Optimization model construction method for improving resolution and forecasting precision of artificial intelligence meteorological large model

Abstract

The invention discloses an optimization model construction method and device for improving the resolution and forecast precision of an artificial intelligence meteorological large model, and the method comprises the steps: constructing a coarse mesh resolution mode initial field through employing a high-resolution training data set, driving the coarse grid resolution target artificial intelligence meteorological large model to carry out continuous forecasting in a preset time period so as to respectively obtain coarse grid resolution to-be-optimized data of a plurality of forecasting steps; inputting the to-be-optimized data corresponding to each forecasting step into a pre-constructed artificial intelligence data processing model, and obtaining an error value corresponding to each forecasting step in combination with the training data; and summing the error values obtained in each forecasting step, and training the artificial intelligence data processing model based on the sum of the error values by taking a minimization objective function as a training objective to obtain an optimized artificial intelligence data processing model. According to the method and the device, the spatial resolution of the meteorological large model can be improved, so that the precision of meteorological forecast is improved.

Description

Technical Field

[0001] The present invention relates to the field of meteorology and artificial intelligence technology, and in particular to an optimization model construction method and device for improving the resolution and forecast accuracy of a large artificial intelligence meteorological model. Background Art

[0002] In recent years, the application of artificial intelligence in the field of weather forecasting has made important technological breakthroughs, and artificial intelligence weather forecasting large models (AIWFMs) have become an emerging direction in the development of atmospheric science. Compared with traditional numerical weather forecasting models, AIWFMs have the outstanding advantages of small resource requirements and fast computing speed, which is convenient for widespread and convenient application and ensemble forecasting, complementing traditional deterministic forecasting.

[0003] Although AIWFMs have obvious advantages, the following problems still need to be solved in actual business applications: the vast majority of AIWFMs are trained using the ERA5 atmospheric reanalysis products of the European Centre for Medium- and Short-Term Forecasts, resulting in obvious homogeneity in the performance of AIWFMs models, especially the spatial resolution is severely limited by the original resolution of the ERA5 atmospheric reanalysis data. Given that the current forecast performance of AIWFMs is comparable to that of traditional numerical weather forecasts, further improving the accuracy of specific forecast elements by designing and training new models requires a lot of additional costs and has limited expected benefits; in addition, the existing AIWFMs are all global large models, and do not pay enough attention to regional forecasts. The sporadic artificial intelligence regional weather forecast large models (AIRWFMs) are also similar to traditional regional numerical weather models, requiring global numerical weather forecast models or AIGWFMs to provide boundary conditions, and the forecast accuracy and computational efficiency are still limited by the forecast accuracy and computational efficiency of the global model, that is, the artificial intelligence large models for global and regional forecasts lack an organic connection.

[0004] In order to improve the spatial resolution of the AI ​​weather forecast model, there are two main existing methods:

[0005] 1) First, high-resolution regional reanalysis data is produced and AIRWFM is directly trained using artificial intelligence (especially deep learning) methods. This method requires a lot of costs and has limited expected benefits. It also requires global models to provide boundary conditions, resulting in a lack of close connection between global and regional artificial intelligence models;

[0006] 2) Use high-resolution (re)analysis data (or observational data) to redevelop existing AIGWFMs, that is, use global high-resolution analysis data (such as global HRES data: https: / / rda.ucar.edu / datasets / ds113.1 / dataaccess / ) to train them and superimpose an artificial intelligence encryption module. For example, FengWu-GHR, Aurora and YanTian all convert HRES data into a data format acceptable to low-resolution models for input (Aurora's 3D Perceiver encoder, YanTian and FengWu-GHR are similar), of course, each model has a different processing module (such as FengWu-GHR's Decompositional and Combinational Transfer Learning, Aurora's 3DSwin Transformer U-Net, YanTian's "variance" module). However, these methods of secondary training models have not fully utilized the spatiotemporal forecast evolution contained in the original AIGWFMs model.

[0007] At present, all the schemes for improving the spatial resolution of large AI weather forecast models have their own drawbacks: directly generating regional high-resolution reanalysis data is extremely costly, and it is impossible to establish an effective connection with the global model and it is difficult to nest and encrypt it. It must rely on the global (traditional or artificial intelligence) model to continuously provide boundary fields; and the secondary development schemes for the existing coarse-resolution AIGWFMs models have failed to effectively reflect the spatiotemporal forecast evolution contained in the original AIGWFMs model.

[0008] In view of this, a method and device for constructing an optimized model for improving the resolution and forecast accuracy of a large meteorological model is provided, in order to at least partially solve the following technical problems existing in the prior art when optimizing the resolution and forecast accuracy of a large meteorological model:

[0009] 1) Directly generate global / regional high-resolution reanalysis data for high-resolution global / regional / artificial intelligence training, which has unbearable computing requirements;

[0010] 2) The artificial intelligence regional meteorological model AIRWFMs and the artificial intelligence global meteorological model AIGWFMs lack effective connections and are difficult to embed organically. The former still needs the latter to improve the boundary field;

[0011] 3) The encryption module developed based on the existing coarse-resolution AIGWFMs model fails to fully utilize the spatiotemporal forecast evolution contained in the AIGWFMs model, and currently can only perform global encryption, failing to achieve local (nested) encryption of the global (or regional) large model. Summary of the invention

[0012] To this end, an embodiment of the present invention provides an optimization model construction method and device for improving the resolution and forecast accuracy of a large artificial intelligence meteorological model to solve at least one of the above-mentioned technical problems.

[0013] In order to achieve the above purpose, the embodiment of the present invention provides the following technical solutions:

[0014] The present invention provides an optimization model construction method for improving the resolution and forecast accuracy of an artificial intelligence meteorological large model, the method comprising:

[0015] The high-resolution training data set is used to construct the initial field of the coarse grid resolution model, and the coarse grid resolution target artificial intelligence meteorological large model is driven to perform continuous forecasts for a preset period of time, so as to obtain the coarse grid resolution data to be optimized for multiple forecast steps respectively;

[0016] The data to be optimized corresponding to each forecast step are respectively input into the pre-built artificial intelligence data processing model, and combined with the training data set to obtain the error value corresponding to each predetermined forecast step;

[0017] The error values ​​obtained in each prediction step are summed, and the artificial intelligence data processing model is trained based on the sum of the error values ​​with minimization of the objective function as the training goal, so as to obtain an optimized artificial intelligence data processing model;

[0018] Each forecast step is divided into multiple preset time periods, and the artificial intelligence data processing model is trained with the coarse resolution indication data at the initial moment within the preset time period, the fine resolution indication data at the initial moment, and / or the coarse resolution indication data at the forecast moment within the preset time period as input data, and with the fine resolution indication data at the forecast moment of the preset time period as output data.

[0019] In some embodiments, constructing a model coarse grid resolution initial field using a high-resolution training data set specifically includes:

[0020] The high-resolution training data is interpolated to obtain the coarse-grid resolution pattern field.

[0021] In some embodiments, driving the target artificial intelligence meteorological big model to perform continuous forecasting for a preset period of time specifically includes:

[0022] Drive the target meteorological model to perform the initial time t j-1 To the forecast time t j Continuous forecast of j-1 =t0+(j-1)Δt,t j =t0+jΔt), j=1,…,N τ , △t is the time step, τ is the time interval of high-resolution training data, N τ is the number of consecutive forecast moments, and t0 is the initial moment of each time interval.

[0023] In some embodiments, the coarse resolution indicator data at the initial moment is the coarse resolution variable at the initial moment, the fine resolution indicator data at the initial moment is the fine resolution variable at the initial moment, the coarse resolution indicator data at the forecast moment is the coarse resolution variable at the forecast moment, and the fine resolution indicator data at the forecast moment is the fine resolution variable at the forecast moment; accordingly, the artificial intelligence data processing model The expression is:

[0024]

[0025] Among them, t j-1 represents the initial time, represents the time residual between the forecast and the initial coarse-resolution forecast variable, represents the spatial residual between the fine and coarse grid resolution variables at the initial moment, t j Indicates the forecast time.

[0026] In some embodiments, the objective function is:

[0027]

[0028] Where k(=1,…,T) refers to the frequency of high-resolution training data, x h,k represents the kth training data and

[0029] In some embodiments, the network architecture of the artificial intelligence data processing model includes but is not limited to a residual network, a U-type network, a graph neural network, or a neural network based on an attention mechanism.

[0030] The present invention also provides an optimization model construction device for improving the resolution and forecast accuracy of an artificial intelligence meteorological large model, the device comprising:

[0031] A data acquisition unit is used to construct an initial field of a coarse grid resolution model using a high-resolution training data set, and drive a coarse grid resolution target artificial intelligence meteorological large model to perform continuous forecasting for a preset period of time, so as to obtain data to be optimized for multiple forecast steps respectively;

[0032] The outer loop unit is used to input the data to be optimized corresponding to each forecast step into the pre-built artificial intelligence data processing model, and obtain the error value corresponding to each forecast step in combination with the training data; sum the error values ​​obtained in each forecast step, and take minimizing the objective function as the training goal, and train the artificial intelligence data processing model based on the sum of the error values ​​to obtain the optimized artificial intelligence data processing model;

[0033] The inner loop unit is used to divide each forecast step into multiple preset time periods. The artificial intelligence data processing model uses the coarse resolution indication data at the initial moment within the preset time period, the fine resolution indication data at the initial moment, and / or the coarse resolution indication data at the forecast moment within the preset time period as input data, and uses the fine resolution indication data at the forecast moment of the preset time period as output data for training.

[0034] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.

[0035] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and the computer program implements the steps of the above method when executed by a processor.

[0036] In one or more of the above specific embodiments, the optimization model construction method and device provided by the present invention for improving the resolution and forecast accuracy of the large meteorological model have the following technical effects:

[0037] Using the global / regional high-precision (re)analysis (or observation, fusion) data in the existing target data set as training data, the AI ​​global (or regional) large model AIG( / R)WFM (DepM l→l (·)) Global and local encryption functions, establishing effective connections between global and regional AI models, and being able to achieve multiple nested encryption through repeated encryption operations, and being compatible with traditional numerical forecasting models;

[0038] Generating data processing models In the process, the original coarse resolution artificial intelligence global (or regional) large model AIG( / R)WFM model (DepM l→l (·)) implies the temporal and spatial evolution of the forecast, that is, the input includes the AIG( / R)WFM model (DepM l→l (·)) Coarse-resolution simulation and forecast results at initial and forecast time;

[0039] Use artificial intelligence algorithms to train data processing models to generate better data optimization models

[0040]

[0041] In general, the present invention relates to fields such as meteorology, ecology, environment, new energy and artificial intelligence, and in particular to a method of improving the spatial resolution of an existing coarse-grid resolution artificial intelligence meteorological model by superimposing artificial intelligence models (such as deep neural networks, etc.), thereby improving the simulation and forecasting accuracy of meteorological, ecological and environmental variables. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0043] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0044] Figure 1 One of the flow charts of the optimization model construction method provided by the present invention for improving the resolution and forecast accuracy of the large meteorological model;

[0045] Figure 2 The second flowchart of the optimization model construction method for improving the resolution and forecast accuracy of the large meteorological model provided by the present invention;

[0046] Figure 3 This is a schematic diagram of the structure of an optimization model building device for improving the resolution and forecast accuracy of a large meteorological model provided by the present invention;

[0047] Figure 4 This is a structural block diagram of a computer device provided by the present invention. DETAILED DESCRIPTION

[0048] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, the optimization model construction method for improving the pre-resolution and forecast accuracy of the meteorological large model provided by the present invention is a general new solution for improving the resolution and forecast accuracy of the AI ​​meteorological large model. The optimization model construction method is applicable to the following two situations:

[0050] First, global encryption: At this time, an artificial intelligence global (or regional) large model AIG( / R)WFM(DepM l→l (·)) and the high-resolution (re)analysis (or observation, fusion) datasets of the same simulation area, namely AIG( / R)WFM(DepM l→l The simulated area of ​​(·)) is the same as that covered by the high-resolution (re)analysis (or observational, fusion) data, except that the AIG( / R)WFM (DepM l→l (·)) model resolution is lower than the (re)analysis (or observation, fusion) data resolution. In this case, the AI ​​global (or regional) large model AIG( / R)WFM (DepM l→l (·))'s global encryption function.

[0051] Second, local encryption: At this time, an artificial intelligence global (or regional) large model AIG( / R)WFM(DepM l→l (·)) and high-resolution regional (re)analysis (or observation, fusion) data sets, while high-resolution (re)analysis (or observation, fusion) data only partially cover the artificial intelligence global (or regional) large model AIG( / R)WFM(DepM l→l (·)) simulation area. This scenario implements the artificial intelligence global (or regional) large model AIG( / R)WFM(DepM l→l (·)) local encryption function, that is, a nested encryption.

[0052] In a specific embodiment, Figure 2 As shown, the optimization model construction method provided by the present invention for improving the resolution and forecast accuracy of the large meteorological model includes the following steps:

[0053] S210: constructing a coarse grid resolution model initial field using a high-resolution training data set (this set of data can use existing reanalysis data products, observation data, and reanalysis data products to drive high-resolution numerical forecast model downscaling generation) to drive the coarse grid resolution target artificial intelligence meteorological large model to perform continuous forecasting for a preset period of time, so as to obtain coarse grid resolution data to be optimized for multiple forecast steps respectively;

[0054] S220: inputting the data to be optimized corresponding to each forecast step into a pre-built artificial intelligence data processing model, respectively, and obtaining the error value corresponding to each forecast step in combination with the training data;

[0055] S230: summing the error values ​​obtained in each prediction step, and taking minimizing the objective function as the training goal, training the artificial intelligence data processing model based on the sum of the error values ​​to obtain an optimized artificial intelligence data processing model.

[0056] Among them, each forecast step is divided into multiple preset time periods, and the artificial intelligence data processing model uses the coarse resolution indication data at the initial moment within the preset time period, the fine resolution indication data at the initial moment, and / or the coarse resolution indication data at the forecast moment within the preset time period as input data, and uses the fine resolution indication data at the forecast moment of the preset time period as output data for training.

[0057] In step S210, constructing a low-grid resolution initial field of the model using a high-resolution training data set specifically includes:

[0058] The high-resolution training data is interpolated to obtain the coarse-grid resolution pattern field.

[0059] Among them, driving the target meteorological model to perform continuous forecasting for a preset period of time specifically includes:

[0060] Drive the coarse grid resolution target meteorological model to perform initial time t j-1 To the forecast time t j Continuous forecast of j-1 =t0+(j-1)Δt,t j =t0+jΔt), j=1,…,N τ , △t is the time step, τ is the time interval of high-resolution training data, N τ is the number of consecutive forecast moments, and t0 is the initial moment of each time interval. It should be understood that this is a cyclic and iterative continuous forecast process, and the data obtained at the forecast moment is used as the data of the initial moment of the next moment.

[0061] In a specific usage scenario, in step S210, firstly, the high-resolution (re)analysis (or observation, fusion) data Simple interpolation to a coarse grid resolution model field Driving the AI ​​Big Model AIG( / R)WFM(DepM l→l (·)) j-1 (=t0+(j-1)Δt) to t j (=t0+jΔt) time continuous forecast, where △t is AIG( / R)WFM(DepM l→l (·)) time step, And τ is the time interval of high-resolution training data.

[0062] It should be understood that the data processing model to be optimized is available for each forecast step. It takes the coarse and fine grid resolution fields at the initial moment and the coarse grid forecast field at the forecast step moment as input, while the fine grid data at the forecast moment is the forecast field, and the training data is only available at relatively large time intervals.

[0063] When designing a data processing model, the input and output of the model may be various indicative data related to resolution, such as the temporal residual between coarse resolution variables or the spatial residual between coarse and fine resolution variables.

[0064] Specifically, the coarse resolution indicator data at the initial moment is the coarse resolution variable at the initial moment, the fine resolution indicator data at the initial moment is the fine resolution variable at the initial moment, the coarse resolution indicator data at the forecast moment is the coarse resolution variable at the forecast moment, and the fine resolution indicator data at the forecast moment is the fine resolution variable at the forecast moment; at this time, the artificial intelligence data processing model The expression is:

[0065]

[0066] Among them, t j-1 represents the initial time, represents the coarse-resolution variable at the initial moment, represents the fine-resolution variable at the initial moment, t j Indicates the forecast time. represents the coarse-resolution variable at the forecast time, represents the fine-resolution variable at the forecast time. In other words, the input of the data processing model is t j-1 Time coarse and fine resolution data With t j Coarse resolution data at the moment in Demonstrating AI encryption model Incorporated into the AIG( / R)WFM Model (DepM l→l (·)) implies the evolution of the spatiotemporal forecasts.

[0067] In some other embodiments, the coarse resolution indicator data at the initial moment is the time residual between the coarse resolution at the forecast moment and the coarse resolution variable at the initial moment, and the fine resolution indicator data at the initial moment is the spatial residual between the fine resolution and the coarse resolution variable at the initial moment; accordingly, the artificial intelligence data processing model The expression is:

[0068]

[0069] Among them, t j-1 represents the initial time, represents the time residual between the forecast and the initial coarse-resolution forecast variable, represents the spatial residual between the fine and coarse grid resolution variables at the initial moment, t j Indicates the forecast time.

[0070] In other words, the above data processing model to be trained The input and output of the model can have various variations. For example, the input can be the AIG( / R)WFM model (DepM l→l (·)) The residual of the forecast t j-1 High-resolution forecast at any time and AIG( / R)WFM(DepM l→l (·)) Model coarse resolution prediction The residual At this time, the coarse resolution result After simple interpolation to a high spatial resolution, the residual is calculated, and the output can be t j High-resolution forecast at any time and AIG( / R)WFM Model (DepM l→l (·))forecast The residual At this time, the coarse resolution result The residual is then calculated after simple interpolation to a high spatial resolution. It should be understood that this case particularly embodies the idea of ​​multi-scale forecasting. The coarse resolution AIG( / R)WFM model (DepM l→l (·)) The model predicts the long-wave information in the atmosphere, and its residual or increment is exactly the forecast increment of this long-wave information; and the residual This shows that the initial high-resolution forecast is better than the AIG( / R)WFM model (DepMl→l (·)) Shortwave increment of coarse resolution forecast.

[0071] In some embodiments, the high-resolution training data in the high-resolution data set (for example, the existing high-resolution (re)analysis (or observation, fusion) data) is used to train using artificial intelligence methods to obtain a data processing model. The training objective is to minimize the following function:

[0072]

[0073] The subscript k (=1,…,T) refers to the frequency of high-resolution training data. h,k represents the kth training data and The data processing model to be determined in the above formula It can be any suitable artificial intelligence model, including but not limited to residual network, U-shaped network, graph neural network and neural network based on attention mechanism. The usual gradient descent method can be used to The model parameters are solved.

[0074] The network architecture of the artificial intelligence data processing model includes but is not limited to a residual network, a U-shaped network, a graph neural network, or a neural network based on an attention mechanism.

[0075] In one or more of the above specific embodiments, the optimization model construction method for improving the resolution and forecast accuracy of the artificial intelligence meteorological model provided by the present invention has the following technical effects:

[0076] Using the global / regional high-precision (re)analysis (or observation, fusion) data in the existing target data set as training data, the AI ​​global (or regional) large model AIG( / R)WFM (DepM l→l (·)) Global and local encryption functions, establishing effective connections between global and regional AI models, and being able to achieve multiple nested encryption through repeated encryption operations, and being compatible with traditional numerical forecasting models;

[0077] Generating data processing models In the process, the original coarse resolution artificial intelligence global (or regional) large model AIG( / R)WFM model (DepM l→l (·)) implies the temporal and spatial evolution of the forecast, that is, the input includes the AIG( / R)WFM model (DepM l→l (·)) Coarse-resolution simulation and forecast results at initial and forecast time;

[0078] Use artificial intelligence algorithms to train data processing models to generate better artificial intelligence data processing models

[0079] In general, the present invention relates to fields such as meteorology, ecology, environment, new energy and artificial intelligence, and in particular to a method of improving the spatial resolution of an existing coarse-grid resolution artificial intelligence meteorological model by superimposing artificial intelligence models (such as deep neural networks, etc.), thereby improving the simulation and forecasting accuracy of meteorological, ecological and environmental variables.

[0080] In addition to the above method, the present invention also provides an optimization model construction device for improving the resolution and forecast accuracy of the artificial intelligence meteorological model, such as Figure 3 As shown, the device comprises:

[0081] The data acquisition unit 310 is used to construct a coarse grid resolution model field using a high-resolution training data set, and drive the target artificial intelligence meteorological large model to perform continuous forecasting for a preset period of time, so as to obtain data to be optimized for multiple forecast steps respectively;

[0082] The outer loop unit 320 is used to input the data to be optimized corresponding to each forecast step into the pre-built artificial intelligence data processing model, and obtain the error value corresponding to each forecast step in combination with the training data; sum the error values ​​obtained in each forecast step, and take minimizing the objective function as the training goal, and train the artificial intelligence data processing model based on the sum of the error values ​​to obtain a data optimization model;

[0083] The inner loop unit 330 is used to divide each forecast step into multiple preset time periods. The artificial intelligence data processing model uses the coarse resolution indication data at the initial moment within the preset time period, the fine resolution indication data at the initial moment, and / or the coarse resolution indication data at the forecast moment within the preset time period as input data, and uses the fine resolution indication data at the forecast moment of the preset time period as output data for training.

[0084] In some embodiments, constructing a model coarse grid resolution initial field using a high-resolution training data set specifically includes:

[0085] The high-resolution training data is interpolated to obtain the coarse-grid resolution pattern field.

[0086] In some embodiments, driving the target artificial intelligence meteorological big model to perform continuous forecasting for a preset period of time specifically includes:

[0087] Drive the coarse grid resolution target meteorological model to perform initial time t j-1 To the forecast time t j Continuous forecast of j-1 =t0+(j-1)Δt,t j =t0+jΔt), j=1,…,Nτ , △t is the time step, τ is the time interval of high-resolution training data, N τ is the number of consecutive forecast moments, and t0 is the initial moment of each time interval.

[0088] In some embodiments, the coarse resolution indicator data at the initial moment is the coarse resolution variable at the initial moment, the fine resolution indicator data at the initial moment is the fine resolution variable at the initial moment, the coarse resolution indicator data at the forecast moment is the coarse resolution variable at the forecast moment, and the fine resolution indicator data at the forecast moment is the fine resolution variable at the forecast moment; accordingly, the artificial intelligence data processing model The expression is:

[0089]

[0090] Among them, t j-1 represents the initial time, Table 1 shows the coarse resolution mode variables at the initial moment, represents the fine-resolution model variable at the initial moment, t j represents the forecast time pattern variable, represents the coarse-resolution model variable at the forecast time, Represents the fine-resolution model variable at the forecast time.

[0091] In some embodiments, the coarse resolution indicator data at the initial time is the time residual between the coarse resolution at the forecast time and the coarse resolution variable at the initial time, and the fine resolution indicator data at the initial time is the spatial residual between the fine resolution at the initial time and the coarse resolution variable; accordingly, the artificial intelligence data processing model The expression is:

[0092]

[0093] Among them, t j-1 represents the initial time, represents the spatial residual between the forecast and the initial coarse-resolution variable forecast, represents the spatial residual between the fine and coarse grid resolution variables at the initial moment, t j Indicates the forecast time.

[0094] In some embodiments, the objective function is:

[0095]

[0096] Where k(=1,…,T) refers to the frequency of high-resolution training data, x h,k represents the kth training data and

[0097] In some embodiments, the network architecture of the artificial intelligence data processing model includes but is not limited to a residual network, a U-shaped network, a graph neural network, or a neural network based on an attention mechanism.

[0098] In one or more of the above specific embodiments, the optimization model construction device for improving the resolution and forecast accuracy of the artificial intelligence meteorological model provided by the present invention has the following technical effects:

[0099] Using the global / regional high-precision (re)analysis (or observation, fusion) data in the existing target data set as training data, the AI ​​global (or regional) large model AIG( / R)WFM (DepM l→l (·)) Global and local encryption functions, establishing effective connections between global and regional AI models, and being able to achieve multiple nested encryption through repeated encryption operations, and being compatible with traditional numerical forecasting models;

[0100] Generating data processing models In the process, the original coarse resolution artificial intelligence global (or regional) large model AIG( / R)WFM model (DepM l→l (·)) implies the temporal and spatial evolution of the forecast, that is, the input includes the AIG( / R)WFM model (DepM l→l (·)) Coarse-resolution simulation and forecast results at initial and forecast time;

[0101] Use artificial intelligence algorithms to train data processing models to generate better data optimization models

[0102]

[0103] In general, the present invention relates to fields such as meteorology, ecology, environment, new energy and artificial intelligence, and in particular to a method of improving the spatial resolution of an existing coarse-grid resolution artificial intelligence meteorological model by superimposing artificial intelligence models (such as deep neural networks, etc.), thereby improving the simulation and forecasting accuracy of meteorological, ecological and environmental variables.

[0104] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4As shown. The computer device includes a processor, a memory and a network interface connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a model prediction. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The model prediction of the computer device is used to store static information and dynamic information data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps in the above method embodiment are implemented.

[0105] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0106] Corresponding to the above embodiment, the embodiment of the present invention further provides a computer storage medium, which contains one or more program instructions, wherein the one or more program instructions are used to execute the above method.

[0107] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the above method.

[0108] In an embodiment of the present invention, the processor is a GPU card for AI training, such as NVIDIA H100, Moore's Law, etc.

[0109] The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and completes the steps of the above method in combination with its hardware.

[0110] The storage medium may be a memory, which may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0111] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0112] The volatile memory may be a random access memory (RAM) which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DRRAM).

[0113] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0114] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by general or special-purpose computers.

[0115] The above specific implementation methods further illustrate the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for constructing an optimization model to improve the resolution and forecast accuracy of an artificial intelligence meteorological large model, characterized in that: The method comprises: The high-resolution training data set is used to construct the initial field of the coarse grid resolution model, and the coarse grid resolution target artificial intelligence meteorological large model is driven to perform continuous forecasts for a preset period of time, so as to obtain the coarse grid resolution data to be optimized for multiple forecast steps respectively; The data to be optimized corresponding to each forecast step are respectively input into the pre-built artificial intelligence data processing model, and the error value corresponding to each forecast step is obtained by combining the training data; The error values ​​obtained in each prediction step are summed, and the artificial intelligence data processing model is trained based on the sum of the error values ​​with minimization of the objective function as the training goal, so as to obtain an optimized artificial intelligence data processing model; Each forecast step is divided into multiple preset time periods, and the artificial intelligence data processing model is trained with the coarse resolution indication data at the initial moment within the preset time period, the fine resolution indication data at the initial moment, and / or the coarse resolution indication data at the forecast moment within the preset time period as input data, and with the fine resolution indication data at the forecast moment of the preset time period as output data.

2. The optimization model construction method for improving the resolution and forecast accuracy of the artificial intelligence meteorological model according to claim 1 is characterized in that: The coarse grid resolution model initial field is constructed using the high-resolution training data set, including: The high-resolution training data is interpolated to obtain the coarse-grid resolution pattern field.

3. The optimization model construction method for improving the resolution and forecast accuracy of the artificial intelligence meteorological model according to claim 1 is characterized in that: Drive the coarse grid resolution target artificial intelligence meteorological model to perform continuous forecasts for a preset period of time, including: Drive the coarse grid resolution target meteorological model to perform initial time t j-1 To the forecast time t j Continuous forecast of j-1 =t0+(j-1)Δt,t j =t0+jΔt), j=1,…,N τ , △t is the time step, τ is the time interval of high-resolution training data, N τ is the number of consecutive forecast moments, and t0 is the initial moment of each time interval.

4. The optimization model construction method for improving the resolution and forecast accuracy of the artificial intelligence meteorological model according to claim 1 is characterized in that: The coarse resolution indicator data at the initial moment is the coarse resolution variable at the initial moment, the fine resolution indicator data at the initial moment is the fine resolution variable at the initial moment, the coarse resolution indicator data at the forecast moment is the coarse resolution variable at the forecast moment, and the fine resolution indicator data at the forecast moment is the fine resolution variable at the forecast moment; accordingly, the artificial intelligence data processing model The expression is: Among them, t j-1 represents the initial time, represents the time residual between the forecast and the initial coarse-resolution forecast variable, represents the spatial residual between the fine and coarse grid resolution variables at the initial moment, t j Indicates the forecast time.

5. The optimization model construction method for improving the resolution and forecast accuracy of a large meteorological model according to claim 4, characterized in that: The objective function is: Where k(=1,…,T) refers to the frequency of high-resolution training data, x h,k represents the kth training data and 6. The optimization model construction method for improving the resolution and forecast accuracy of a large meteorological model according to claim 5, characterized in that: The network architecture of the artificial intelligence data processing model includes but is not limited to a residual network, a U-shaped network, a graph neural network, or a neural network based on an attention mechanism.

7. An optimization model construction device for improving the resolution and forecast accuracy of an artificial intelligence meteorological large model, characterized in that: The device comprises: A data acquisition unit is used to construct an initial field of a coarse grid resolution model using a high-resolution training data set, and drive a coarse grid resolution target artificial intelligence meteorological large model to perform continuous forecasting for a preset period of time, so as to obtain data to be optimized for multiple forecast steps respectively; The outer loop unit is used to input the data to be optimized corresponding to each forecast step into the pre-built artificial intelligence data processing model, and obtain the error value corresponding to each forecast step in combination with the training data; sum the error values ​​obtained in each forecast step, and take minimizing the objective function as the training goal, and train the artificial intelligence data processing model based on the sum of the error values ​​to obtain the optimized artificial intelligence data processing model; The inner loop unit is used to divide each forecast step into multiple preset time periods. The artificial intelligence data processing model uses the coarse resolution indication data at the initial moment within the preset time period, the fine resolution indication data at the initial moment, and / or the coarse resolution indication data at the forecast moment within the preset time period as input data, and uses the fine resolution indication data at the forecast moment of the preset time period as output data for training.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images