gate is used to add the historical sea ice area data to be added from the cell state according to the historical sea ice area data input at the current moment and the state at the previous moment; The output gate is used to determine the target historical sea ice area data to be output according to the cell state, and obtain the sea ice area characteristics of the time dimension.
3. The method according to claim 1, characterized in that The step of inputting the sea ice area feature in the time dimension into the attention mechanism model for attention processing to obtain the sea ice area attention feature in the time dimension, and predicting the current sea ice area based on the sea ice area attention feature in the time dimension includes: Calculate the attention weight for each input position; Multiply the representation of each input position by the corresponding attention weight and sum all weighted results; The output is calculated based on the weighted input representation and other model parameters to predict the current sea ice extent.
4. The method according to any one of claims 1 to 3, characterized in that: MAE is used as the loss function to optimize the LSTM model.
5. The method according to any one of claims 1 to 3, characterized in that: The LSTM model includes a hidden layer of 128 neurons.
6. The method according to claim 5, characterized in that The attention mechanism model has 32 and 16 neurons in the first fully connected layer and the second fully connected layer, respectively.
7. The method according to claim 6, characterized in that Adam is used as the optimizer for optimizing weights, and the learning rate and batch size are set to 0.001 and 128, respectively.
8. A sea ice area prediction device, characterized in that: The device comprises: An acquisition module is used to obtain the historical sea ice area time series data of the target ocean area; An extraction module is used to input the historical sea ice area time series data into a trained long short-term memory network LSTM model for feature extraction to obtain sea ice area features in the time dimension; The prediction module is used to input the sea ice area characteristics of the time dimension into the attention mechanism model for attention processing, obtain the sea ice area attention characteristics of the time dimension, and predict the current sea ice area based on the sea ice area attention characteristics of the time dimension.
9. An electronic device, characterized in that: include: one or more processors; A storage unit, used to store one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 can be implemented.
