New energy extreme meteorological data prediction method and device, server and storage medium

Abstract

The invention discloses a new energy extreme meteorological data prediction method and device, a server and a storage medium. The method comprises the steps of obtaining historical meteorological data of a new energy station; constructing an assimilation model based on the generative diffusion model, and optimizing the assimilation model by using the constructed gradient; generating a plurality of training weather initial fields based on the assimilation model; training a meteorological model based on the training weather initial field, wherein the meteorological model is iteratively optimized by using a flat troposphere loss function; and using the meteorological model to obtain a weather prediction result according to an input weather initial field. And the prediction accuracy and stability can be further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of supply and demand forecasting, and in particular to a new energy extreme weather data forecasting method, device, server and storage medium. Background Art

[0002] New energy power stations (wind power stations and photovoltaic power stations) have very strict requirements on the accuracy of weather forecasts during operation, especially accurate weather forecasts under extreme weather conditions are particularly important. Traditional NWP models mainly rely on physical laws and empirical formulas, which are complex and computationally intensive, and sometimes cannot fully reflect actual weather changes in actual forecasts. In recent years, deep learning technology, especially deep learning-based large meteorological models such as Pangu-Weather and Graphcast, has received widespread attention and application due to its advantages in processing large-scale meteorological data and improving prediction accuracy. These models perform better than traditional NWP models in terms of root mean square error (RMSE) in medium-term weather forecasts, and have shown better performance in some extreme weather forecasts, such as typhoon paths.

[0003] However, the above models are prone to fuzzy prediction results in data-driven weather forecasting, and cannot accurately reflect the details of weather changes, especially the prediction level under extreme weather conditions. At the same time, the training data used is more dependent on the provided data set, and is not closely integrated with the specific site weather, which leads to low accuracy of fine weather forecasts for new energy power stations. Summary of the invention

[0004] The embodiments of the present invention provide a new energy extreme weather data prediction method, device, server and storage medium to solve the technical problem of low prediction accuracy of extreme weather for new energy power stations in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides a new energy extreme weather data prediction method, comprising:

[0006] Obtain historical meteorological data of new energy stations;

[0007] Construct an assimilation model based on the generative diffusion model and optimize the assimilation model using the constructed gradient;

[0008] generating a plurality of training weather initial fields based on the assimilation model;

[0009] Training a meteorological model based on an initial training weather field, wherein the meteorological model is iteratively optimized using a flat tropospheric loss function;

[0010] The meteorological model is used to obtain a weather forecast result according to the input initial weather field.

[0011] In a second aspect, an embodiment of the present invention further provides a new energy extreme weather data prediction device, comprising:

[0012] Acquisition module, used to obtain historical meteorological data of new energy stations;

[0013] A construction module is used to construct an assimilation model based on the generative diffusion model and optimize the assimilation model using the constructed gradient;

[0014] A generation module, used for generating a plurality of training weather initial fields based on the assimilation model;

[0015] A training module, used for training a meteorological model based on an initial training weather field, wherein the meteorological model is iteratively optimized using a flat tropospheric loss function;

[0016] The obtaining module is used to obtain the weather forecast result according to the input initial weather field by using the meteorological model.

[0017] In a third aspect, an embodiment of the present invention further provides a server, the server comprising:

[0018] one or more processors;

[0019] a storage device for storing one or more programs,

[0020] When the one or more programs are executed by the one or more processors, the one or more processors implement any new energy extreme weather data prediction method provided in the above embodiments.

[0021] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute the new energy extreme weather data prediction method provided in the above embodiment.

[0022] The extreme weather data prediction method, device, server and storage medium provided by the embodiment of the present invention obtain the historical weather data of the new energy station; construct an assimilation model based on the generative diffusion model, and optimize the assimilation model using the constructed gradient; generate multiple training weather initial fields based on the assimilation model; train the meteorological model based on the training weather initial field, and the meteorological model is iteratively optimized using the flat tropospheric loss function; and obtain the weather forecast result according to the input weather initial field using the meteorological model. By introducing the flat tropospheric error loss function, the prediction accuracy of the model in extreme weather is improved, and a variety of extreme weather elements are introduced for model training so that it can be applied to the new energy extreme weather prediction scenario. At the same time, combined with the initial field assimilation technology based on the generative model, multiple initial fields are generated by integrating the measured data of the new energy station, and ensemble forecasting is performed to further improve the prediction accuracy and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0024] Figure 1 It is a flow chart of a new energy extreme weather data prediction method provided in the first embodiment of the present invention;

[0025] Figure 2 It is a flow chart of a new energy extreme weather data prediction method provided in Embodiment 2 of the present invention;

[0026] Figure 3 It is a structural schematic diagram of a new energy extreme weather data prediction device provided in Embodiment 3 of the present invention;

[0027] Figure 4 It is a schematic diagram of the structure of the server provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0029] Embodiment 1

[0030] Figure 1 : is a flow chart of a method for predicting extreme weather data for new energy provided in Example 1 of the present invention. This embodiment is applicable to the case of accurately predicting extreme weather at new energy stations. The method can be executed by a new energy extreme weather data prediction device, and specifically includes the following steps:

[0031] Step 110, obtaining historical meteorological data of new energy stations.

[0032] In this embodiment, it is necessary to provide a more accurate extreme weather forecast for the specific location of the new energy station. Therefore, in this embodiment, it is first necessary to obtain the historical weather data of the new energy station.

[0033] Exemplarily, the historical meteorological data may include: measured data of new energy sites such as wind speed, temperature, humidity, air pressure, total radiation, direct radiation, diffuse radiation and other meteorological data.

[0034] Furthermore, the method may also add the following steps: cleaning and standardizing the historical meteorological data of the new energy stations.

[0035] For example, the measured data is first cleaned, for example, dead values, missing values ​​and outliers are removed from the measured data. After cleaning, the data is standardized. The data is converted to a uniform scale. For example, the above method is used to achieve:

[0036]

[0037] Where X norm Represents the data after standardization, X is the original data, μ is the data mean, and σ is the data standard deviation.

[0038] Step 120: construct an assimilation model based on the generative diffusion model, and optimize the assimilation model using the constructed gradient.

[0039] When training, it is necessary not only to use the historical meteorological data of the new energy station, but also other meteorological big data to expand the types of data sources and capture the extreme meteorological characteristics. However, the two data sources cannot be directly mixed and used. Therefore, in this embodiment, it is necessary to assimilate the two types of data first so that the formed training set can fully retain the meteorological characteristics of the new energy station and the characteristics of the extreme meteorological data in the big data.

[0040] For example, the measured data of new energy stations such as wind speed, temperature, humidity, air pressure, total radiation, direct radiation, diffuse radiation, etc., as well as other global meteorological measured data, historical initial field and reanalysis meteorological data are integrated to form a training data set, and the assimilation model is used to assimilate it.

[0041] In this embodiment, the assimilation model may adopt the SDA assimilation model, or may adopt an assimilation model such as DiffDA.

[0042] First, an assimilation model is constructed based on a generative diffusion model. Exemplarily, the generative diffusion model is implemented in the following manner:

[0043]

[0044] Among them, p(x0|y) represents the posterior probability of the initial field x0 given the observation data y; p(y|x0) is the likelihood of the observation data; p(x0) is the prior distribution of the initial field. Furthermore, the score matching method can be used to construct the gradient in the calculation assimilation process in the SDA model as follows:

[0045]

[0046] Gradients can provide an update direction so that the assimilation model can better assimilate the observed data by gradually adjusting the data state, thereby improving the accuracy and reliability of the model. Specifically, in the diffusion model, through back propagation and multi-step update process, gradients can be used to gradually denoise the data, and finally generate a noise-free state that is more consistent with the observed data, so that it gradually approaches the target state over multiple time steps, ensuring that the final generated result is consistent with both the physical prior and the observed data.

[0047] Step 130: Generate multiple training weather initial fields based on the assimilation model.

[0048] The trained SDA model input is used to integrate the measured data of new energy stations in real time to generate different initial fields, providing diverse initial conditions for ensemble forecasts. And a variety of possible initial fields are generated through the diffusion process:

[0049] x t =x0+t·∈

[0050] Among them, x t represents the intermediate state in the diffusion process, t is the time step, and ∈ is the noise.

[0051] Furthermore, the method of gradually reversing the generation of the initial field at adjacent moments is:

[0052] x t-1 =x t -η·Score(x t ,y).

[0053] Step 140, training a meteorological model based on the training weather initial field, wherein the meteorological model is iteratively optimized using a flat tropospheric loss function.

[0054] Based on the GraphCast model, a prediction model based on extreme weather optimization is trained. The model is trained using historical reanalysis data meteorological data, and advection loss is used to replace the traditional RMSE loss function to improve the prediction accuracy under extreme weather conditions, resulting in Graphcast_ACEloss.

[0055] For example, based on the pre-processed historical reanalysis data, in terms of the selection of elements, compared with the existing Graphcast pre-trained model, new elements such as ERA5 snowfall, water vapor content of the entire atmosphere, cloud water content of each pressure layer, cloud ice content of each pressure layer, 100m wind speed, gusts, etc. are added to increase the applicability of the model to extreme weather such as ice cover, strong winds, and snow accumulation. The dust and particulate matter concentration in the CAMS reanalysis data are added to increase the applicability of predictions for haze, dust, etc. The accuracy of extreme weather data prediction can be further enhanced by assimilating multiple data.

[0056] In this embodiment, the original ACE loss function in the model is replaced by the flat tropospheric loss function, so as to utilize the influence of the change in flat convection in meteorological changes on the meteorology and improve the accuracy of extreme meteorological data prediction.

[0057] Exemplarily, the flat tropospheric loss function is implemented in the following manner:

[0058]

[0059] Among them, AE represents advection error and CE represents convection error.

[0060]

[0061] Among them, I o The input image represents the initial field, that is, the weather state at the beginning of the forecast. f Represents the real target field image, that is, the actual weather conditions at the prediction time point. represents the target field image predicted by the model; f v represents the advection error calculation function,

[0062]

[0063] Among them, v x ,v y Represents the horizontal wind speed component of the real target field, represents the horizontal wind speed component predicted by the model. re It is a reconstruction function used to remove the advection effect from the target field image and extract the convection component.

[0064] The model is iteratively optimized using the flat tropospheric loss function to obtain the trained meteorological model.

[0065] Step 150, using the meteorological model to obtain a weather forecast result according to the input initial weather field.

[0066] Using the trained extreme weather prediction model Graphcast_ACEloss, and the assimilation model provided in the previous steps, multiple initial fields are obtained based on the recently acquired meteorological information of new energy stations and the latest big data meteorological information. Multiple generated initial fields are input to perform multiple sets of ensemble forecasts, thereby improving the reliability and stability of the forecast results. Ensemble forecasts generate multiple sets of forecast results, for example,

[0067] P={p1,p2,...,p n}

[0068] Where P is the ensemble forecast result set, p iIndicates the i-th group of forecast results. Combined with the specific conditions of the initial field, multiple groups of forecast results are averaged by weight to generate extreme weather forecast results for new energy stations.

[0069] This embodiment obtains historical meteorological data of new energy stations; constructs an assimilation model based on a generative diffusion model, and optimizes the assimilation model using the constructed gradient; generates multiple training weather initial fields based on the assimilation model; trains the meteorological model based on the training weather initial field, and the meteorological model is iteratively optimized using a flat tropospheric loss function; and uses the meteorological model to obtain weather forecast results based on the input weather initial field. By introducing the advection-convection error loss function, the prediction accuracy of the model in extreme weather is improved, and a variety of extreme weather elements are introduced for model training so that it can be applied to new energy extreme weather prediction scenarios. At the same time, combined with the initial field assimilation technology based on the generative model, multiple initial fields are generated using data measured by the new energy station, and ensemble forecasts are performed to further improve the prediction accuracy and stability.

[0070] Embodiment 2

[0071] Figure 2 : is a flow chart of a new energy extreme weather data prediction method provided by the second embodiment of the present invention. This embodiment is optimized based on the above embodiment, and the meteorological model is iteratively optimized using the flat tropospheric loss function. The specific optimization is: using the ACE loss function to evaluate the prediction performance of the model, optimizing the model parameters, ensuring the high accuracy of the prediction results under extreme weather conditions, and optimizing the ACE loss function to minimize the ACE loss through the gradient descent algorithm, and updating the model parameters: Among them, θ is the model parameter and η is the learning rate.

[0072] See also Figure 2 , the new energy extreme weather data prediction method comprises:

[0073] Step 210, obtaining historical meteorological data of the new energy station.

[0074] Step 220: construct an assimilation model based on the generative diffusion model, and optimize the assimilation model using the constructed gradient.

[0075] Step 230: Generate multiple training weather initial fields based on the assimilation model.

[0076] Step 240, train the meteorological model based on the initial training weather field, use the ACE loss function to evaluate the prediction performance of the model, optimize the model parameters, ensure the high accuracy of the prediction results under extreme weather conditions, and optimize the ACE loss function to minimize the ACE loss through the gradient descent algorithm and update the model parameters.

[0077] In this embodiment, the loss function can be used to iteratively optimize various parameters of the evaluation model. Exemplarily, this is achieved in the following way:

[0078]

[0079] Among them, θ is the model parameter and η is the learning rate.

[0080] Step 250, using the meteorological model to obtain weather forecast results based on the input initial weather field.

[0081] In this embodiment, the meteorological model is iteratively optimized using the flat tropospheric loss function. Specifically, the optimization is as follows: the prediction performance of the model is evaluated using the ACE loss function, the model parameters are optimized, and the high accuracy of the prediction results under extreme weather conditions is ensured. The ACE loss function optimization minimizes the ACE loss through the gradient descent algorithm and updates the model parameters: Among them, θ is the model parameter and η is the learning rate. Using the above method, the parameters can be optimized step by step by setting the learning rate, so that the loss function gradually approaches the minimum, and the learning rate can be set to meet different training requirements of fast or small step iteration.

[0082] Embodiment 3

[0083] Figure 3 is a schematic diagram of the structure of the new energy extreme weather data prediction device provided by the third embodiment of the present invention, see Figure 3 , the new energy extreme weather data prediction device comprises:

[0084] An acquisition module 310 is used to acquire historical meteorological data of new energy stations;

[0085] A construction module 320 is used to construct an assimilation model based on the generative diffusion model and optimize the assimilation model using the constructed gradient;

[0086] A generating module 330, configured to generate a plurality of training weather initial fields based on the assimilation model;

[0087] A training module 340 is used to train a meteorological model based on a training weather initial field, wherein the meteorological model is iteratively optimized using a flat tropospheric loss function;

[0088] The obtaining module 350 is used to obtain the weather forecast result according to the input initial weather field by using the meteorological model.

[0089] The new energy extreme weather data prediction device provided in this embodiment obtains historical weather data of new energy stations; constructs an assimilation model based on a generative diffusion model, and optimizes the assimilation model using the constructed gradient; generates multiple training weather initial fields based on the assimilation model; trains the meteorological model based on the training weather initial field, and the meteorological model is iteratively optimized using a flat tropospheric loss function; and uses the meteorological model to obtain weather forecast results based on the input weather initial field. By introducing the advection-convection error loss function, the prediction accuracy of the model under extreme weather is improved, and a variety of extreme weather elements are introduced for model training so that it can be applied to new energy extreme weather prediction scenarios. At the same time, combined with the initial field assimilation technology based on the generative model, multiple initial fields are generated using the data measured by the new energy station, and ensemble forecasts are performed to further improve the prediction accuracy and stability.

[0090] Based on the above embodiments, the calculation module is implemented in the following manner:

[0091]

[0092] Where I(·,·) represents the mutual information operation function, β is the Lagrange multiplier, V t is the user task vector, Z is the user authority vector, S is the user state vector, and Z * is the user permission vector that satisfies the least privileged state.

[0093] Based on the above embodiments, the adjustment module includes:

[0094] A restoration unit, used to restore the user authority vector to a current task authority list using an inverse mapping function;

[0095] A conversion unit, used to convert the digital code in the current task authority list into a corresponding authority;

[0096] The updating unit is used to update the user rights according to the corresponding rights, allocate new rights and reclaim unused rights.

[0097] Based on the above embodiments, the fusion module realizes the fusion of identity information vector and authority vector in the following manner:

[0098] s=V u ·M·V p , where s is the user's current user state vector, V u and V p are the user identity vector and user authority vector after mapping, respectively. M is a learnable weight matrix obtained by random initialization.

[0099] Based on the above embodiments, the generative diffusion model is implemented in the following manner:

[0100]

[0101] Among them, p(x0|y) represents the posterior probability of the initial field x0 given the observed data y; p(y|x0) is the likelihood of the observed data; and p(x0) is the prior distribution of the initial field.

[0102] Based on the above embodiments, the gradient of the construction is as follows:

[0103]

[0104] On the basis of the above embodiments, the generation of multiple training weather initial fields based on the assimilation model is implemented in the following manner:

[0105] x t =x0+t·∈,

[0106] Among them, x t represents the intermediate state in the diffusion process, t is the time step, and ∈ is the noise;

[0107] Correspondingly, the method of gradually reversing the generation of the initial field at adjacent moments is:

[0108] x t-1 =x t -η·Score(x t ,y).

[0109] On the basis of the above embodiments, the flat tropospheric loss function is implemented in the following manner:

[0110]

[0111] Among them, AE represents advection error, CE represents convection error,

[0112]

[0113] Among them, I o The input image represents the initial field, that is, the weather state at the beginning of the forecast. f Represents the real target field image, that is, the actual weather conditions at the prediction time point. represents the target field image predicted by the model; f v represents the advection error calculation function,

[0114]

[0115] Among them, v x ,v yRepresents the horizontal wind speed component of the real target field, represents the horizontal wind speed component predicted by the model. re It is a reconstruction function used to remove the advection effect from the target field image and extract the convection component.

[0116] Based on the above embodiments, the meteorological model is iteratively optimized using the flat tropospheric loss function, including:

[0117] Use the ACE loss function to evaluate the prediction performance of the model, optimize the model parameters, and ensure the high accuracy of the prediction results under extreme weather conditions. The ACE loss function optimization minimizes the ACE loss through the gradient descent algorithm and updates the model parameters:

[0118]

[0119] Among them, θ is the model parameter and η is the learning rate.

[0120] Based on the above embodiments, the device further includes:

[0121] The standardized processing module is used to clean and standardize the historical meteorological data of new energy stations.

[0122] The new energy extreme weather data prediction device provided in the embodiment of the present invention can execute the new energy extreme weather data prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0123] Embodiment 4

[0124] Figure 4 A schematic diagram of the structure of a server provided in Embodiment 4 of the present invention. Figure 4 A block diagram of an exemplary server 12 suitable for use in implementing embodiments of the present invention is shown. Figure 4 The server 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0125] like Figure 4 As shown, the server 12 is in the form of a general purpose computing device. The components of the server 12 may include, but are not limited to: one or more processors or processing units 16, server memory 28, and a bus 18 connecting different server components (including the server memory 28 and the processing unit 16).

[0126] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0127] The server 12 typically includes a variety of computer server readable media. These media can be any available media that can be accessed by the server 12, including volatile and non-volatile media, removable and non-removable media.

[0128] Server memory 28 may include computer server readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. Server 12 may further include other removable / non-removable, volatile / non-volatile computer server storage media. By way of example only, storage server 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 4 not shown, usually called a "hard drive"). Although Figure 4 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The server memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0129] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the server memory 28, such program modules 42 including, but not limited to, operating a server, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0130] The server 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), may communicate with one or more devices that enable a user to interact with the server 12, and / or may communicate with any device that enables the server 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the server 12 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the server 12 via a bus 18. It should be understood that, although not shown in the figure, the server 12 may be assembled to use other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID servers, tape drives, and data backup storage servers, etc.

[0131] The processing unit 16 executes various functional applications and data processing by running the programs stored in the server memory 28, such as implementing the new energy extreme weather data prediction method provided in the embodiment of the present invention.

[0132] Embodiment 5

[0133] Embodiment 5 of the present invention further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute any of the new energy extreme weather data prediction methods provided in the above embodiments.

[0134] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer 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. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0135] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0136] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0137] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or device. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0138] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A new energy extreme weather data prediction method, characterized in that: include: Obtain historical meteorological data of new energy stations; Construct an assimilation model based on the generative diffusion model and optimize the assimilation model using the constructed gradient; generating a plurality of training weather initial fields based on the assimilation model; Training a meteorological model based on an initial training weather field, wherein the meteorological model is iteratively optimized using a flat tropospheric loss function; The meteorological model is used to obtain a weather forecast result according to the input initial weather field.

2. The method according to claim 1, characterized in that The generative diffusion model is implemented in the following way: Among them, p(x0|y) represents the posterior probability of the initial field x0 given the observed data y; p(y|x0) is the likelihood of the observed data; and p(x0) is the prior distribution of the initial field.

3. The method according to claim 1, characterized in that The gradients for the build are as follows:

4. The method according to claim 1, characterized in that: The generation of multiple training weather initial fields based on the assimilation model is achieved by: x t =x0+t·∈, Among them, x t represents the intermediate state in the diffusion process, t is the time step, and ∈ is the noise; Correspondingly, the method of gradually reversing the generation of the initial field at adjacent moments is: x t-1 =x t -η·Score(x t ,y)。 5. The method according to claim 1, characterized in that The flat tropospheric loss function is implemented in the following way: Among them, AE represents advection error, CE represents convection error, Among them, I o The input image represents the initial field, that is, the weather state at the beginning of the forecast. f Represents the real target field image, that is, the actual weather conditions at the prediction time point. represents the target field image predicted by the model; f v represents the advection error calculation function, Among them, v x ,v y Represents the horizontal wind speed component of the real target field, represents the horizontal wind speed component predicted by the model. re It is a reconstruction function used to remove the advection effect from the target field image and extract the convection component.

6. The method according to claim 1, characterized in that The meteorological model is iteratively optimized using a flat tropospheric loss function, including: Use the ACE loss function to evaluate the prediction performance of the model, optimize the model parameters, and ensure the high accuracy of the prediction results under extreme weather conditions. The ACE loss function optimization minimizes the ACE loss through the gradient descent algorithm and updates the model parameters: Among them, θ is the model parameter and η is the learning rate.

7. The method according to claim 1, characterized in that The method further comprises: Clean and standardize historical meteorological data of new energy stations.

8. A new energy extreme weather data prediction device, characterized in that: include: Acquisition module, used to obtain historical meteorological data of new energy stations; A construction module is used to construct an assimilation model based on the generative diffusion model and optimize the assimilation model using the constructed gradient; A generation module, used for generating a plurality of training weather initial fields based on the assimilation model; A training module, used for training a meteorological model based on an initial training weather field, wherein the meteorological model is iteratively optimized using a flat tropospheric loss function; The obtaining module is used to obtain the weather forecast result according to the input initial weather field by using the meteorological model.

9. A server, characterized in that: The server comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the new energy extreme weather data prediction method as described in any one of claims 1-7.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the new energy extreme weather data prediction method as described in any one of claims 1-8.

Images