Climate prediction correction method, system and equipment based on machine learning and medium

Abstract

The invention relates to the technical field of climate prediction, in particular to a climate prediction correction method, system and device based on machine learning and a medium, and the method comprises the steps: operating a CMA-CPSv3 climate prediction system, and obtaining an original climate mode prediction data sequence; inputting the original climate mode prediction data sequence into a pre-trained intelligent grid climate prediction model, and outputting a corrected rainfall prediction data sequence; obtaining an actually measured rainfall observation data sequence and an original climate mode prediction data sequence of a corresponding time period after rainfall in the target prediction area; and evaluating the prediction performance of the intelligent grid climate prediction model based on the actually measured rainfall observation data sequence and the corresponding original climate mode prediction data sequence, and updating the intelligent grid climate prediction model according to an evaluation result. According to the method, the gridding data are output through the intelligent grid climate prediction model, the parameters and the closed-loop process are dynamically updated based on the actually measured data, and high-precision, self-adaptive and efficient correction of regional rainfall prediction is realized.

Description

Background Art

[0002] Climate prediction is crucial for regional disaster prevention, mitigation, and agricultural production. The current climate prediction system (such as CMA-CPSv3) generates raw climate model prediction data through numerical model operations that couple physical processes such as the atmosphere and ocean. However, there are significant uncertainties in its predictions over long time windows, such as the sub-seasonal scale (15 - 60 days). This uncertainty stems from initial value errors, simplification of physical processes, and limited spatial resolution. In particular, regional climate is affected by local underlying surface characteristics (such as terrain and sea surface temperature) and forms specific patterns, which are difficult for the raw predictions to accurately capture, leading to relatively large biases in regional precipitation predictions. Therefore, it is necessary to correct the raw prediction results to improve regional applicability.

[0003] Existing technologies correct the raw prediction data of climate prediction systems through statistical or dynamical methods. For example, techniques such as non-parametric methods, Kalman filtering, or historical bias correction are used to adjust the model output; methods such as atmospheric low-frequency weather maps, circulation anomaly similarity methods, and spatio-temporal projection models (STPM) are also used to establish historical statistical relationships or physical mapping mechanisms to optimize regional prediction results.

[0004] However, the existing methods for correcting the raw prediction data of climate prediction systems have the following limitations: Insufficient spatial resolution: Traditional methods are difficult to efficiently process high-dimensional grid data, and the spatial fineness of the correction results is lower than that of the raw model grid, unable to fully represent regional climate patterns; Lack of dynamic adaptability: There is a lack of an evaluation mechanism based on real-time measured data and corresponding raw prediction sequences, and the model parameters cannot be dynamically optimized with the addition of new data; Process fragmentation and inefficiency: The correction process does not form a closed-loop system, consuming a large amount of computing resources and being difficult to continuously adapt to operational requirements. Summary of the Invention

[0005] Aiming at the technical problems of insufficient spatial resolution, lack of dynamic adaptability, and inefficiency caused by process fragmentation in the existing methods for correcting the raw prediction data of climate prediction systems, this application provides a climate prediction correction method, system, device, and medium based on machine learning, which realizes high-precision, adaptive, and efficient correction of regional precipitation prediction through an intelligent grid climate prediction model to output high-resolution grid data, dynamically update parameters based on measured data, and a closed-loop process.

[0006] In the first aspect, this application provides a climate prediction correction method based on machine learning, including the following steps: S1. Run the CMA-CPSv3 climate prediction system to obtain the raw climate model prediction data sequence of the target prediction area during the target prediction period; The original climate model prediction data sequence is the time-series data of the original climate model prediction data for the target prediction area within a preset time window; S2. Input the original climate model prediction data sequence into a pre-trained intelligent grid climate prediction model, and output the corrected precipitation prediction data sequence for the target prediction area during the target prediction period; The intelligent grid climate prediction model is constructed based on a machine learning algorithm, and its calculation is based on a regular longitude-latitude grid covering the geographical area where the target prediction area is located; The corrected precipitation prediction data sequence is the regular longitude-latitude grid precipitation prediction time-series data output by the intelligent grid climate prediction model and having the same spatial resolution as the calculation grid of the intelligent grid climate prediction model in the spatial dimension; S3. After precipitation occurs in the target prediction area, obtain the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence for the corresponding period; S4. Based on the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluate the prediction performance of the intelligent grid climate prediction model, and update the intelligent grid climate prediction model according to the evaluation results.

[0007] It should be further noted that in step S1, the original climate model prediction data includes: The height fields of different pressure levels, the zonal winds of different pressure levels, the meridional winds of different pressure levels, the sea-level pressure, the precipitation amount, and the air temperature at a height of 2 meters.

[0008] It should be further noted that the time window length of the target prediction period is 15 days to 60 days.

[0009] It should be further noted that in step S2, the corrected precipitation prediction data sequence includes: The time-series data of the grid precipitation probability prediction values and the time-series data of the grid precipitation amount prediction values for the target prediction area during the target prediction period.

[0010] It should be further noted that in step S2, the intelligent grid climate prediction model is constructed based on a long short-term memory network, a gated recurrent unit, a temporal convolutional network, or a Transformer model based on an attention mechanism.

[0011] It should be further noted that in step S2, the training steps of the intelligent grid climate prediction model include: Obtain the historical CMA-CPSv3 original climate model prediction data and the corresponding historical precipitation observation data for the target prediction area to form a historical data set; In the historical dataset, multiple groups of sample pairs are split according to a preset time window. Each group of sample pairs includes a CMA-CPSv3 original climate model prediction data sequence and its corresponding historical precipitation observation data sequence; Use the sample pairs to construct a training set, a validation set, and a test set; Initialize the parameters of the intelligent grid climate prediction model; Input the training set into the intelligent grid climate prediction model for iterative training; During the training process, use the validation set to monitor the model performance and implement an early stopping strategy to avoid overfitting; After the training is completed, use the test set to evaluate the generalization performance of the intelligent grid climate prediction model.

[0012] It should be further noted that the loss functions used in the iterative training include mean squared error, mean absolute error, or continuous ranked probability score; Among them, mean squared error and mean absolute error are applicable to the regression task of precipitation prediction values, and continuous ranked probability score is applicable to the probability prediction evaluation of precipitation probability prediction values.

[0013] It should be further noted that step S4 includes: S401. Input the CMA-CPSv3 original climate model prediction data sequence corresponding to the measured precipitation observation data sequence into the pre-trained intelligent grid climate prediction model to obtain the corresponding corrected precipitation prediction data sequence; S402. Calculate the evaluation indicators between the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; The evaluation indicators include correlation coefficient, root mean square error, and skill score; Among them, the correlation coefficient and root mean square error are used to evaluate the accuracy of the precipitation prediction values in the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; The skill score is used to evaluate the improvement degree of the corrected precipitation prediction data sequence relative to the CMA-CPSv3 original climate model prediction data sequence; S403. Set the thresholds of the evaluation indicators; when at least one evaluation indicator exceeds the corresponding preset threshold, the original climate model prediction data sequence and the measured precipitation observation data sequence corresponding to this evaluation are stored as new sample pairs in the incremental dataset; S404. When the number of new sample pairs in the incremental dataset exceeds the preset quantity threshold, use the incremental dataset to update the parameters of the intelligent grid climate prediction model through incremental learning, or retrain the intelligent grid climate prediction model using the extended dataset containing the incremental dataset.

[0014] Second aspect, the present application provides a climate prediction correction system based on machine learning for implementing the above climate prediction correction method, including: A data acquisition module, configured to run the CMA-CPSv3 climate prediction system to obtain the original climate model prediction data sequence of the target prediction area during the target prediction period, and after precipitation occurs in the target prediction area, obtain the measured precipitation observation data sequence and the CMA-CPSv3 original climate model prediction data sequence of the corresponding period; An intelligent grid prediction module, configured to input the original climate model prediction data sequence into a pre-trained intelligent grid climate prediction model, and output the corrected precipitation prediction data sequence of the target prediction area during the target prediction period; A dynamic evaluation and update module, based on the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluates the prediction performance of the intelligent grid climate prediction model, and updates the intelligent grid climate prediction model according to the evaluation result.

[0015] Third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor is configured to implement the steps of the above climate prediction correction method when executing the computer program.

[0016] Fourth aspect, the present application provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above climate prediction correction method are implemented.

[0017] From the above technical solutions, it can be seen that the present application has the following advantages: 1. The present application directly outputs the regular longitude and latitude grid precipitation prediction time series data with the same spatial dimension as the model calculation grid through the intelligent grid climate prediction model, solves the problem of insufficient spatial resolution of traditional methods, and realizes the high-precision spatial matching of precipitation prediction and regional underlying surface rules.

[0018] 2. The present application updates the model through the evaluation result based on the measured precipitation observation data sequence and the corresponding original prediction sequence, solves the problem of lack of dynamic adaptability in the prior art, realizes the real-time adjustment of model parameters with the addition of regional data, and strengthens the continuous learning ability of regional climate rules.

[0019] 3. The present application integrates the entire process closed-loop of "running the original prediction system - machine learning model correction - measured data evaluation - model update", solves the problems of process fragmentation and low calculation efficiency of traditional methods, realizes the efficient integrated correction of regional sub-seasonal precipitation prediction, and significantly improves the timeliness of business services. Description of the Drawings

[0020] To more clearly illustrate the technical solutions of this application, the accompanying drawings required for description will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0021] Figure 1 is a flowchart of a climate prediction correction method based on machine learning in an embodiment of this application.

[0022] Figure 2 is a schematic block diagram of a climate prediction correction system based on machine learning in an embodiment of this application.

[0023] Figure 3 is a schematic diagram of the hardware structure of an electronic device in an embodiment of this application. Detailed implementation manners

[0024] To make the application objectives, features, and advantages of this application more obvious and understandable, specific embodiments and accompanying drawings will be used below to clearly and completely describe the technical solutions protected by this application. Obviously, the embodiments described below are only some embodiments of this application, rather than all embodiments. Based on the embodiments in this patent, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this patent.

[0025] The climate prediction correction method related to this application will be described in detail below. For the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are proposed to thoroughly understand the embodiments of this application. However, those skilled in the art should understand that this application can also be implemented in other embodiments without these specific details.

[0026] In the climate prediction correction method related to this application, the term "including" indicates the existence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations. The terms "including", "comprising", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0027] To facilitate a clear description of the technical solutions of this application, terms such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and roles. Those skilled in the art can understand that the terms "first" and "second" do not limit the quantity and execution order, and the terms "first" and "second" do not necessarily limit to being different.

[0028] Statements such as "one embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in the embodiment are included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in some other embodiments", "in still some other embodiments", etc. that appear in different parts of this application do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.

[0029] The following are some noun explanations in this solution to facilitate a better understanding of this solution: CMA-CPSv3 Climate Prediction System: The CMA-CPSv3 Climate Prediction System is a sub-seasonal - seasonal - interannual scale integrated climate model prediction system developed by the China Meteorological Administration. It consists of a global high-resolution climate model subsystem, a multi-sphere coupled assimilation subsystem, and an ensemble prediction subsystem. It has the ability to predict temperature, precipitation, atmospheric circulation, and Madden-Julian Oscillation (MJO) at pentad and monthly scales, as well as the ability to predict key indicators such as temperature, precipitation, El Niño - Southern Oscillation (ENSO) index, Asian summer monsoon index, Western Pacific subtropical high index, and global sea ice area at seasonal scales. It can generate prediction graphic products covering spatial scales such as the global, Asian, and Chinese scales, including key elements of the atmosphere, land surface, ocean, and sea ice spheres and important climate indices.

[0030] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this application.

[0031] The climate prediction correction method provided in the embodiments of this application is executed by a computer device. Correspondingly, the climate prediction correction system for the dynamic balance of the ship's center of gravity runs in the computer device.

[0032] Figure 1 It is a flowchart of a machine learning-based climate prediction correction method according to an embodiment of this application. Among them, Figure 1 The execution subject can be a climate prediction correction system. According to different requirements, the order of the steps in this flowchart can be changed, and some can be omitted.

[0033] As Figure 1 shown, the machine learning-based climate prediction correction method includes: Step S1, run the CMA-CPSv3 climate prediction system to obtain the original climate model prediction data sequence of the target prediction area in the target prediction period; The original climate model prediction data sequence is the time-series data of the original climate model prediction data of the target prediction area within a preset time window.

[0034] Obtain the original climate model prediction data sequence of the target prediction area within the preset time window by running the CMA-CPSv3 climate prediction system, providing the multi-dimensional meteorological element basic data covering the target geographical area for subsequent machine learning correction, and ensuring the consistency of the data source in the correction process with the business prediction system.

[0035] In some specific embodiments, the original climate model prediction data includes: The height fields of different pressure levels, the zonal winds of different pressure levels, the meridional winds of different pressure levels, the sea-level pressure, the precipitation, and the air temperature at 2 meters height.

[0036] By defining that the original climate model prediction data includes the height fields of different pressure levels, zonal winds, meridional winds, sea-level pressure, precipitation, and air temperature at 2 meters height, multi-dimensional meteorological element inputs are provided for the intelligent grid climate prediction model, enhancing the model's ability to capture the characteristics of the atmospheric circulation and the underlying surface, and improving the physical rationality of precipitation prediction.

[0037] Step S2: Input the original climate model prediction data sequence into the pre-trained intelligent grid climate prediction model, and output the corrected precipitation prediction data sequence of the target prediction area during the target prediction period; The intelligent grid climate prediction model is constructed based on machine learning algorithms, and its calculation is based on the regular longitude and latitude grids covering the geographical area where the target prediction area is located; The corrected precipitation prediction data sequence is the regular longitude and latitude grid-based precipitation prediction time series data output by the intelligent grid climate prediction model, which has the same spatial resolution as the calculation grid of the intelligent grid climate prediction model in the spatial dimension.

[0038] By inputting the original climate model prediction data sequence into the intelligent grid climate prediction model constructed based on machine learning algorithms, directly outputting the regular longitude and latitude grid-based precipitation prediction time series data with the same resolution as the model calculation grid in the spatial dimension, the efficient and refined correction of the original prediction is realized, overcoming the limitations of traditional methods in grid data processing efficiency and resolution.

[0039] In some specific embodiments, the time window length of the target prediction period is 15 days to 60 days.

[0040] By setting the time window length of the target prediction period to be 15 days to 60 days, the sub-seasonal scale prediction range is clarified, enabling the model training and prediction period to match the actual business needs, and ensuring the applicability of the correction results within the sub-seasonal scale. In some specific embodiments, the corrected precipitation prediction data sequence includes: The time series data of the gridded precipitation probability prediction values and the time series data of the gridded precipitation amount prediction values of the target prediction area during the target prediction period.

[0041] By defining that the corrected sub-seasonal precipitation prediction data sequence includes the time-series data of grid-based precipitation probability prediction values and the time-series data of grid-based precipitation amount prediction values, probabilistic prediction and quantitative prediction are provided simultaneously, meeting the diverse needs of different application scenarios for precipitation prediction information.

[0042] In some specific embodiments, the intelligent grid climate prediction model is constructed based on a long short-term memory network, a gated recurrent unit, a temporal convolutional network, or a Transformer model based on an attention mechanism.

[0043] By constructing an intelligent grid climate prediction model using a long short-term memory network, a gated recurrent unit, a temporal convolutional network, or a Transformer model, and leveraging the powerful ability of deep learning algorithms to extract spatio-temporal sequence features, the non-linear relationship modeling of high-dimensional grid-based climate data is realized, improving the prediction accuracy.

[0044] In some specific embodiments, the training steps of the intelligent grid climate prediction model include: Obtain the historical CMA-CPSv3 raw climate model prediction data of the target prediction area and the corresponding historical precipitation observation data to form a historical data set; In the historical data set, split out multiple groups of sample pairs according to a preset time window, and each group of sample pairs includes a CMA-CPSv3 raw climate model prediction data sequence and its corresponding historical precipitation observation data sequence; Use the sample pairs to construct a training set, a validation set, and a test set; Initialize the parameters of the intelligent grid climate prediction model; Input the training set into the intelligent grid climate prediction model for iterative training; During the training process, use the validation set to monitor the model performance and implement an early stopping strategy to avoid overfitting; After the training is completed, use the test set to evaluate the generalization performance of the intelligent grid climate prediction model.

[0045] By constructing sample pairs based on the historical CMA-CPSv3 raw climate model prediction data and the corresponding precipitation observation data, dividing the training set, the validation set, and the test set for model training and early stopping strategy control, the model is ensured to have generalization ability and avoid the risk of overfitting.

[0046] In some specific embodiments, the loss functions used in the iterative training include mean squared error, mean absolute error, or continuous ranked probability score; Among them, mean squared error and mean absolute error are applicable to the regression task of precipitation amount prediction values, and continuous ranked probability score is applicable to the probability prediction evaluation of precipitation probability prediction values.

[0047] By using the mean square error or mean absolute error as the regression loss function for precipitation prediction values and the continuous ranked probability score as the evaluation index for precipitation probability prediction values, the model training objective is optimized for different prediction tasks, enhancing the statistical rationality and business usability of the prediction results.

[0048] Step S3, after precipitation occurs in the target prediction area, obtain the observed precipitation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence for the corresponding time period.

[0049] By obtaining the observed precipitation data sequence and the corresponding original climate model prediction data sequence after precipitation occurs in the target prediction area, a validation data set with strict spatio-temporal matching is provided for model performance evaluation, ensuring the objectivity and reliability of the evaluation results.

[0050] Step S4, based on the observed precipitation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluate the prediction performance of the intelligent grid climate prediction model, and update the intelligent grid climate prediction model according to the evaluation results.

[0051] By updating the intelligent grid climate prediction model according to the evaluation results based on the observed precipitation data sequence and the corresponding original prediction sequence, dynamic optimization and adaptive adjustment of the model parameters are achieved, ensuring continuous improvement of the prediction performance with the addition of new data.

[0052] In some specific embodiments, step S4 includes: S401. Input the CMA-CPSv3 original climate model prediction data sequence corresponding to the observed precipitation data sequence into the pre-trained intelligent grid climate prediction model to obtain the corresponding corrected precipitation prediction data sequence; S402. Calculate the evaluation index between the corrected precipitation prediction data sequence and the observed precipitation data sequence; The evaluation index includes the correlation coefficient, root mean square error, and skill score; Among them, the correlation coefficient and root mean square error are used to evaluate the accuracy of the precipitation prediction value in the corrected precipitation prediction data sequence and the observed precipitation data sequence; The skill score is used to evaluate the improvement degree of the corrected precipitation prediction data sequence relative to the CMA-CPSv3 original climate model prediction data sequence; S403. Set the threshold of the evaluation index; when at least one evaluation index exceeds the corresponding preset threshold, the original climate model prediction data sequence and the observed precipitation data sequence corresponding to this evaluation are used as new sample pairs and stored in the incremental data set; S404. When the number of newly added sample pairs in the incremental dataset exceeds a preset quantity threshold, update the parameters of the intelligent grid climate prediction model through incremental learning using the incremental dataset, or retrain the intelligent grid climate prediction model using an extended dataset that includes the incremental dataset.

[0053] By calculating the correlation coefficient, root mean square error, and skill score between the correction result and the measured data, set a threshold to trigger the storage of the incremental dataset, and use incremental learning or extended dataset retraining to update the model, realizing a closed-loop optimization mechanism based on real-time prediction performance feedback.

[0054] In a specific embodiment, the steps of the climate prediction correction method based on machine learning include: Step S1, run the CMA-CPSv3 climate prediction system to obtain the original climate model prediction data sequence of the target prediction area during the target prediction period; The original climate model prediction data sequence is the time-series data of the original climate model prediction data of the target prediction area within a preset time window; The target prediction area is Shandong Province; The target prediction period is the flood season (June - August), and the length of the time window is the sub-seasonal scale (the next 3 - 12 pentads, i.e., 15 - 60 days) The original climate model prediction data includes: 200hPa geopotential height field, 500hPa geopotential height field; 200hPa zonal wind, 200hPa meridional wind, 850hPa zonal wind, 850hPa meridional wind; Sea level pressure; Precipitation; Average temperature.

[0055] Step S2, input the original climate model prediction data sequence into the pre-trained intelligent grid climate prediction model, and output the corrected precipitation prediction data sequence of the target prediction area during the target prediction period; The intelligent grid climate prediction model is constructed based on the Temporal Convolutional Network (TCN), and its calculation is based on a regular longitude and latitude grid covering the geographical area where the target prediction area (Shandong Province) is located; The corrected precipitation prediction data sequence is the regular longitude and latitude grid precipitation prediction time-series data output by the intelligent grid climate prediction model and having the same spatial resolution as the calculation grid of the intelligent grid climate prediction model in the spatial dimension; The corrected precipitation prediction data sequence includes: The time-series data of the grid precipitation probability prediction value and the time-series data of the grid precipitation amount prediction value of the target prediction area (Shandong Province) during the target prediction period (the next 3 - 12 pentads); The training steps of the smart grid climate prediction model include: Obtain historical data sets: historical CMA-CPSv3 original climate model prediction data (including the variables described in step S1) for the target prediction area (Shandong Province) during the flood season (June-August) from 2008 to 2020 and corresponding daily precipitation observation data of 122 stations in Shandong Province during the same period; Construct sample pairs: In the historical data set, multiple groups of sample pairs are split according to the time window (daily data, the goal is to predict precipitation in the next 3-12 months). Each sample pair includes a CMA-CPSv3 original climate model prediction data sequence and its corresponding historical precipitation observation data sequence (precipitation at 122 stations in Shandong Province); Dataset division: Use sample pairs to construct training sets, validation sets, and test sets; Model construction and initialization: A time series convolutional network is used to establish the mapping relationship between the historical CMA-CPSv3 model prediction data and the historical observed precipitation at 122 stations in Shandong Province, forming a sub-seasonal smart grid climate prediction model and initializing the model parameters; The training set is input into the smart grid climate prediction model for iterative training. The loss functions used in iterative training include: Mean Square Error (MSE): Applicable to the regression task of precipitation prediction value; Continuous Graded Probability Score (CRPS): Applicable to the probabilistic forecast assessment of precipitation probability forecast values; During training, use the validation set to monitor model performance and implement early stopping strategies to avoid overfitting; After training, the test set is used to evaluate the generalization performance of the smart grid climate prediction model.

[0056] Step S3, after precipitation occurs in the target prediction area, obtain the measured precipitation observation data sequence and the CMA-CPSv3 original climate model prediction data sequence for the corresponding period; Among them, the measured precipitation observation data series is the daily precipitation data of 122 stations in Shandong Province; The CMA-CPSv3 original climate model prediction data sequence for the corresponding period is the original prediction data output by the CMA-CPSv3 system corresponding to the measured precipitation period; Step S4, based on the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluate the prediction performance of the smart grid climate prediction model, and update the smart grid climate prediction model according to the evaluation results, specifically including: S401. Inputting the CMA-CPSv3 original climate model prediction data sequence corresponding to the measured precipitation observation data sequence into the pre-trained smart grid climate prediction model to obtain the corresponding revised precipitation prediction data sequence; S402. Calculate the evaluation indicators between the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; The evaluation indicators include: Correlation coefficient: Evaluate the correlation between the precipitation prediction values in the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; Root Mean Square Error (RMSE): Evaluate the error between the precipitation prediction values in the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; Skill score: Evaluate the improvement degree of the corrected precipitation prediction data sequence relative to the CMA-CPSv3 original climate model prediction data sequence; S403. Set the thresholds of the evaluation indicators; when at least one evaluation indicator exceeds the corresponding preset threshold, use the corresponding original climate model prediction data sequence and the measured precipitation observation data sequence of this evaluation as a new sample pair and store it in the incremental data set; S404. When the number of new sample pairs in the incremental data set exceeds the preset quantity threshold, use the incremental data set to update the parameters of the intelligent grid climate prediction model through incremental learning, or retrain the intelligent grid climate prediction model using the extended data set containing the incremental data set.

[0057] The following is an embodiment of the climate prediction correction system based on machine learning provided by the embodiments of the present application. The climate prediction correction system based on machine learning and the climate prediction correction methods of the above embodiments belong to the same inventive concept. For the details not described in detail in the embodiments of the climate prediction correction system, reference may be made to the embodiments of the climate prediction correction method based on machine learning.

[0058] As Figure 2 shown, the climate prediction correction system based on machine learning includes: A data acquisition module, configured to run the CMA-CPSv3 climate prediction system, obtain the original climate model prediction data sequence of the target prediction area during the target prediction period, and after precipitation occurs in the target prediction area, obtain the measured precipitation observation data sequence and the CMA-CPSv3 original climate model prediction data sequence of the corresponding period; An intelligent grid prediction module, configured to input the original climate model prediction data sequence into a pre-trained intelligent grid climate prediction model, and output the corrected precipitation prediction data sequence of the target prediction area during the target prediction period; A dynamic evaluation and update module, based on the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluate the prediction performance of the intelligent grid climate prediction model, and update the intelligent grid climate prediction model according to the evaluation result.

[0059] The climate prediction correction system of this embodiment is used to implement a climate prediction correction method based on machine learning.

[0060] This application also provides an electronic device for implementing each embodiment of this application. Figure 3 As shown in the schematic hardware structure diagram of an electronic device for implementing each embodiment of this application, Figure 3 the electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor.

[0061] Those skilled in the art can understand that the structure of the electronic device involved in the embodiments of this application does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0062] In the embodiments of this application, the electronic device includes, but is not limited to, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device may also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described herein and / or claimed.

[0063] In the embodiments of this application, the processor may be implemented by using at least one of an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to execute the functions described herein. In some cases, such an implementation may be implemented in the controller. For a software implementation, an implementation of a process or function may be implemented with a separate software module that allows execution of at least one function or operation. The software code may be implemented by a software application (or program) written in any suitable programming language, and the software code may be stored in the memory and executed by the controller.

[0064] In addition, the electronic device includes some functional modules not shown herein, which will not be elaborated further.

[0065] Those skilled in the art to which the present application pertains can understand that various aspects of the electronic device provided by the present application can be implemented as a system, a method, or a program product. Therefore, various aspects of the present application can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which can be collectively referred to herein as "circuit", "module", or "system".

[0066] The present application also provides a storage medium in which a program product capable of implementing a climate prediction correction method for dynamic balance of the center of gravity of a ship is stored. In some possible implementation manners, various aspects of the present application can also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0067] The storage medium can be any combination of one or more readable media. The readable media can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0068] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A climate prediction correction method based on machine learning, characterized in that, Including: S1. Run the CMA-CPSv3 climate prediction system to obtain the original climate model prediction data sequence for the target prediction area during the target prediction period; The original climate model prediction data sequence is the time-series data of the original climate model prediction data for the target prediction area within a preset time window; S2. Input the original climate model prediction data sequence into the pre-trained intelligent grid climate prediction model to output the corrected precipitation prediction data sequence for the target prediction area during the target prediction period; The intelligent grid climate prediction model is constructed based on a machine learning algorithm, and its calculation is based on a regular longitude-latitude grid covering the geographical area where the target prediction area is located; The corrected precipitation prediction data sequence is the regular longitude-latitude grid precipitation prediction time-series data output by the intelligent grid climate prediction model and having the same spatial resolution as the calculation grid of the intelligent grid climate prediction model in the spatial dimension; S3. After precipitation occurs in the target prediction area, obtain the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence for the corresponding period; S4. Based on the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluate the prediction performance of the intelligent grid climate prediction model, and update the intelligent grid climate prediction model according to the evaluation results.

2. The climate prediction correction method according to claim 1, wherein, In step S1, the original climate model prediction data includes: The height fields of different pressure levels, the zonal winds of different pressure levels, the meridional winds of different pressure levels, the sea-level pressure, the precipitation, and the air temperature at 2 meters height.

3. The climate prediction correction method according to claim 1, wherein, In step S2, the corrected precipitation prediction data sequence includes: The time-series data of the gridded precipitation probability prediction values and the time-series data of the gridded precipitation amount prediction values for the target prediction area during the target prediction period.

4. The climate prediction correction method according to claim 1, wherein, In step S2, the intelligent grid climate prediction model is constructed based on a long short-term memory network, a gated recurrent unit, a temporal convolutional network, or a Transformer model based on an attention mechanism.

5. The climate prediction correction method according to claim 1, characterized in that In step S2, the training steps of the intelligent grid climate prediction model include: Obtain the historical CMA-CPSv3 original climate model prediction data and the corresponding historical precipitation observation data for the target prediction area to form a historical data set; In the historical data set, split out multiple groups of sample pairs according to a preset time window, and each group of sample pairs includes a CMA-CPSv3 original climate model prediction data sequence and its corresponding historical precipitation observation data sequence; Use the sample pairs to construct a training set, a validation set, and a test set; Initialize the parameters of the intelligent grid climate prediction model; Input the training set into the intelligent grid climate prediction model for iterative training; During the training process, use the validation set to monitor the model performance and implement an early stopping strategy to avoid overfitting; After the training is completed, use the test set to evaluate the generalization performance of the intelligent grid climate prediction model.

6. The climate prediction correction method according to claim 5, wherein, The loss functions used in the iterative training include the mean squared error, the mean absolute error, or the continuous ranked probability score; Among them, the mean squared error and the mean absolute error are applicable to the regression task of the precipitation amount prediction value, and the continuous ranked probability score is applicable to the probability prediction evaluation of the precipitation probability prediction value.

7. The climate prediction correction method according to claim 1, characterized in that, Step S4 includes: S401. Input the CMA-CPSv3 original climate model prediction data sequence corresponding to the measured precipitation observation data sequence into the pre-trained intelligent grid climate prediction model to obtain the corresponding corrected precipitation prediction data sequence; S402. Calculate the evaluation indicators between the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; The evaluation indicators include the correlation coefficient, root mean square error, and skill score; Among them, the correlation coefficient and root mean square error are used to evaluate the accuracy of the precipitation prediction value in the corrected precipitation prediction data sequence and the measured precipitation observation data sequence; The skill score is used to evaluate the improvement degree of the corrected precipitation prediction data sequence relative to the CMA-CPSv3 original climate model prediction data sequence; S403. Set the threshold of the evaluation indicator; when at least one evaluation indicator exceeds the corresponding preset threshold, the original climate model prediction data sequence and the measured precipitation observation data sequence corresponding to this evaluation are used as a new sample pair and stored in the incremental data set; S404. When the number of new sample pairs in the incremental data set exceeds the preset quantity threshold, use the incremental data set to update the parameters of the intelligent grid climate prediction model through incremental learning, or retrain the intelligent grid climate prediction model using the extended data set containing the incremental data set.

8. A climate prediction correction system based on machine learning, characterized in that, For implementing the climate prediction correction method according to any one of claims 1-7, including: A data acquisition module, configured to run the CMA-CPSv3 climate prediction system to obtain the original climate model prediction data sequence of the target prediction area during the target prediction period, and after precipitation occurs in the target prediction area, obtain the measured precipitation observation data sequence and the CMA-CPSv3 original climate model prediction data sequence of the corresponding period; An intelligent grid prediction module, configured to input the original climate model prediction data sequence into the pre-trained intelligent grid climate prediction model and output the corrected precipitation prediction data sequence of the target prediction area during the target prediction period; A dynamic evaluation and update module, based on the measured precipitation observation data sequence and the corresponding CMA-CPSv3 original climate model prediction data sequence, evaluates the prediction performance of the intelligent grid climate prediction model, and updates the intelligent grid climate prediction model according to the evaluation results.

9. An electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor is configured to implement the steps of the climate prediction correction method according to any one of claims 1-7 when executing the computer program.

10. A storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the climate prediction correction method according to any one of claims 1-7 are implemented.

Images