Climate prediction methods and equipment based on observation constraints and multi-time sliding windows

Abstract

This invention discloses a climate prediction method and apparatus based on observation constraints and a multi-time-dependent sliding window. The method includes: acquiring the first climate data for the current and historical years predicted by a specified climate system model; for each month in the historical year: using the first climate data of that month and a predetermined number of previous months as input features, and the actual climate data of that month as labels, generating sample pairs for that month and adding them to a training sample library; training a pre-constructed climate prediction model using the training sample library with the objective of minimizing the joint loss, the joint loss including the root mean square error, spatial anomaly correlation coefficient, and prediction score characterizing the model's prediction performance; for any month to be predicted in the current year: inputting the first climate data of that month and a predetermined number of previous months into the trained climate prediction model, and outputting the climate prediction result for that month. This invention improves the reliability and fine spatiotemporal resolution of climate prediction.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence and climate prediction, and more specifically, relates to a climate prediction method and device based on observation constraints and multiple time-dependent sliding windows. Background Technology

[0002] Climate prediction is a crucial decision-making basis for disaster prevention and mitigation, water resource management, and agricultural production planning. However, current climate prediction still faces significant challenges, with its accuracy and reliability needing substantial improvement. Mainstream dynamic climate prediction systems, such as the Beijing Climate Center Climate System Model (BCC-CSM) and Climate Prediction System version 2 (CFSv2), are core tools for seasonal climate prediction. However, these complex numerical models generally exhibit significant systematic biases due to insufficient understanding of parameterization schemes such as cloud physics and land surface processes, as well as initial field errors. Furthermore, the dispersion among members in ensemble forecasts with multiple initial values ​​or multiple physical processes is often too large, resulting in a low signal-to-noise ratio and reduced certainty and reliability of ensemble average results, thus limiting their direct application value in refined services.

[0003] To correct systematic errors in numerical models (i.e., dynamic climate prediction systems), model output statistics (MOS) or regression methods are often employed. While these methods improve prediction bias to some extent, they essentially establish linear or simple nonlinear statistical relationships between forecast factors and observations. Faced with the complex nonlinear processes of ocean-land-atmosphere interactions in the climate system and the coupling effects across multiple temporal and spatial scales, traditional statistical models struggle to fully extract and utilize the effective predictive information contained in model products, limiting the potential for improvement in prediction performance.

[0004] In recent years, deep learning technology has provided new insights for climate prediction due to its powerful nonlinear fitting capabilities. However, existing deep learning-based climate prediction methods suffer from the following major problems: Most existing studies use fixed historical datasets for model training, which gradually become outdated over time, leading to a decline in predictive performance; during training, the model typically optimizes the loss function grid-by-grid, lacking explicit constraints on large-scale spatial correlations and co-evolutionary patterns, making the training process prone to spatial structure bias; optimizing only the mean square error (MSE) or root mean square error (RMSE) without physical structure constraints only ensures that the predicted field is numerically close to the actual field, but cannot guarantee the physical rationality of the predicted variables; most methods are designed to directly predict the average total amount or average state of the entire season, lacking the ability to predict the monthly evolution process within the season. This coarse-grained prediction cannot meet the needs for refined services such as key agricultural periods and flood seasons within the season.

[0005] Therefore, there is an urgent need for a new forecasting framework that can integrate numerical model information, dynamically adapt to climate change, embed physical constraints, and support refined timescale output, in order to break through the current technical bottlenecks in seasonal climate forecasting. Summary of the Invention

[0006] The main objective of this invention is to provide a climate prediction method and device based on observation constraints and multiple time-dependent sliding windows, so as to overcome the shortcomings of the prior art.

[0007] The first aspect of this invention provides a climate prediction method based on observation constraints and multiple time-dependent sliding windows, comprising: acquiring the first climate data of the current and historical years predicted by a specified climate system model; for each month in the historical year: generating sample pairs for that month using the first climate data of that month and a predetermined number of previous months as input features and the actual climate data of that month as labels; combining the sample pairs of each month to form a training sample library; training a pre-constructed climate prediction model using the training sample library with the objective of minimizing the joint loss, wherein the joint loss includes the root mean square error, spatial anomaly correlation coefficient, and prediction score characterizing the model's prediction performance; for any month to be predicted in the current year: inputting the first climate data of the month to be predicted and a predetermined number of previous months into the trained climate prediction model, and outputting the climate prediction result for the month to be predicted.

[0008] Preferably, the joint loss is:

[0009] ;

[0010] in, For the joint loss; This is the first hyperparameter; This is the second hyperparameter; and The root mean square error, the spatial anomaly correlation coefficient, and the predicted score are all of the same order of magnitude. The root mean square error (RMSE) represents the error between the predicted value and the actual value. The spatial anomaly correlation coefficient characterizes the spatial consistency between the predicted and actual values. The prediction score characterizes the degree of closeness between the predicted value and the actual value.

[0011] Preferably, the predicted score is:

[0012] ;

[0013] in, For the predicted score, The total number of observation stations, This represents the number of observation stations where the predicted percentage anomaly and the actual percentage anomaly have the same sign. The absolute values ​​of both the predicted and actual anomaly percentages are within [a certain range]. The number of observation stations within the interval, The absolute values ​​of both the predicted and actual anomaly percentages are within [a certain range]. The number of observation stations within the interval, For the absolute value of the predicted value's percentage anomaly to be no less than 1 and the absolute value of the actual value's percentage anomaly to be within 1, the following conditions must be met: The number of observation stations within the interval, The first preset parameter, This is the second preset parameter. , where a is the first coefficient, b is the second coefficient, and c is the third coefficient.

[0014] Preferably, the spatial anomaly correlation coefficient is:

[0015] ;

[0016] ;

[0017] in, The spatial anomaly correlation coefficient is... The weights corresponding to observation station i, Let i be the latitude corresponding to observation station i. For the actual climate data of observation station i, This is the average of the actual climate data from all observation stations. The model prediction results for observation station i. This is the average of the model predictions for all observation sites.

[0018] Preferably, obtaining the first climate data for the current and historical years predicted by the specified climate system model specifically includes: obtaining the data of each ensemble member for the current and historical years predicted by the specified climate system model; for each year: performing ensemble averaging on the data of each ensemble member for that year to obtain the climate factor field for that year, interpolating and resampling the climate factor field to a preset standard spatial grid to generate the first climate data for that year.

[0019] Preferably, the climate prediction model includes: an input layer for preprocessing the data input to the climate prediction model into a specified format; a multi-scale convolution module for capturing spatial features of the preprocessed data at different scales to obtain feature maps at different scales; a residual module including multiple cascaded residual blocks, each residual block including a convolutional layer and a batch normalization layer to process the feature maps; an adaptive pooling module for pooling the feature maps output by the residual module to output feature maps of a set size; and a fully connected module for expanding the feature maps output by the adaptive pooling module and mapping them to the final output space to output the climate prediction results for each observation station.

[0020] Preferably, a self-attention mechanism module is further connected between the residual module and the adaptive pooling module; the self-attention mechanism module is used to capture long-range dependencies in the feature map sequence output by the residual module to adjust the influence of different regions on the prediction result; the adaptive pooling module is used to perform pooling processing on the feature map output by the self-attention mechanism module to output a feature map of a set size.

[0021] Preferably, the method further includes: the trained climate prediction model outputs the climate prediction results for each month of the current year, and calculates the quarterly climate prediction results based on the climate prediction results for the corresponding months.

[0022] A second aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the climate prediction method based on observation constraints and multiple time-dependent sliding windows as described above.

[0023] A third aspect of the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the climate prediction method based on observation constraints and multiple time-dependent sliding windows as described above.

[0024] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a climate prediction method and device based on observation constraints and multi-time-dependent sliding windows. The training sample library constructed is a dynamic sliding training set, ensuring that the data of the year to be predicted does not participate in any learning or optimization process, completely eliminating the risk of future information leakage and improving the objectivity and credibility of the prediction model. The invention introduces spatial anomaly correlation coefficients and prediction scores into the joint loss. This constraint mechanism directly guides the model to learn and maintain a spatial covariance structure consistent with the actual observation field during training, thereby effectively improving the spatial consistency and physical rationality of the prediction results, making the spatial distribution of the model output closer to actual atmospheric processes. The model can output climate prediction results for each month, achieving a breakthrough in refined prediction from seasonal to monthly scales, providing richer and more decision-guiding prediction information. It possesses high automation and scalability, supporting automatic updates of the training set and model iteration based on newly added observation data each year, realizing continuous learning and performance evolution of the prediction model, ensuring that the model can adapt to climate change, and guaranteeing the long-term stable and efficient operation of the operational forecasting system. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart of a climate prediction method based on observation constraints and multiple time-dependent sliding windows provided in an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of the sliding training set construction method in this invention.

[0028] Figure 3 This is a schematic diagram illustrating the working mechanism and structure of the sliding input window used in this invention.

[0029] Figure 4 This is a schematic diagram of the structure of the climate prediction model provided in an embodiment of the present invention.

[0030] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0031] In view of the shortcomings of the prior art, the inventors of this invention, through long-term research and extensive practice, have proposed the technical solution of this invention. The following will further explain and illustrate this technical solution, its implementation process, and its principles.

[0032] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0033] Furthermore, in the description of this invention, it should be understood that the terms "upper," "lower," "inner," "outer," "horizontal," "vertical," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0034] In the description of this specification, the references to terms such as "an embodiment," "a particular embodiment," or "the embodiment" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0035] Figure 1 A flowchart illustrating the climate prediction method based on observation constraints and multi-time-dependent sliding windows provided in this embodiment of the invention. (See also...) Figure 1 The climate prediction method based on observation constraints and multi-time-dependent sliding windows provided by the present invention includes the following operations S100-S400.

[0036] Operate S100 to obtain the first climate data for the current and historical years predicted by the specified climate system model.

[0037] Operation S200: For each month in a historical year: use the first climate data of the current month and a set number of previous months as input features, and the actual climate data of the current month as labels to generate sample pairs for that month; combine the sample pairs of each month to form a training sample library.

[0038] The S300 operation aims to minimize the joint loss by training a pre-built climate prediction model using a training sample library. The joint loss includes the root mean square error, spatial anomaly correlation coefficient, and prediction score, which characterize the model's prediction performance.

[0039] Operation S400: For any month to be predicted in the current year: Input the first climate data of the month to be predicted and the previous set number of months into the trained climate prediction model, and output the climate prediction result of the month to be predicted.

[0040] Preferably, operation S100 specifically includes the following sub-operations S110-S120.

[0041] In suboperation S110, data of each set member predicted by the specified climate system model for the current and historical years are obtained.

[0042] Specifically, based on a preset file path mapping, it can automatically read data from each ensemble member of a climate system model (e.g., BCC-CSM) within a specified year range. The specified year could be, for example, the current year and the 10 years preceding it.

[0043] In sub-operation S120, for each year: the data of each set member under that year are ensembled to obtain the climate factor field of that year, and the climate factor field is interpolated and resampled to a preset standard spatial grid to generate the first climate data of that year.

[0044] Specifically, the data of each set member is parsed, and the ensemble average of all set member data for the same year is performed to obtain the climate factor field representing the climate state of that year. This climate factor field is then interpolated and resampled to a preset standard spatial grid (e.g., 1°×1°) to eliminate the differences in spatial resolution of the original data and ensure that all input data are spatially aligned.

[0045] Before performing operation S200, the method also includes reading and filtering observational data (actual climate data). Specifically, it reads the station observational data files, parses the time information and climate data of each station, and filters out the actual climate data of historical years (e.g., the 10 years before the current year), which will then be used as labels for model training.

[0046] The process of constructing the sliding training set is as follows: Figure 2 As shown. Taking the current year as 2026, the historical years as 2016-2025, and a set number of samples of 3 as an example, the sample pairs generated in operation S200 are explained. For each month in the 2016-2025 period, such as June 2024, the first climate data of June, May, April, and March 2024 are used as input features, and the actual climate data of June 2024 are used as labels to generate "input feature-label" sample pairs for June 2024; and so on, to construct a training sample library covering sample pairs from January 2016 to December 2025. The working mechanism and structure of the sliding input window are as follows. Figure 3 As shown.

[0047] Furthermore, the feature tensors, observation labels, and corresponding metadata (such as year) of all generated sample pairs can be saved as files in a standardized format, completing the automated pipeline processing from raw data to a model-ready dataset. In this embodiment, the sample pairs in the training sample library can be randomly divided into training and validation sets, and the sample pairs of the current year can be used as the test set for the optimization training of the climate prediction model.

[0048] In this embodiment of the invention, a climate prediction model is pre-built before performing operation S300. This climate prediction model can be a newly built climate prediction model for each year, or it can be a climate prediction model optimized and trained in the previous year.

[0049] Preferably, the climate prediction model includes: an input layer for preprocessing the data input to the climate prediction model into a specified format; a multi-scale convolutional module for capturing spatial features of the preprocessed data at different scales to obtain feature maps at different scales; a residual module including multiple cascaded residual blocks, each residual block including convolutional layers and batch normalization layers to process the feature maps; an adaptive pooling module for pooling the feature maps output by the residual module to output feature maps of a set size; and a fully connected module for expanding the feature maps output by the adaptive pooling module and mapping them to the final output space to output the climate prediction results for each observation station.

[0050] Further preferably, a self-attention mechanism module is connected between the residual module and the adaptive pooling module. The self-attention mechanism module is used to capture long-range dependencies in the feature map sequence output by the residual module, thereby adjusting the influence of different regions on the prediction results; the adaptive pooling module is used to pool the feature maps output by the self-attention mechanism module to output feature maps of a set size. The structure of the climate prediction model is as follows: Figure 4 As shown, a multi-scale residual network architecture, combined with self-attention mechanism and adaptive pooling, is used to improve the model's ability to model spatial and temporal features when processing climate data.

[0051] by Figure 4 For example, the specific architecture of a climate prediction model is illustrated.

[0052] In the input layer, the input data is preprocessed meteorological data, in the form of... Where N is the number of samples, C is the number of channels (e.g., different advance months), H is the latitude, and W is the longitude.

[0053] The multi-scale convolution module uses convolution operations with kernel sizes of 3, 5, and 7 to capture spatial features at different scales. Multiple kernels can extract spatial patterns within different ranges, enhancing the model's ability to perceive different spatial scales. The multi-scale convolution module helps the model extract information from features at different scales, making it particularly suitable for complex spatial structures in climate prediction. The input is processed by convolution kernels of sizes 3x3, 5x5, and 7x7 to obtain feature maps at different scales. The feature maps at each scale are merged through channel concatenation and used as input for subsequent networks.

[0054] The residual module contains multiple residual blocks, each consisting of two convolutional layers and a batch normalization layer. Input data, after passing through convolution, activation functions, and pooling, is directly added to the output via residual connections. Residual blocks can alleviate the vanishing gradient problem, improve the model's convergence speed and performance, and enable effective training even with deeper networks.

[0055] The self-attention mechanism module enables the model to dynamically capture long-range dependencies in the input sequence. In climate prediction, the model needs to handle the long-term spatial dependencies of different regions. The self-attention mechanism can adaptively adjust the impact of different regions on the prediction, for example, allowing the model to focus its attention on the most relevant parts (such as areas prone to heavy rainfall).

[0056] The adaptive pooling module performs pooling on the input feature map, making its size adaptable to the input requirements of the fully connected layer. Unlike traditional pooling methods, adaptive pooling automatically adjusts the size of the pooling region based on the size of the input features, outputting a feature map of a fixed size.

[0057] The model uses fully connected layers to unfold the pooled feature maps and map them to the final output space. The output dimension corresponds to the number of observation stations, representing the predicted value for each observation station. For example, the output is the precipitation prediction for 160 observation stations.

[0058] It should be noted that, Figure 4 The climate prediction model shown is a preferred example, and other structures can also be used in the climate prediction model in the embodiments of the present invention.

[0059] In this embodiment of the invention, the joint loss of the climate prediction model includes the root mean square error, spatial anomaly correlation coefficient, and prediction score, which characterize the model's prediction performance. This ensures that during the training process of the precipitation prediction model, not only is numerical accuracy optimized (through the root mean square error), but also the spatial consistency of the prediction results is maintained. By introducing the spatial anomaly correlation coefficient constraint, the model can better capture and maintain the spatial structure and patterns between the prediction results and the observed data, thereby improving the physical rationality and spatial coherence of the prediction. By introducing the prediction score, the ability to predict sudden extreme events is strengthened, making the overall model score robust and improving the anomaly sign consistency rate between predictions and observations.

[0060] Preferably, the joint loss is:

[0061] ;

[0062] in, For joint losses; This is the first hyperparameter; This is the second hyperparameter; and The root mean square error, spatial anomaly correlation coefficient, and prediction score are of the same order of magnitude; Root mean square error (RMSE) represents the error between the predicted and actual values. The spatial anomaly correlation coefficient characterizes the spatial consistency between predicted and actual values. A prediction score characterizes the degree of closeness between the predicted and actual values. Choosing a reasonable score is crucial. and The model achieves a good balance between root mean square error, spatial anomaly correlation coefficient, and prediction score.

[0063] The root mean square error measures the error between the predicted value and the actual value (true observation), and its goal is to minimize the numerical difference between the predicted value and the true observation.

[0064] Preferably, the spatial anomaly correlation coefficient for:

[0065] ;

[0066] ;

[0067] in, The weights corresponding to observation station i, Let i be the latitude corresponding to observation station i. For the actual climate data of observation station i, This is the average of the actual climate data from all observation stations. The model prediction results for observation station i. This is the average of the model predictions for all observation sites.

[0068] The spatial anomaly correlation coefficient measures the spatial consistency between predicted and actual values, with a range of [value range missing]. The closer the value is to 1, the closer the predicted spatial anomaly is to the actual data.

[0069] The prediction score is determined by classifying the percentage error between the predicted results and the actual observations at each observation station, and combining this with the number of stations at each error level, reflecting how close the predicted results are to the actual observations. The PS value ranges from 0 to 1, for example, with a value closer to 1 indicating better model prediction performance.

[0070] Preferred score for:

[0071] ;

[0072] in, The total number of observation stations, This represents the number of observation stations where the predicted percentage anomaly and the actual percentage anomaly have the same sign. The absolute values ​​of both the predicted and actual anomaly percentages are within [a certain range]. The number of observation stations within the interval, The absolute values ​​of both the predicted and actual anomaly percentages are within [a certain range]. The number of observation stations within the interval, For the absolute value of the predicted value's percentage anomaly to be no less than 1 and the absolute value of the actual value's percentage anomaly to be within 1, the following conditions must be met: The number of observation stations within the interval, The first preset parameter, This is the second preset parameter. , where 'a' is the first coefficient, 'b' is the second coefficient, and 'c' is the third coefficient. For example... , a=2, b=2, c=4. The predicted percentage anomaly and the actual percentage anomaly may be positive, negative, or 0.

[0073] Preferably, the method further includes: the trained climate prediction model outputs the climate prediction results for each month of the current year, and calculates the quarterly climate prediction results based on the climate prediction results for the corresponding months.

[0074] Taking the current year as 2026 as an example, the trained climate prediction model directly outputs monthly climate prediction results. , , , , , , , , , , , ,in, This represents the climate prediction result for the i-th site in the m-th month of 2026. For example, As part of the first quarter's climate forecasts, This is a result of summer climate forecasts.

[0075] The prediction results of the method of this invention were compared with those of existing climate models (BCC-CSM in this embodiment), and summer climate data for the years 2021-2025 were predicted respectively. The prediction results show that the average ACC of the existing method for the years 2021-2025 is 0.13, while the average ACC of the method of this invention is 0.26, representing a doubling of the original ACC; the average PS score of the existing method for the years 2021-2025 is 72.02, while the average PS score of the method of this invention is 78.08, representing an improvement of approximately 8.4% compared to the existing method.

[0076] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the climate prediction method based on observation constraints and multiple time-dependent sliding windows as described in any of the above embodiments.

[0077] Figure 5 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 510, a memory 520, an input / output interface 530, a communication interface 540, and a bus 550. The processor 510, memory 520, input / output interface 530, and communication interface 540 are interconnected internally via the bus 550.

[0078] The processor 510 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0079] The memory 520 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 520 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 520 and is called and executed by the processor 510.

[0080] Input / output interface 530 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0081] The communication interface 540 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0082] Bus 550 includes a pathway for transmitting information between various components of the device, such as processor 510, memory 520, input / output interface 530, and communication interface 540.

[0083] It should be noted that although the above-described device only shows the processor 510, memory 520, input / output interface 530, communication interface 540, and bus 550, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0084] The electronic devices described above are used to implement the corresponding climate prediction methods based on observation constraints and multi-time-dependent sliding windows in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0085] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the climate prediction method based on observation constraints and multiple time-dependent sliding windows as described in any of the above embodiments.

[0086] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0087] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the climate prediction method based on observation constraints and multiple time-dependent sliding windows as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0088] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of the invention as described above, which are not provided in the details for the sake of brevity.

[0089] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of the invention, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of the invention, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of the invention will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of the invention, it will be apparent to those skilled in the art that the embodiments of the invention may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0090] Although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0091] It should be understood that the above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A climate prediction method based on observation constraints and multi-time-scale sliding windows, characterized in that, include: Obtain the first climate data for the current and historical years predicted by a specified climate system model; For each month in a historical year: using the first climate data of that month and a predetermined number of previous months as input features and the actual climate data of that month as labels, generate sample pairs for that month; combine the sample pairs from each month to form a training sample library; With the goal of minimizing the joint loss, a pre-built climate prediction model is trained using the aforementioned training sample library. The joint loss includes the root mean square error, spatial anomaly correlation coefficient, and prediction score, all characterizing the model's prediction performance. The spatial anomaly correlation coefficient is: ； ； wherein, is the spatial distance correlation coefficient, is the weight corresponding to observation site i, is the latitude corresponding to observation site i, is the actual climate data of observation site i, is the average of the actual climate data of all observation sites, is the model prediction result of observation site i, is the average of the model prediction results of all observation sites; For any month to be predicted in the current year: input the first climate data of the month to be predicted and the previous set number of months into the trained climate prediction model, and output the climate prediction result for the month to be predicted.

2. The climate prediction method based on observation constraints and multi-time-scale sliding windows of claim 1, wherein, The joint loss is: ； wherein, is the joint loss; is a first hyper-parameter; is a second hyper-parameter; and a magnitude of the root mean square error, the spatial distance correlation coefficient, and the prediction score are controlled to be the same; is the root mean square error, representing an error between a predicted value and an actual value; is the spatial distance correlation coefficient, representing a consistency in space between a predicted value and an actual value; is the prediction score, representing a closeness between a predicted value and an actual value.

3. The climate prediction method based on observation constraints and multi-time-dependent sliding windows according to claim 1 or 2, characterized in that, The predicted score is: ； in, For the predicted score, The total number of observation stations, This represents the number of observation stations where the predicted percentage anomaly and the actual percentage anomaly have the same sign. The absolute values ​​of both the predicted and actual anomaly percentages are within [a certain range]. The number of observation stations within the interval, The absolute values ​​of both the predicted and actual anomaly percentages are within [a certain range]. The number of observation stations within the interval, For the absolute value of the predicted value's percentage anomaly to be no less than 1 and the absolute value of the actual value's percentage anomaly to be within 1, the following conditions must be met: The number of observation stations within the interval, The first preset parameter, This is the second preset parameter. , where a is the first coefficient, b is the second coefficient, and c is the third coefficient.

4. The climate prediction method based on observation constraints and multi-time-dependent sliding windows according to claim 1, characterized in that, Obtain first-year climate data predicted by a specified climate system model for the current and historical years, specifically including: Obtain the data of each set member predicted by the specified climate system model for the current and historical years; For each year: the data of each set member in that year are averaged to obtain the climate factor field for that year. The climate factor field is then interpolated and resampled to a preset standard spatial grid to generate the first climate data for that year.

5. The climate prediction method based on observation constraints and multi-time-dependent sliding windows according to claim 1, characterized in that, The climate prediction model includes: The input layer is used to preprocess the data input to the climate prediction model into a specified format; The multi-scale convolution module is used to capture the spatial features of the preprocessed data at different scales and obtain feature maps at different scales. The residual module includes multiple cascaded residual blocks, each residual block comprising a convolutional layer and a batch normalization layer, for processing the feature map; An adaptive pooling module is used to perform pooling processing on the feature map output by the residual module, and output a feature map of a set size. The fully connected module is used to expand the feature map output by the adaptive pooling module and map it to the final output space to output the climate prediction results of each observation station.

6. The climate prediction method based on observation constraints and multi-time-dependent sliding windows according to claim 5, characterized in that, A self-attention mechanism module is also connected between the residual module and the adaptive pooling module; The self-attention mechanism module is used to capture long-range dependencies in the feature map sequence output by the residual module, so as to adjust the influence of different regions on the prediction results; The adaptive pooling module is used to perform pooling processing on the feature map output by the self-attention mechanism module to output a feature map of a set size.

7. The climate prediction method based on observation constraints and multi-time-dependent sliding windows according to claim 1, characterized in that, The method further includes: the trained climate prediction model outputs the climate prediction results for each month of the current year, and calculates the quarterly climate prediction results based on the climate prediction results for the corresponding months.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the climate prediction method based on observation constraints and multi-time-dependent sliding windows as described in any one of claims 1-7.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the climate prediction method based on observation constraints and multiple time-dependent sliding windows as described in any one of claims 1-7.

Images