Sea temperature prediction method and device and electronic equipment

Through deep learning of the sea temperature prediction model, combined with the spatiotemporal characteristics of ocean temperature, the problem of low prediction accuracy of traditional numerical models is solved, and more accurate sea temperature prediction is achieved, which is suitable for marine resource management and climate change research.

CN118940210APending Publication Date: 2024-11-12HUANENG GUANGDONG SHANTOU OFFSHORE WIND POWER CO LTD +3
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410964252.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The traditional sea temperature prediction method is based on numerical models and is limited by complex ocean dynamics and meteorological processes, resulting in low prediction accuracy.

Method used

Adopting a deep learning-based sea temperature prediction model, combining the powerful fitting ability of neural networks and modeling ability of nonlinear relationships, comprehensively considering the spatiotemporal characteristics of ocean temperature, including ocean surface temperature, deep temperature distribution, current velocity, wind speed, etc. Factors, automatic learning and adapting to complex patterns and laws of ocean temperature changes.

Benefits of technology

It improves the accuracy of sea temperature prediction, can capture the dynamic changes of ocean temperature more accurately, and is suitable for predictions in different regions and seasons, providing important support for marine resource development, ecological protection and climate change research.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940210A_ABST
    Figure CN118940210A_ABST
Patent Text Reader

Abstract

The invention provides a sea temperature prediction method and device, and electronic equipment, and the method comprises the steps: obtaining a multi-group sample mapping relation in a target region; according to the feature set data and the label data corresponding to the sea temperature data indicated by each group of sample mapping relations, training a sea temperature prediction model to obtain a trained sea temperature prediction model; determining first sea temperature data and feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area; wherein each piece of real-time sea temperature data carries a time characteristic; and according to the feature set data corresponding to the first sea temperature data, using the trained sea temperature prediction model to predict second sea temperature data of which the time feature is after the time feature of the first sea temperature data. Wherein each group of sample mapping relations fully considers the spatial features and the time features of the sea temperature data, so that the sea temperature prediction model can more accurately capture the dynamic change process of the sea temperature, and the prediction accuracy of the sea temperature prediction model is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of ocean observation technology, and in particular to a sea temperature prediction method, device and electronic equipment. Background Art

[0002] Sea temperature (ocean temperature) is one of the key factors of the marine environment. It characterizes the absorption and release of heat by seawater, reflects the energy exchange in the global atmosphere and marine ecosystem, and has a great impact on many environmental and climate issues. Ocean temperature prediction is a key meteorological and oceanographic task, which is of great significance for marine resource management, shipping safety, fishery decision-making, and climate change research. In recent years, with the intensification of climate change, accurate prediction of ocean temperature has become particularly critical.

[0003] Traditional sea temperature prediction methods are mainly based on numerical models, but these models are usually limited by complex ocean dynamics and meteorological processes, resulting in unsatisfactory prediction accuracy. Summary of the invention

[0004] The present invention provides a sea temperature prediction method, device and electronic device to solve one of the technical problems in the related art at least to a certain extent. The technical solution of the present invention is as follows:

[0005] According to a first aspect of an embodiment of the present invention, a sea temperature prediction method is provided, comprising: obtaining multiple groups of sample mapping relationships within a target area; wherein any group of the sample mapping relationships is used to indicate the correspondence between feature set data and label data corresponding to any sea temperature data within the target area; training a sea temperature prediction model according to the feature set data and label data corresponding to each of the sea temperature data indicated by each group of the sample mapping relationships to obtain a trained sea temperature prediction model; determining first sea temperature data and feature set data corresponding to the first sea temperature data from real-time sea temperature data within the target area; wherein each of the real-time sea temperature data carries a time feature; and according to the feature set data corresponding to the first sea temperature data, using the trained sea temperature prediction model to predict second sea temperature data whose time feature is after the time feature of the first sea temperature data.

[0006] According to a second aspect of an embodiment of the present invention, a sea temperature prediction device is provided, comprising: an acquisition module, used to acquire multiple groups of sample mapping relationships in a target area; wherein any group of the sample mapping relationships is used to indicate the correspondence between feature set data and label data corresponding to any sea temperature data in the target area; a training module, used to train a sea temperature prediction model according to the feature set data and label data corresponding to each of the sea temperature data indicated by each group of the sample mapping relationships, so as to obtain a trained sea temperature prediction model; a determination module, used to determine first sea temperature data and feature set data corresponding to the first sea temperature data from real-time sea temperature data in the target area; wherein each of the real-time sea temperature data carries a time feature; a prediction module, used to predict second sea temperature data whose time feature is after the time feature of the first sea temperature data using the trained sea temperature prediction model according to the feature set data corresponding to the first sea temperature data.

[0007] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the sea temperature prediction method as described in the embodiment of the first aspect of the present invention.

[0008] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the sea temperature prediction method as described in the embodiment of the first aspect of the present invention.

[0009] According to a fifth aspect of an embodiment of the present invention, there is provided a computer program product, comprising: a computer program, wherein when the computer program is executed by a processor, the sea temperature prediction method as described in the embodiment of the first aspect of the present invention is implemented.

[0010] The technical solution provided by the embodiments of the present invention brings at least the following beneficial effects:

[0011] In this technical solution, multiple groups of sample mapping relationships in the target area are obtained; the sea temperature prediction model is trained according to the feature set data and label data corresponding to each sea temperature data indicated by each group of sample mapping relationships to obtain a trained sea temperature prediction model; the first sea temperature data and the feature set data corresponding to the first sea temperature data are determined from the real-time sea temperature data in the target area; wherein each real-time sea temperature data carries a time feature; according to the feature set data corresponding to the first sea temperature data, the trained sea temperature prediction model is used to predict the second sea temperature data whose time feature is after the time feature of the first sea temperature data. Among them, each group of sample mapping relationships fully considers the spatial and temporal features of the sea temperature data, which enables the sea temperature prediction model to more accurately capture the dynamic change process of the ocean temperature and effectively improve the prediction accuracy of the sea temperature prediction model.

[0012] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0014] Figure 1 A schematic diagram of a flow chart of a sea temperature prediction method provided by an embodiment of the present invention;

[0015] Figure 2 A schematic diagram of the principle of a sea temperature prediction model provided by an embodiment of the present invention;

[0016] Figure 3 A schematic diagram of the structure of a sea temperature prediction device provided by an embodiment of the present invention;

[0017] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0019] Traditional sea temperature prediction methods are mainly based on numerical models, but these models are usually limited by complex ocean dynamics and meteorological processes, resulting in unsatisfactory prediction accuracy. Machine learning methods can learn complex ocean temperature change patterns through a large amount of historical data, thereby achieving more accurate predictions. In this context, the present invention proposes a sea temperature prediction model based on deep learning, which combines the powerful fitting ability of neural networks and the effective modeling ability of nonlinear relationships, as well as professional knowledge in the field of oceanography.

[0020] In the present invention, the sea temperature prediction model utilizes the spatiotemporal characteristics of ocean temperature, comprehensively considers multiple factors such as ocean surface temperature, deep temperature distribution, ocean current speed, wind speed, and the complex interactions between them. By comprehensively analyzing and modeling these factors, the dynamic change process of ocean temperature can be captured more accurately, thereby achieving accurate prediction of future ocean temperature. Compared with traditional physical models, the sea temperature prediction model can automatically learn and adapt to the complex patterns and laws of ocean temperature changes, and has strong versatility in predictions for different regions and seasons.

[0021] In addition, the sea temperature prediction model can also provide important support and reference for marine resource development, ecological protection and climate change research. At the same time, the sea temperature prediction model also has a wide range of application prospects, and can be used for marine disaster warning, shipping route planning, fishery resource management, etc., which is of great significance to the development of marine science and technology.

[0022] The sea temperature prediction method, device and electronic device according to the embodiments of the present invention are described below with reference to the accompanying drawings.

[0023] Figure 1 A schematic flow chart of a sea temperature prediction method provided in an embodiment of the present invention.

[0024] like Figure 1 As shown, the sea temperature prediction method includes the following steps:

[0025] Step 101: Acquire mapping relationships of multiple groups of samples in a target area.

[0026] Among them, any set of sample mapping relationships is used to indicate the corresponding relationship between the feature set data and the label data corresponding to any sea temperature data in the target area.

[0027] The feature set data consists of continuous time series of sea temperature data, and the label data consists of one sea temperature data or continuous time series of sea temperature data. It should be noted that the number of sea temperature data in the feature set data corresponding to the sea temperature data indicated by each group of sample mapping relationships is the same, and the number of sea temperature data in the label data corresponding to the sea temperature data indicated by each group of sample mapping relationships is also the same.

[0028] In a possible implementation, when executing step 101 to obtain mapping relationships of multiple groups of samples in the target area, the following steps may be specifically performed:

[0029] Step 1011, obtaining a sea temperature dataset in a target area and a target period.

[0030] The sea temperature data set includes multiple sea temperature data, each of which carries longitude features, latitude features, time features and depth features.

[0031] Among them, the longitude feature is used to indicate the longitude of the sea temperature data, the latitude feature is used to indicate the latitude of the sea temperature data, the time feature is used to indicate the acquisition time of the sea temperature data, and the depth feature is used to indicate the depth level of the sea temperature data.

[0032] As a possible implementation method, for any sea temperature data, it is possible to determine whether the sea temperature data is in the target area based on the longitude indicated by the longitude feature of the sea temperature data and the latitude indicated by the latitude feature of the sea temperature data, and to determine whether the sea temperature data is within the target time period based on the acquisition time indicated by the time feature of the sea temperature data.

[0033] As an example, the depth levels may be 0m, 50m, 100m, 200m and 300m respectively, so that the depth feature of each sea temperature data in the sea temperature dataset may be any value among 0m, 50m, 100m, 200m and 300m.

[0034] Step 1012: determining at least one sea temperature data group from the sea temperature data set according to the depth characteristics of each sea temperature data in the sea temperature data set.

[0035] Among them, any sea temperature data group includes multiple sea temperature data (in the sea temperature data set).

[0036] As a possible implementation method, multiple sea temperature data with the same depth feature value in the sea temperature data set can be determined as a sea temperature data group according to the depth feature of each sea temperature data in the sea temperature data set. Thus, if there are several values ​​of the depth feature of each sea temperature data in the sea temperature data set, several groups of sea temperature data groups can be determined, and the depth feature values ​​of each sea temperature data in each sea temperature data group are the same.

[0037] Step 1013, for any sea temperature data group, determine the feature set data and label data corresponding to the target sea temperature data in the sea temperature data group according to the time characteristics of each sea temperature data in the sea temperature data group.

[0038] As a possible implementation method, for any sea temperature data group, each sea temperature data in the sea temperature data group can be normalized to obtain a normalized value corresponding to each sea temperature data in the sea temperature data group, thereby determining feature set data and label data corresponding to at least one target sea temperature data in the sea temperature data group from the normalized values ​​corresponding to each sea temperature data in the sea temperature data group based on the time characteristics of each sea temperature data in the sea temperature data group.

[0039] Optionally, the normalized value corresponding to each sea temperature data in each sea temperature data group may be determined by the following formula:

[0040] x normal =xx max

[0041] x max -x min

[0042] Among them, x normalrepresents the normalized value corresponding to any sea temperature data x in any sea temperature data set, x max represents the maximum value of the sea temperature data in any sea temperature data group, x min Indicates the minimum value of the sea temperature data in any sea temperature data group.

[0043] Optionally, when determining feature set data and label data from normalized values ​​corresponding to each sea temperature data in the sea temperature data group according to the time characteristics of each sea temperature data in the sea temperature data group, the first sea temperature data whose time characteristics are before the time characteristics of the target sea temperature data and the second sea temperature data whose time characteristics are after the time characteristics of the target sea temperature data can be determined for any target sea temperature data in the sea temperature data group according to the time characteristics of each sea temperature data in the sea temperature data group, wherein the target sea temperature data is the sea temperature data in which there are at least a first set number of first sea temperature data and at least a second set number of second sea temperature data in the sea temperature data group. Furthermore, the normalized value corresponding to the target sea temperature data and the normalized value corresponding to the first set number of first sea temperature data adjacent to the target sea temperature data can be determined as the feature set data corresponding to the target sea temperature data, and the normalized value corresponding to the second set number of second sea temperature data adjacent to the target sea temperature data can be determined as the label data corresponding to the target sea temperature data.

[0044] Among them, the first sea temperature data refers to the sea temperature data whose acquisition time indicated by the time characteristics is before the acquisition time indicated by the time characteristics of the target sea temperature data, and the second sea temperature data refers to the sea temperature data whose acquisition time indicated by the time characteristics is after the acquisition time indicated by the time characteristics of the target sea temperature data.

[0045] The first set number and the second set number can be any set value, which is not limited in the present invention. For example, assuming that the first set number is 12 and the second set number is 1, the target sea temperature data is the sea temperature data having at least 12 first sea temperature data and at least 1 second sea temperature data in the sea temperature data group, so that for any target sea temperature data, the normalized value corresponding to the target sea temperature data and the normalized values ​​corresponding to the 12 first sea temperature data adjacent to the target sea temperature data can be determined as the feature set data corresponding to the target sea temperature data, and the normalized value corresponding to the 1 second sea temperature data adjacent to the target sea temperature data can be determined as the label data corresponding to the target sea temperature data. Assuming that the target sea temperature data is the sea temperature data with an acquisition time of t indicated by the time feature, the 12 first sea temperature data adjacent to the target sea temperature data are the sea temperature data with an acquisition time of t-1 to t-12 indicated by the time feature, and the 1 second sea temperature data adjacent to the target sea temperature data is the sea temperature data with an acquisition time of t+1 indicated by the time feature.

[0046] In summary, the feature set data corresponding to each sea temperature data is the sea temperature data of the continuous time series including the sea temperature data, and the label data corresponding to each sea temperature data is the sea temperature data after the sea temperature data or the sea temperature data of the continuous time series.

[0047] Step 1014: for any target sea temperature data in the sea temperature data group, determine a sample mapping relationship between feature set data corresponding to the target sea temperature data and label data corresponding to the target sea temperature data.

[0048] Among them, the sample mapping relationship is used to train the sea temperature prediction model.

[0049] In summary, at least one sea temperature data group can be determined based on the spatial characteristics of multiple sea temperature data in the sea temperature data set within the target area and target period, and for each sea temperature data group, based on the time characteristics of each sea temperature data in the sea temperature data group, the feature set data and label data corresponding to at least one target sea temperature data are determined, so that at least one set of sample mapping relationships can be determined.

[0050] It should be noted that, in order to improve the prediction accuracy of the sea temperature prediction model in the present invention, in addition to determining at least one set of sample mapping relationships in the sea temperature data set acquired in the target area and target period, a sample generator can also be used to dynamically generate sea temperature data to increase the number of sample mapping relationships. In a possible implementation, the following steps can be specifically performed:

[0051] Step 1015, based on the feature set data and label data corresponding to the target sea temperature data in the sea temperature data group, train the sample generator corresponding to the sea temperature data group, so that the sample generator corresponding to the trained sea temperature data group learns the correspondence between the feature set data and the label data.

[0052] As a possible implementation method, for any target sea temperature data in the sea temperature data group, the feature set data corresponding to the target sea temperature data can be used as the input of the sample generator corresponding to the sea temperature data group, and the label data corresponding to the target sea temperature data can be used as the output of the sample generator corresponding to the sea temperature data group. The sample generator corresponding to the sea temperature data group can be trained so that the trained sample generator corresponding to the sea temperature data group can learn the correspondence between the feature set data and the label data.

[0053] Step 1016, determining a sample mapping relationship between at least one set of feature set data and label data based on each sea temperature data in the sea temperature data group and the sea temperature data generated by the sample generator corresponding to the trained sea temperature data group.

[0054] As a possible implementation, the normalized value corresponding to each sea temperature data in the sea temperature data group and / or the sea temperature data generated by the sample generator corresponding to the trained sea temperature data group can be used as the input of the sample generator corresponding to the trained sea temperature data group to obtain the sea temperature data output by the sample generator corresponding to the trained sea temperature data group. Then, the input of the sample generator corresponding to the trained sea temperature data group is used as the feature set data, and the output of the sample generator corresponding to the trained sea temperature data group is used as the label data to determine the sample mapping relationship between at least one set of feature set data and the label data.

[0055] Optionally, when the sample generator is used to generate sea temperature data, the last sea temperature data in the sea temperature data group can be used as the target sea temperature data, the feature set data corresponding to the target sea temperature data is determined, and the feature set data corresponding to the target sea temperature data is input into the sample generator corresponding to the trained sea temperature data group to obtain the sea temperature data output by the sample generator corresponding to the trained sea temperature data group. Optionally, the feature set data corresponding to the target sea temperature data can be the normalized value corresponding to the target sea temperature data and the normalized values ​​corresponding to the first set number of first sea temperature data adjacent to the target sea temperature data, wherein the first sea temperature data is the sea temperature data whose time feature is before the time feature of the target sea temperature data, so that the sea temperature data output by the sample generator corresponding to the trained sea temperature data group is the second set number of second sea temperature data adjacent to the target sea temperature data, wherein the second sea temperature data is the sea temperature data whose time feature is after the time feature of the target sea temperature data.

[0056] After that, any sea temperature data output by the sample generator corresponding to the trained sea temperature data group can be used as the target sea temperature data, the feature set data corresponding to the target sea temperature data can be determined, and the feature set data corresponding to the target sea temperature data can be input into the sample generator corresponding to the trained sea temperature data group to obtain the sea temperature data output by the sample generator corresponding to the trained sea temperature data group. Thus, by using the sample generator to dynamically generate sea temperature data in batches, the amount of sea temperature data can be increased, and then the number of sample mapping relationships can be increased, thereby improving the prediction accuracy of the sea temperature prediction model.

[0057] In summary, multiple groups of sample mapping relationships in the target area can be obtained, and each group of sample mapping relationships fully considers the spatial and temporal characteristics of the SST data, which is conducive to more accurately capturing the dynamic change process of the SST data.

[0058] Step 102 , training a sea temperature prediction model according to the feature set data and label data corresponding to each sea temperature data indicated by each group of sample mapping relationships, to obtain a trained sea temperature prediction model.

[0059] In a possible implementation, the sea temperature training model includes at least one first module, each first module includes a plurality of second modules, and when executing step 102, the sea temperature prediction model is trained according to the feature set data and label data corresponding to each sea temperature data indicated by each group of sample mapping relationship to obtain the trained sea temperature prediction model, the following steps may be specifically performed:

[0060] Step 1021, using the feature set data corresponding to the sea temperature data indicated by any set of sample mapping relationships as the input of the first second module in each first module to obtain the forward prediction result and the backward prediction result output by the first second module.

[0061] Among them, the forward prediction result refers to the result of predicting the future time series of sea temperature data along the time series development direction of the feature set data corresponding to the sea temperature data, and the backward prediction result refers to the result of predicting the historical time series of sea temperature data along the opposite direction of the time series development direction of the feature set data corresponding to the sea temperature data.

[0062] For example, assuming that the sea temperature data is the sea temperature data obtained at time t as indicated by the time feature, the feature set data corresponding to the sea temperature data is the sea temperature data obtained at time t-1 to t-12 as indicated by the time feature, and the label data corresponding to the sea temperature data is the sea temperature data obtained at time t+1 as indicated by the time feature. Since the label data is the sea temperature data after the sea temperature data, the sea temperature prediction model is used to predict a future sea temperature data after the sea temperature data. Therefore, the forward prediction result is the result of the sea temperature data obtained at time t+1 as indicated by the prediction time feature, and the backward prediction result is the result of the sea temperature data obtained at time t-13 as indicated by the prediction time feature.

[0063] As a possible implementation method, the second module includes a fully connected layer, a linear projection layer and a bias layer. The input of the second module can be subjected to feature integration through the fully connected layer, and then two branches are used to perform corresponding linear transformation through the linear projection layer to generate forward prediction coefficients and backward prediction coefficients respectively. The forward prediction coefficients and backward prediction coefficients are then respectively mapped through the bias layers of the corresponding branches to obtain the forward prediction results and backward prediction results output by the second module respectively.

[0064] Step 1022, for any second module except the first second module in each first module, determine the input of the second module according to the input of the previous second module of the second module and the backward prediction result of the previous second module output of the second module, so as to obtain the forward prediction result and the backward prediction result output by the second module.

[0065] As a possible implementation method, when determining the input of each second module except the first second module in each first module, the difference between the input of the previous second module of the second module and the backward prediction result of the previous second module output of the second module can be determined as the input of the second module.

[0066] Optionally, the process of the second module obtaining the forward prediction result and the backward prediction result output by the second module based on the input of the second module can be referred to the description of the above step 1021, which will not be repeated here.

[0067] Step 1023: for each first module, the forward prediction results output by each second module are integrated to obtain the forward prediction result output by the first module.

[0068] In the present invention, the output of each second module is a forward prediction result and a backward prediction result. The forward prediction results output by each second module in each first module are integrated to obtain the output result of the first module. That is, the forward prediction results output by the first module are obtained by integrating the forward prediction results output by each second module in each first module.

[0069] Step 1024, integrating the forward prediction results output by each first module to obtain the sea temperature prediction result output by the sea temperature prediction model.

[0070] Furthermore, by integrating the forward prediction results output by each first module, the sea temperature prediction result output by the sea temperature prediction model can be obtained.

[0071] As an example, Figure 2 As shown, the input of the sea temperature prediction model is the historical sea temperature data, and the output of the sea temperature prediction model is the future sea temperature data. The sea temperature prediction model is composed of multiple first modules, and each first module is composed of multiple second modules in series, wherein the input of the first second module of each first module is the input of the sea temperature prediction model (the feature set data corresponding to the sea temperature data indicated by any set of sample mapping relationships), and the output is two parts, one is the forward prediction result of the future sea temperature data, and the other is the backward prediction result of the historical sea temperature data. The input of the remaining second modules is the residual of the input of the previous second module minus the backward prediction result output by the previous second module. Afterwards, for each first module, the forward prediction results output by each second module in the first module are integrated to obtain the forward prediction result output by the first module, and then, the forward prediction results output by each first module are integrated to obtain the sea temperature prediction result output by the sea temperature prediction model.

[0072] Among them, each second module can output forward prediction results and backward prediction results based on the input through the following process: first, the input (time series data) passes through 4 layers of fully connected layers; then, after passing through the linear projection layer, a linear transformation is performed to generate forward prediction coefficients and backward prediction coefficients respectively; finally, the forward prediction coefficients and backward prediction coefficients are respectively output through the bias layer to output forward prediction and backward prediction results. The forward prediction result is the result of predicting the future sequence direction along the development direction of the time series. The backward prediction result is the restored historical time series, which is used for signal separation.

[0073] Therefore, the output of each second module is a forward prediction result and a backward prediction result. For any first module, the forward prediction results output by each second module in the first module are integrated to be the output result of the first module; the backward prediction results output by each second module in the first module exist between the second modules in the form of residuals, and the second modules are connected, which is conducive to the back propagation of the gradient and makes the prediction work of the next second module easier.

[0074] In summary, in the present invention, the residual transfer mechanism between the second modules in the first module can make each second module, as a supplement to the previous second module, constantly try to fit the residual information that the previous second module failed to capture, so that the second module inside each first module can gradually approach the real time series, which is conducive to improving the prediction performance of the sea temperature prediction model. In addition, although the inputs of different first modules are the same, their respective processing methods and purposes are different, resulting in different features and information captured by them. This allows the sea temperature prediction model to flexibly capture the complex features in the time series and further improve the accuracy of the prediction.

[0075] Step 1025 , training the sea temperature prediction model according to the difference between the sea temperature prediction result output by the sea temperature prediction model and the label data corresponding to the sea temperature data indicated by the sample mapping relationship, so as to obtain a trained sea temperature prediction model.

[0076] During the training process, the sea temperature prediction model can be trained based on the difference between the sea temperature prediction result output by the sea temperature prediction model and the label data corresponding to the sea temperature data indicated by the sample mapping relationship, thereby obtaining a trained sea temperature prediction model.

[0077] Step 103: determine first sea temperature data and feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area.

[0078] Among them, each real-time sea temperature data carries a time feature.

[0079] In some embodiments, after obtaining a trained sea temperature prediction model, feature set data corresponding to a sea temperature data can be determined from real-time sea temperature data in the target area, so as to predict future sea temperature data based on the feature set data using the trained sea temperature prediction model.

[0080] In a possible implementation, when executing step 103 to determine the first sea temperature data and the feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area, the following steps may be specifically performed:

[0081] Step 1031, selecting any sea temperature data from the real-time sea temperature data in the target area as the first sea temperature data.

[0082] Step 1032: determine a set number of consecutive sea temperature data whose time features are adjacent to the time features of the first sea temperature data and before the time features of the first sea temperature data and the first sea temperature data as feature set data corresponding to the first sea temperature data.

[0083] It should be noted that the number of sea temperature data in the feature set data corresponding to the first sea temperature data is the same as the number of sea temperature data in the feature set data corresponding to the sea temperature data indicated by each group of sample mapping relationships.

[0084] As an example, assuming that the number of sea temperature data in the feature set data corresponding to the sea temperature data indicated by each group of sample mapping relationships is 13, and the first sea temperature data is the sea temperature data with an acquisition time of t indicated by the time feature, then the sea temperature data with an acquisition time of t-1 to t-12 indicated by the time feature and the first sea temperature data can be determined as the feature set data corresponding to the first sea temperature data.

[0085] Step 104 , based on the feature set data corresponding to the first sea temperature data, a trained sea temperature prediction model is used to predict second sea temperature data whose time feature is subsequent to the time feature of the first sea temperature data.

[0086] The number of the second sea temperature data is the same as the number of the sea temperature data in the label data corresponding to the sea temperature data indicated by each group of sample mapping relationships.

[0087] As a possible implementation method, the feature set data corresponding to the first sea temperature data can be input into a trained sea temperature prediction model to obtain second sea temperature data whose time characteristics output by the trained sea temperature prediction model are after the time characteristics of the first sea temperature data.

[0088] In summary, by obtaining multiple groups of sample mapping relationships in the target area; training the sea temperature prediction model according to the feature set data and label data corresponding to each sea temperature data indicated by each group of sample mapping relationships to obtain a trained sea temperature prediction model; determining the first sea temperature data and the feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area; wherein each real-time sea temperature data carries a time feature; according to the feature set data corresponding to the first sea temperature data, using the trained sea temperature prediction model, predict the second sea temperature data whose time feature is after the time feature of the first sea temperature data. Among them, each group of sample mapping relationships fully considers the spatial and temporal features of the sea temperature data, which enables the sea temperature prediction model to more accurately capture the dynamic change process of the ocean temperature and effectively improve the prediction accuracy of the sea temperature prediction model.

[0089] In order to implement the above embodiment, the present invention also proposes a sea temperature prediction device.

[0090] Figure 3 A schematic diagram of the structure of a sea temperature prediction device provided in an embodiment of the present invention.

[0091] like Figure 3 As shown, the sea temperature prediction device includes: an acquisition module 31, a first training module 32, a first determination module 33 and a prediction module 34.

[0092] An acquisition module 31 is used to acquire multiple groups of sample mapping relationships in the target area; wherein any group of sample mapping relationships is used to indicate the corresponding relationship between feature set data and label data corresponding to any sea temperature data in the target area;

[0093] The first training module 32 is used to train the sea temperature prediction model according to the feature set data and label data corresponding to each sea temperature data indicated by each group of sample mapping relationships, so as to obtain a trained sea temperature prediction model;

[0094] A first determination module 33 is used to determine first sea temperature data and feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area; wherein each real-time sea temperature data carries a time feature;

[0095] The prediction module 34 is used to predict second sea temperature data whose time feature follows the time feature of the first sea temperature data by using a trained sea temperature prediction model according to the feature set data corresponding to the first sea temperature data.

[0096] Further, in a possible implementation of the embodiment of the present invention, the sea temperature training model includes at least one first module, each first module includes a plurality of second modules; the first training module 32 includes:

[0097] A first processing unit, used to use the feature set data corresponding to the sea temperature data indicated by any set of sample mapping relationships as input to the first second module in each first module, so as to obtain a forward prediction result and a backward prediction result output by the first second module;

[0098] a second processing unit, for determining, for any second module except the first second module in each first module, an input of the second module according to an input of a previous second module of the second module and a backward prediction result output by a previous second module of the second module, so as to obtain a forward prediction result and a backward prediction result output by the second module;

[0099] A first integration unit is used to integrate the forward prediction results output by each second module for each first module to obtain the forward prediction result output by the first module;

[0100] A second integration unit is used to integrate the forward prediction results output by each first module to obtain a sea temperature prediction result output by the sea temperature prediction model;

[0101] The training unit is used to train the sea temperature prediction model according to the difference between the sea temperature prediction result output by the sea temperature prediction model and the label data corresponding to the sea temperature data indicated by the sample mapping relationship, so as to obtain a trained sea temperature prediction model.

[0102] Furthermore, in a possible implementation manner of the embodiment of the present invention, the second processing unit is further configured to:

[0103] The difference between the input of the previous second module of the second module and the backward prediction result of the previous second module output of the second module is determined as the input of the second module.

[0104] Further, in a possible implementation of the embodiment of the present invention, the second module includes a fully connected layer, a linear projection layer and a bias layer; the first training module 32 is further used for:

[0105] The input of the second module is passed through a fully connected layer for feature integration, and then two branches are used to perform corresponding linear transformations through a linear projection layer to generate forward prediction coefficients and backward prediction coefficients respectively;

[0106] The forward prediction coefficient and the backward prediction coefficient are respectively mapped through the bias layer of the corresponding branch to obtain the forward prediction result and the backward prediction result output by the second module respectively.

[0107] Furthermore, in a possible implementation of the embodiment of the present invention, the acquisition module 31 includes:

[0108] An acquisition unit is used to acquire a sea temperature data set in a target area and a target period; wherein the sea temperature data set includes a plurality of sea temperature data, each of which carries a longitude feature, a latitude feature, a time feature and a depth feature;

[0109] A first determination unit is used to determine at least one sea temperature data group from the sea temperature data set according to the depth characteristics of each sea temperature data in the sea temperature data set; wherein any sea temperature data group includes a plurality of sea temperature data;

[0110] A second determination unit is used to determine, for any sea temperature data group, feature set data and label data corresponding to at least one target sea temperature data in the sea temperature data group according to the time characteristics of each sea temperature data in the sea temperature data group;

[0111] The third determination unit is used to determine, for any target sea temperature data in the sea temperature data group, a sample mapping relationship between feature set data corresponding to the target sea temperature data and label data corresponding to the target sea temperature data.

[0112] Furthermore, in a possible implementation of the embodiment of the present invention, the above-mentioned device further includes:

[0113] The second training module is used to train the sample generator corresponding to the sea temperature data group according to the feature set data and label data corresponding to the target sea temperature data in the sea temperature data group, so that the sample generator corresponding to the trained sea temperature data group learns the corresponding relationship between the feature set data and the label data;

[0114] The second determination module is used to determine the sample mapping relationship between at least one set of feature set data and label data based on each sea temperature data in the sea temperature data group and the sea temperature data generated by the sample generator corresponding to the trained sea temperature data group.

[0115] Further, in a possible implementation of the embodiment of the present invention, the sea temperature data carries a time feature; the first determination module 33 is further used to:

[0116] Select any sea temperature data from the real-time sea temperature data in the target area as the first sea temperature data;

[0117] A set number of continuous sea temperature data whose time characteristics are adjacent to the time characteristics of the first sea temperature data and before the time characteristics of the first sea temperature data and the first sea temperature data are determined as feature set data corresponding to the first sea temperature data.

[0118] Furthermore, in a possible implementation of the embodiment of the present invention, the prediction module 34 is further configured to:

[0119] The feature set data corresponding to the first sea temperature data is input into the trained sea temperature prediction model to obtain the second sea temperature data output by the trained sea temperature prediction model.

[0120] It should be noted that the aforementioned explanation of the sea temperature prediction method embodiment is also applicable to the sea temperature prediction device of this embodiment, and will not be repeated here.

[0121] The sea temperature prediction device of the embodiment of the present invention obtains multiple groups of sample mapping relationships in the target area; trains the sea temperature prediction model according to the feature set data and label data corresponding to each sea temperature data indicated by each group of sample mapping relationships to obtain a trained sea temperature prediction model; determines the first sea temperature data and the feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area; wherein each real-time sea temperature data carries a time feature; and according to the feature set data corresponding to the first sea temperature data, uses the trained sea temperature prediction model to predict the second sea temperature data whose time feature is after the time feature of the first sea temperature data. Wherein, each group of sample mapping relationships fully considers the spatial features and time features of the sea temperature data, which enables the sea temperature prediction model to more accurately capture the dynamic change process of the ocean temperature, and effectively improves the prediction accuracy of the sea temperature prediction model.

[0122] In order to implement the above embodiments, the present invention also proposes an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the sea temperature prediction method proposed in any of the above embodiments of the present invention.

[0123] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0124] It should be noted that Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0125] like Figure 4 As shown, the electronic device includes:

[0126] A memory 41 , a processor 42 , and a computer program stored in the memory 41 and executable on the processor 42 .

[0127] When the processor 42 executes the program, the sea temperature prediction method provided in any of the above embodiments is implemented.

[0128] Furthermore, the electronic device further comprises:

[0129] The communication interface 43 is used for communication between the memory 41 and the processor 42 .

[0130] The memory 41 is used to store computer programs that can be run on the processor 42 .

[0131] The memory 41 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0132] The processor 42 is used to implement the sea temperature prediction method described in any of the above embodiments when executing the program.

[0133] If the memory 41, the processor 42 and the communication interface 43 are implemented independently, the communication interface 43, the memory 41 and the processor 42 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0134] Optionally, in a specific implementation, if the memory 41, the processor 42 and the communication interface 43 are integrated on a chip, the memory 41, the processor 42 and the communication interface 43 can communicate with each other through an internal interface.

[0135] The processor 42 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0136] In order to implement the above embodiments, the present invention further proposes a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the sea temperature prediction method proposed in any of the above embodiments of the present invention.

[0137] In order to implement the above embodiments, the present invention further proposes a computer program product, including a computer program, which, when executed by a processor, implements the sea temperature prediction method proposed in any of the above embodiments of the present invention.

[0138] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0139] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0140] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0141] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0142] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0143] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0144] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0145] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A sea temperature prediction method, characterized in that: include: Acquire multiple groups of sample mapping relationships in the target area; wherein any group of the sample mapping relationships is used to indicate the corresponding relationship between the feature set data and the label data corresponding to any sea temperature data in the target area; Training a sea temperature prediction model according to the feature set data and label data corresponding to each of the sea temperature data indicated by each group of the sample mapping relationships to obtain a trained sea temperature prediction model; Determining first sea temperature data and feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area; wherein each of the real-time sea temperature data carries a time feature; The trained sea temperature prediction model is used according to the feature set data corresponding to the first sea temperature data to predict second sea temperature data whose time feature is after the time feature of the first sea temperature data.

2. The method according to claim 1, characterized in that The sea temperature training model includes at least one first module, each of which includes a plurality of second modules; the sea temperature prediction model is trained according to the feature set data and label data corresponding to each of the sea temperature data indicated by each group of the sample mapping relationships to obtain a trained sea temperature prediction model, including: Using the feature set data corresponding to the sea temperature data indicated by any set of the sample mapping relationships as the input of the first of the second modules in each of the first modules, so as to obtain the forward prediction result and the backward prediction result output by the first of the second modules; For any second module except the first second module in each of the first modules, determining the input of the second module according to the input of the previous second module of the second module and the backward prediction result output by the previous second module of the second module, so as to obtain the forward prediction result and the backward prediction result output by the second module; For each of the first modules, integrating the forward prediction results output by each of the second modules to obtain the forward prediction result output by the first module; Integrating the forward prediction results output by each of the first modules to obtain the sea temperature prediction result output by the sea temperature prediction model; The sea temperature prediction model is trained according to the difference between the sea temperature prediction result output by the sea temperature prediction model and the label data corresponding to the sea temperature data indicated by the sample mapping relationship to obtain a trained sea temperature prediction model.

3. The method according to claim 2, characterized in that The step of determining the input of the second module according to the input of the previous second module of the second module and the backward prediction result of the previous second module output of the second module comprises: The difference between the input of the previous second module of the second module and the backward prediction result of the previous second module output of the second module is determined as the input of the second module.

4. The method according to claim 2 or 3, characterized in that: The second module includes a fully connected layer, a linear projection layer and a bias layer; the second module obtains a forward prediction result and a backward prediction result output by the second module based on the input of the second module, including: The input of the second module is passed through the fully connected layer for feature integration, and then two branches are used to perform corresponding linear transformation through the linear projection layer to generate forward prediction coefficients and backward prediction coefficients respectively; The forward prediction coefficient and the backward prediction coefficient are respectively mapped through the bias layer of the corresponding branch to obtain the forward prediction result and the backward prediction result output by the second module respectively.

5. The method according to claim 1, characterized in that The obtaining of mapping relationships between multiple groups of samples in the target area includes: Acquire a sea temperature data set within a target area and a target period; wherein the sea temperature data set includes a plurality of sea temperature data, each of which carries a longitude feature, a latitude feature, a time feature, and a depth feature; Determine at least one sea temperature data group from the sea temperature data set according to the depth characteristics of each of the sea temperature data in the sea temperature data set; wherein any of the sea temperature data groups includes a plurality of the sea temperature data; For any of the sea temperature data groups, determining feature set data and label data corresponding to at least one target sea temperature data in the sea temperature data group according to the time characteristics of each of the sea temperature data in the sea temperature data group; For any of the target sea temperature data in the sea temperature data group, a sample mapping relationship between the feature set data corresponding to the target sea temperature data and the label data corresponding to the target sea temperature data is determined.

6. The method according to claim 5, characterized in that After determining the sample mapping relationship between the feature set data corresponding to any of the target sea temperature data in the sea temperature data group and the label data corresponding to the target sea temperature data, the method further includes: According to the feature set data and label data corresponding to the target sea temperature data in the sea temperature data group, training the sample generator corresponding to the sea temperature data group, so that the trained sample generator corresponding to the sea temperature data group learns the corresponding relationship between the feature set data and the label data; According to each of the sea temperature data in the sea temperature data group and the sea temperature data generated by the sample generator corresponding to the trained sea temperature data group, a sample mapping relationship between at least one set of feature set data and label data is determined.

7. The method according to claim 1, characterized in that The sea temperature data carries a time feature; the determining of the first sea temperature data and the feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area includes: Select any sea temperature data from the real-time sea temperature data in the target area as the first sea temperature data; A set number of continuous sea temperature data whose time features are adjacent to the time features of the first sea temperature data and before the time features of the first sea temperature data and the first sea temperature data are determined as feature set data corresponding to the first sea temperature data.

8. The method according to claim 1, characterized in that The method of using the trained sea temperature prediction model to predict second sea temperature data whose time feature is after the time feature of the first sea temperature data according to the feature set data corresponding to the first sea temperature data includes: The feature set data corresponding to the first sea temperature data is input into the trained sea temperature prediction model to obtain the second sea temperature data output by the trained sea temperature prediction model.

9. A sea temperature prediction device, characterized in that: include: An acquisition module, used to acquire multiple groups of sample mapping relationships in a target area; wherein any group of the sample mapping relationships is used to indicate a corresponding relationship between feature set data and label data corresponding to any sea temperature data in the target area; A training module, used for training a sea temperature prediction model according to the feature set data and label data corresponding to each of the sea temperature data indicated by each group of the sample mapping relationships, so as to obtain a trained sea temperature prediction model; A determination module, used to determine first sea temperature data and feature set data corresponding to the first sea temperature data from the real-time sea temperature data in the target area; wherein each of the real-time sea temperature data carries a time feature; A prediction module is used to predict second sea temperature data whose time characteristics are after the time characteristics of the first sea temperature data using the trained sea temperature prediction model according to the feature set data corresponding to the first sea temperature data.

10. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.