A wind resource dynamic downscaling method based on a global stretched grid pattern

Abstract

The application relates to the technical field of new energy resource simulation, and discloses a wind resource dynamic downscaling method based on a global stretching grid mode, which introduces grid stretching transformation in an initial global mode framework adopting an original cubic spherical uniform grid, realizes gradual grid encryption of spatial resolution in an arbitrary target region while keeping smoothness and physical consistency in a numerical calculation process, and realizes seamless transition from large-scale driving data to a local high-precision wind resource field while retaining large-scale meteorological information. Meanwhile, the short-time strong gust and extreme wind speed event simulation capability is improved by embedding a theoretical gust wind speed calculation module in the global stretching grid mode. Further, the target global stretching grid mode is used for downscaling processing of wind resources of a target region to be downscaled, so that seamless transition from large scale to a local high-precision wind resource field is realized.

Description

Technical Field

[0001] This invention relates to the field of new energy resource simulation technology, specifically to a wind resource dynamics downscaling method based on a global stretched grid model. Background Technology

[0002] To address climate change and promote energy structure transformation, the development of renewable energy sources such as wind and solar power has accelerated globally. As a direct energy source for new energy power generation, high-resolution, refined simulation and forecasting of wind resources has become a crucial link in improving power generation efficiency and ensuring the stable operation of the power system. Downscaling techniques, as an important means of obtaining high spatial resolution wind resource information, play a core role in the field of wind resource simulation and forecasting. Generally speaking, downscaling methods can be mainly divided into statistical downscaling methods, dynamic downscaling methods, and dynamic-statistical downscaling methods.

[0003] (1) Statistical downscaling method: Based on the statistical relationship between historical observation data and the output of large-scale climate models, local climate prediction is carried out by establishing regression models or machine learning algorithms. The model is simple, the calculation speed is fast and the efficiency is high. However, this method relies on a large amount of historical data to capture the correlation between climate variables. It is constrained by historical statistical relationships, cannot describe the patterns that do not appear in historical data, and has limited ability to simulate extreme weather events.

[0004] (2) Dynamic downscaling method: This method utilizes the output of global climate models to provide initial predictive boundary conditions for regional climate models (usually WRF models). It uses physical equations to simulate the climate system, obtaining results with higher spatial resolution for the region. By constraining based on physical and mathematical equations, it better reflects the interaction between local variability and the large-scale background field, making it more suitable for simulating extreme weather and complex terrain. However, it is computationally expensive; the computational cost increases exponentially with increasing model resolution. Furthermore, the systematic errors between global and regional climate models amplify the uncertainty of the downscaling results.

[0005] (3) Dynamic-Statistical Downscaling Method: There are two main implementation paths. One is to obtain a high-resolution regional climate field by driving RCM through GCM, and then combine it with observations to construct a statistical model to further extrapolate station-scale information, which is often used for climate change impact assessment. The other method directly establishes a statistical mapping relationship between large-scale and local variables between GCM-RCM outputs and applies it to other GCM outputs to achieve efficient downscaling without repeating RCM simulations, which is suitable for multi-model and multi-scenario predictions. Although it combines the physical consistency of dynamic downscaling with the computational efficiency of statistical methods, it is essentially a concatenation of the two methods, and there are limitations in physical consistency and error propagation handling. At the same time, this method depends on the consistency of outputs between different GCMs. If the differences between models are large, the generalization ability and prediction stability of this method may decrease. Summary of the Invention

[0006] In view of this, the present invention provides a wind resource dynamic downscaling method based on a global stretched grid model to solve the limitations of traditional statistical downscaling methods, dynamic downscaling methods, and dynamic-statistical downscaling methods.

[0007] In a first aspect, the present invention provides a wind resource dynamics downscaling method based on a global stretched grid model, the method comprising:

[0008] An initial global model framework is obtained, which adopts the original cubic spherical uniform grid. In the initial global model framework, the original cubic spherical uniform grid is stretched using the Schmidt transform method to generate an initial global stretched grid model. A theoretical gust calculation model is added to the diagnostic calculation module of the initial global stretched grid model to generate a target global stretched grid model. The wind resources of the target area to be downscaled are downscaled using the target global stretched grid model to obtain a high-resolution wind resource information set.

[0009] This invention provides a wind resource dynamics downscaling method based on a global stretched grid model. Within an initial global model framework using an original cubic spherical uniform grid, a Schmidt transform is employed to stretch the original grid and generate an initial global stretched grid model. This achieves progressive refinement of the spatial resolution of the target region, overcoming the boundary discontinuities inherent in traditional nested dynamics downscaling. Furthermore, compared to traditional regional models, it eliminates the need for manually setting nested boundaries, preserving large-scale global climate background information while enhancing local small-scale simulation capabilities, thus resolving the difficulty in continuously describing the interaction between large-scale circulation and local processes. Further, a theoretical gust calculation model is added to the diagnostic calculation module of the initial global stretched grid model to generate the target global stretched grid model, enabling direct output of gust wind speeds. This enhances the model's ability to characterize extreme wind events, providing crucial data for wind farm safe operation and extreme wind speed early warning, addressing the problem that traditional models cannot meet the engineering requirements for extreme wind speed information. Simultaneously, the coupling process maintains the overall physical consistency of the model, avoiding the introduction of errors caused by separate calculations of gusts in later stages. This allows the target global stretched grid model to not only output average wind speeds but also reflect extreme gust characteristics through an embedded target gust calculation model, solving the problem that traditional numerical models struggle to characterize extreme wind events and enhancing its adaptability to wind resource characteristics under complex terrain. Furthermore, by defining the target area to be scaled down, computational resources are concentrated on the area of ​​demand. Moreover, by using the wind resources of the target area to be scaled down from the target global stretched grid model for downscaling, a seamless transition from large-scale to local high-precision wind resource fields is achieved.

[0010] In one alternative implementation, obtaining an initial global model framework includes:

[0011] Obtain a preset non-hydrostatic global numerical model; configure a time integration scheme based on horizontal explicit and vertical semi-implicit methods in the dynamic framework of the preset non-hydrostatic global numerical model to obtain the global numerical model; use a preset physics scheme package to configure the physical process parameters in the global numerical model to obtain the initial global model framework.

[0012] This invention provides a wind resource dynamics downscaling method based on a global stretched grid model. By pre-setting a finite-volume cubic spherical grid framework for a non-hydrostatic global numerical model, it avoids the polar convergence and singularity problems of traditional latitude and longitude grids, ensuring the numerical stability and parallel scalability of global simulations and solving the computational efficiency bottleneck in high-resolution simulations. Furthermore, by configuring a time integration scheme based on horizontal explicit and vertical semi-implicit methods, it can balance the computational efficiency of horizontal small-scale explicit methods with the stability of vertical acoustic implicit constraints, maintaining a reasonable time step even in high-resolution regions. This solves the problem of extremely small time steps and drastically increased computational costs caused by rapid vertical perturbations in traditional explicit schemes. Furthermore, it utilizes a pre-set physics scheme package to achieve modular invocation of physical process parameters. Simultaneously, the same physics scheme is used globally and locally, ensuring physical consistency and solving the error problems caused by chaotic physics schemes and inconsistent parameterization in different regions in traditional models.

[0013] In one alternative implementation, within the initial global model framework, the original cubic spherical uniform mesh is stretched using the Schmidt transform method to generate an initial global stretched mesh model, including:

[0014] In the initial global model framework, the original cubic spherical uniform mesh is analytically deformed based on the preset mesh stretching coefficient and the Schmidt transform method to obtain the target stretched mesh. The six-sided topology of the target stretched mesh is consistent with the six-sided topology of the original cubic spherical uniform mesh. Based on the target stretched mesh, the initial global stretched mesh model is generated through a preset physical process parameterization scheme, which has resolution-aware characteristics.

[0015] The wind resource dynamics downscaling method based on a global stretched grid model provided by this invention

[0016] In one optional implementation, within the initial global model framework, the original cubic spherical uniform mesh is analytically deformed based on a preset mesh stretching factor and a Schmidt transform method to obtain the target stretched mesh, including:

[0017] The grid coordinate transformation coefficient is calculated using the grid stretching coefficient. In the initial global model framework, the latitude of the original cubic spherical uniform grid is stretched according to the grid coordinate transformation coefficient to obtain multiple stretched grid latitudes. The multiple original grid latitudes of the densification region corresponding to the original cubic spherical uniform grid are rotated to multiple stretched grid latitudes using rigid body rotation to obtain the target stretched grid. The densification region is the area near Antarctica in the original cubic spherical uniform grid that is concentrated due to the transformation of the original grid latitude.

[0018] This invention provides a wind resource dynamics downscaling method based on a global stretched grid model. Based on the stretching coefficient and Schmidt transform, it achieves directional refinement of the target region with smooth grid transitions, maintaining global grid coherence and solving the problem of physical process discontinuity caused by abrupt resolution changes in traditional nested models. Simultaneously, compared to traditional regional models, it can improve the accuracy of the target region while preserving the basic characteristics of other global regions, avoiding the loss of large-scale information. Furthermore, it only transforms the latitude direction, keeping the longitude unchanged, simplifying calculations while ensuring grid topology stability, avoiding numerical disturbances that may be caused by full coordinate transformations, and guaranteeing the stability of model calculations. Furthermore, by moving the South Pole to the center of the target region through rigid body rotation, it achieves resolution refinement in any region, allowing the model to be flexibly applied to any target region globally, overcoming the limitation of traditional models that can only perform nested simulations in fixed regions. Finally, the resulting target stretched grid has the highest resolution in the central region, gradually transitioning outwards, balancing accuracy and efficiency, and providing an optimized grid foundation for subsequent terrain matching and simulation operations. Therefore, by implementing this invention, a progressive densification of the spatial resolution of the target area is achieved, enabling the grid to transition continuously from the center outwards, maintaining the smoothness and physical consistency of differential calculations, balancing global coverage and local accuracy, improving the simulation accuracy of key areas, and preserving large-scale information features.

[0019] In one alternative implementation, the method further includes:

[0020] The initial gust calculation model is obtained based on the preset gust definition; the initial gust factor is determined based on the initial gust calculation model, peak value coefficient, wind speed standard deviation, and turbulence intensity index; the target gust factor is determined based on the initial gust factor and the influence factors of terrain height and surface roughness; the theoretical gust calculation model is determined based on the target gust factor, average wind speed, and the initial gust calculation model.

[0021] This invention provides a wind resource dynamics downscaling method based on a global stretched grid model. By pre-defining gust definitions to obtain an initial gust calculation model, wind calculations have a clear standard basis, solving the problems of vague gust definitions and lack of unified calculation standards in traditional models. Furthermore, by combining peak power coefficient, wind speed standard deviation, and turbulence intensity index, the relationship between gusts and average wind speed can be scientifically quantified, thus transforming gust estimation from empirical to theoretical, improving the accuracy of gust calculations and overcoming the rough estimation defects of traditional methods. Furthermore, by introducing the influence factors of terrain height and surface roughness, gust calculations are adapted to complex underlying surface characteristics, solving the problem that traditional gust models ignore terrain and surface attributes and have large errors in complex terrain areas. Finally, the theoretical gust calculation model determined based on the target gust factor, average wind speed, and the initial gust calculation model can more accurately reflect actual gust characteristics, providing a reliable reference for extreme wind speed design in wind farms and enhancing the engineering practicality of wind resource information.

[0022] In one optional implementation, the wind resources of the target region to be downscaled are downscaled using a target global stretched grid model to obtain a high-resolution wind resource information set, including:

[0023] Acquire the initial high-resolution terrain dataset and the preset driving dataset; using the center latitude and longitude of the target region as the center of the grid stretching, perform a stretching transformation on the original cubic spherical uniform grid using the Schmidt transform method to obtain the target stretched grid of the target region; process the initial high-resolution terrain dataset to obtain the target basic terrain dataset; based on the preset driving dataset, the target stretched grid, and the target basic terrain dataset, perform a downscaling operation on the target global stretched grid model to obtain the downscaled high-resolution wind resource information set of the target region.

[0024] This invention provides a wind resource dynamics downscaling method based on a global stretched grid model. By identifying the target region to be downscaled, it provides a targeted approach for subsequent grid stretching and refined simulations, solving the problem that traditional global models have uniform resolution and cannot efficiently improve accuracy for specific regions. This allows computational resources to be concentrated on the required region. Furthermore, by using a Schmidt transform to obtain the target stretched grid, it achieves progressive refinement of the spatial resolution of the target region, overcoming the boundary discontinuity defect in traditional nested dynamics downscaling. Simultaneously, compared to traditional regional models, it eliminates the need for manually setting nested boundaries, preserving global large-scale climate background information while improving local small-scale simulation capabilities, solving the problem of the difficulty in continuously describing the interaction between large-scale circulation and local processes. Furthermore, by matching topographic data with the target stretched grid, it achieves a high degree of matching between topographic relief and the model grid, improving the simulation accuracy of complex topographic regions and avoiding numerical instability caused by mismatch between topographic data and the grid. This solves the problem of coarse topographic processing in traditional models, which cannot accurately reflect the impact of topography on the wind field. Furthermore, based on the obtained preset driving dataset, target stretched grid, and target basic terrain dataset, the target global stretched grid model is downscaled to obtain the target area downscaled wind resource information set, realizing a seamless transition from large-scale to local high-precision wind resource field.

[0025] Secondly, the present invention provides a wind resource dynamics downscaling device based on a global stretched grid model, the device comprising:

[0026] The system comprises the following modules: an acquisition module for acquiring an initial global model framework using a primitive cubic spherical uniform grid; a first generation module for stretching the primitive cubic spherical uniform grid within the initial global model framework using the Schmidt transform method to generate an initial global stretched grid model; a second generation module for adding a theoretical gust calculation model to the diagnostic calculation module of the initial global stretched grid model and generating a target global stretched grid model; and a processing module for downscaling the wind resources of the target region to be downscaled using the target global stretched grid model to obtain a high-resolution wind resource information set.

[0027] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the wind resource dynamics downscaling method based on the global stretched grid model described in the first aspect or any corresponding embodiment.

[0028] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the wind resource dynamics downscaling method based on a global stretched grid model as described in the first aspect or any corresponding embodiment thereof.

[0029] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the wind resource dynamics downscaling method based on a global stretched grid model as described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0030] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a flowchart illustrating a wind resource dynamics downscaling method based on a global stretched grid model according to an embodiment of the present invention.

[0032] Figure 2 This is a flowchart illustrating a wind resource dynamics downscaling method based on a global stretched grid model with a finite-volume cubic spherical grid as the dynamic framework, according to an embodiment of the present invention.

[0033] Figure 3 This is a schematic diagram of the global stretched mesh pattern construction process according to an embodiment of the present invention;

[0034] Figure 4 This is a schematic diagram of the construction process of the theoretical gust calculation model according to an embodiment of the present invention;

[0035] Figure 5 This is a schematic diagram of the downscaling test preparation process according to an embodiment of the present invention;

[0036] Figure 6 This is a schematic diagram of the downscaling test operation process according to an embodiment of the present invention;

[0037] Figure 7 This is a schematic diagram of downscaling technique evaluation according to an embodiment of the present invention;

[0038] Figure 8 This is a structural block diagram of a wind resource dynamics downscaling device based on a global stretched grid model according to an embodiment of the present invention;

[0039] Figure 9 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] As a crucial component of renewable energy, wind power is gradually transitioning from a supplementary energy source to a primary energy source. However, wind energy resources are characterized by uneven spatial distribution and strong spatiotemporal variations, posing significant technical challenges to the site selection, development, construction, and subsequent operation and maintenance of wind power projects. Most existing climate models and reanalysis data typically have a spatial resolution of tens to hundreds of kilometers, making it difficult to analyze regional-scale topographic relief, land-sea differences, and boundary layer processes. This is particularly true in complex underlying surface areas such as mountains, hills, and coastal zones, where the ability to simulate local wind speed variations, abrupt changes, and extreme events is significantly insufficient, failing to directly meet the engineering application requirements for refined wind resource assessment and multi-scale forecasting. Therefore, from finely characterizing wind resource distribution at the macroscopic scale to accurately forecasting wind speed trends at the medium and short-term scales, higher demands are placed on wind resource simulation and prediction technologies. Against this backdrop, wind resource simulation and prediction technologies are continuously developing towards higher resolution and greater precision to better support key aspects of the entire wind power lifecycle, including wind farm site selection, wind resource classification, wind turbine layout optimization, power generation prediction, and grid load dispatching. To extract more locally representative and engineering-applicable wind resource information from coarse-resolution climate models or reanalysis data, downscaling techniques have become indispensable. Broadly speaking, downscaling methods can be categorized into statistical downscaling, dynamic downscaling, and dynamic-statistical downscaling.

[0042] Statistical downscaling methods primarily infer regional-scale climate characteristics by constructing statistical mapping relationships between large-scale climate variables and local meteorological elements. These methods offer advantages such as high computational efficiency and flexibility. Commonly used methods include: linear regression models (establishing a linear relationship between large-scale and local climate variables using linear regression), multiple regression analysis (extending linear regression models to consider the impact of multiple large-scale climate variables on local climate), quantile regression (for extreme climate events, quantile regression models can better describe the distribution characteristics of climate variables), canonical correlation analysis, and statistical models constructed using principal component analysis, empirical orthogonal functions, singular value decomposition, and other statistical methods. Although statistical downscaling methods are widely used in practice, they heavily rely on historical sample data and are difficult to adapt to the current climate state under the background of drastic climate change, especially in handling sudden, nonlinear, and extreme weather events, where they have certain limitations.

[0043] Furthermore, dynamic downscaling methods typically utilize regional climate models, solving mathematical and physical equations within the numerical model to analyze the state of the climate system. Common regional climate models include WRF, RegCM, and COSMO models. These methods usually use the outputs of global climate models (GCM) or reanalysis data as initial and boundary conditions, driving regional climate models (RCM) to perform high-resolution simulations within a specific target region. This allows for the analysis of the modulating effect of complex underlying surface heterogeneity on local climate. This method can effectively capture local climate characteristics and small-scale meteorological processes, significantly improving the ability to characterize the spatial distribution of local wind resources, small-scale disturbances, and extreme events. The advantage of dynamic downscaling is that its simulation process is constrained by physical equations, does not rely on historical statistical relationships, and possesses stronger extrapolation capabilities and the potential to adapt to future climate scenario changes. However, dynamic downscaling still has several limitations. On the one hand, existing dynamic downscaling methods typically employ a nested structure of GCM-driven RCM, which is a non-integrated system with insufficient physical consistency. Furthermore, the boundary settings of regional models are somewhat subjective, and due to limited spatial coverage, they struggle to fully respond to and reflect global climate change. This results in an inability to continuously describe the interaction between large-scale circulation and local processes, easily leading to the transmission, accumulation, and amplification of systematic biases during the nesting process. On the other hand, high-resolution simulations require enormous computational resources. As spatial resolution increases, the computational load grows exponentially, further limiting their efficiency and operability in practical wind energy resource forecasting and operational applications. While dynamic downscaling methods theoretically possess the advantage of physical consistency, their simulation performance and computational efficiency still heavily depend on the structural design and parameterization scheme configuration strategies of the numerical model itself. Firstly, different numerical models differ in dynamic cores, grid structures, and time integration schemes, directly affecting their analytical capabilities in complex terrain and boundary layer environments. Secondly, even within the same model framework, the type, combination, and parameter settings of physical parameterization schemes significantly influence the simulation results for wind speed structure, small-scale disturbances, and extreme wind events. Therefore, when developing based on numerical models, the scientific selection and targeted configuration of physical schemes are key technical issues in constructing dynamic downscaling systems, and also important factors in determining model adaptability and simulation performance.

[0044] Furthermore, the dynamical-statistical downscaling method combines the advantages of both dynamical and statistical downscaling, typically employing two implementation paths. One common approach is to first use the output of the regional climate model (GCM) to drive the regional climate model (RCM), initially obtaining high-resolution climate variables at the regional scale. Then, based on the historical statistical relationships between these outputs and observational data, statistical methods (the aforementioned statistical models or artificial intelligence models) are used to further acquire local climate information at the station scale or smaller scales. This method can, to a certain extent, balance the physical consistency of dynamical downscaling with the high efficiency of statistical methods and is widely used in climate change impact assessments. Another approach establishes statistical mapping relationships between large-scale and local-scale climate variables between multiple GCMs and their corresponding RCM outputs, and applies this statistical model to the outputs of other GCMs, thus achieving efficient downscaling without requiring further RCM simulations. This method significantly saves computational resources and has strong scalability, making it particularly suitable for multi-model, multi-scenario climate prediction analysis. However, the effectiveness of this method is highly dependent on the consistency and representativeness of the outputs from different GCMs. Due to the significant system differences among GCMs, if the statistical relationships obtained based on a certain set of GCM-RCMs are directly applied to other GCMs, their generalization performance may be weakened, thereby affecting the accuracy and stability of the prediction results.

[0045] According to an embodiment of the present invention, a wind resource dynamics downscaling method based on a global stretched grid model is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0046] This embodiment provides a wind resource dynamics downscaling method based on a global stretched grid model, which can be used in electronic devices such as computers, mobile phones, and tablets. Figure 1 This is a flowchart of a wind resource dynamics downscaling method based on a global stretched grid model according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps:

[0047] Step S101: Obtain the initial global model framework.

[0048] The initial global model framework adopts the UFS global model developed by NOAA for debugging and application. It uses a dynamic framework based on a finite volume cubic spherical mesh. This dynamic framework approximates the three-dimensional compressible non-hydrostatic atmospheric equations on the mesh using the finite volume method. At the same time, by dividing the sphere into six faces, the topological structure of the six-faced cubic sphere ensures the continuity and scalability of the global mesh. This mesh avoids the convergence and singularity problems of traditional latitude and longitude meshes in polar regions and supports large-scale parallel computing, providing a benchmark global model framework for the subsequent construction of a global stretched mesh model.

[0049] Furthermore, the original cubic spherical uniform grid serves as the model's basic grid architecture, with its initial resolution determined by the number of grid points (each face has a uniform grid distribution; for example, a C100 grid represents 100*100 grid points per face, resulting in an initial resolution of approximately 100 km). In the model integration, each sphere is calculated independently, and boundary physical quantity information such as momentum, heat, and water vapor fluxes is exchanged at the boundaries between spheres, providing a baseline grid for subsequent regional resolution refinement via Schmitt transform.

[0050] Step S102: In the initial global model framework, the original cubic spherical uniform mesh is stretched using the Schmidt transform method to generate the initial global stretched mesh model.

[0051] The Schmidt transform method represents a technique that analytically deforms the original cubic spherical uniform grid by introducing a grid stretching coefficient and a grid coordinate mapping method to the grid and the original latitude and longitude grid points. By determining the stretching coefficient and according to the grid coordinate mapping formula, the latitude and longitude coordinates of the original cubic spherical grid are mapped to the South Pole. Only the latitude direction of the original grid needs to be transformed, while the longitude remains unchanged. Then, the entire latitude and longitude coordinate system is moved to the center point of the target region's latitude and longitude coordinates through rigid body rotation, thereby forming a global stretched grid model with regional spatial resolution densification.

[0052] Specifically, by introducing the concept of Schmidt transformation on the basis of the global model framework, the original cubic spherical uniform grid is analytically deformed without changing the six-sided topology of the original spherical grid. The global grid is gradually densified in the target region and correspondingly sparsed in the opposite region, which can then further generate the corresponding initial global stretched grid model.

[0053] Step S103: Add a theoretical gust calculation model to the diagnostic calculation module of the initial global stretched mesh model and generate the target global stretched mesh model.

[0054] Specifically, the theoretical gust calculation model is used to simulate and calculate instantaneous gust speeds by referring to the WMO (World Meteorological Organization) definition of gusts (the maximum instantaneous wind speed with a 3-second moving average within a certain time window T), based on parameters such as horizontal wind speed and surface roughness output by the global stretched grid model at a finite time resolution, and with the help of theoretical and empirical formulas.

[0055] Furthermore, by adding a theoretical gust calculation module to the diagnostic calculation module of the initial global stretched grid model, the shortcomings of traditional numerical models in the time simulation of short-term strong winds and extreme wind speeds are made up, and the model's ability to output extreme characteristics of wind resources is enhanced.

[0056] Step S104: Use the target global stretched grid model to downscale the wind resources of the target area to be downscaled, and obtain a high-resolution wind resource information set.

[0057] Specifically, based on research needs, the original global uniform grid of the target global stretched grid pattern that matches the preset driving data spatial resolution can be obtained to determine the target region to be scaled down.

[0058] Furthermore, the target global stretched grid model can be invoked and run on a supercomputing platform. In addition, the target global stretched grid model automatically solves atmospheric dynamic equations and physical processes during operation, and simultaneously runs the embedded target gust calculation model.

[0059] Furthermore, during the operation of the target global stretched grid model, wind resource information of the target area is calculated and stored in real time, including three-dimensional wind vectors (zonal wind, meridional wind), wind speed, wind direction, wind power density, and instantaneous maximum gust wind speed generated by the target gust calculation model. Finally, a downscaled wind resource information set is formed, realizing a seamless transition from large-scale driving data to local high-precision wind resource field.

[0060] This embodiment provides a wind resource dynamics downscaling method based on a global stretched grid model. Using an initial global model framework with an original cubic spherical uniform grid, a Schmidt transform is employed to stretch the original cubic spherical uniform grid and generate an initial global stretched grid model. This achieves progressive refinement of the spatial resolution of the target region, overcoming the boundary discontinuities inherent in traditional nested dynamics downscaling. Furthermore, compared to traditional regional models, it eliminates the need for manually setting nested boundaries, preserving large-scale global climate background information while enhancing local small-scale simulation capabilities, thus resolving the difficulty in continuously describing the interaction between large-scale circulation and local processes. Further, a theoretical gust calculation model is added to the diagnostic calculation module of the initial global stretched grid model to generate the target global stretched grid model, enabling direct output of gust wind speeds. This enhances the model's ability to characterize extreme wind events, providing crucial data for wind farm safe operation and extreme wind speed early warning, addressing the problem that traditional models cannot meet the engineering requirements for extreme wind speed information. Simultaneously, the coupling process maintains the overall physical consistency of the model, avoiding the introduction of errors caused by separate calculations of gusts in later stages. This allows the target global stretched grid model to not only output average wind speeds but also reflect extreme gust characteristics through an embedded target gust calculation model, solving the problem that traditional numerical models struggle to characterize extreme wind events and enhancing its adaptability to wind resource characteristics under complex terrain. Furthermore, by defining the target area to be scaled down, computational resources are concentrated on the area of ​​demand. Moreover, by using the wind resources of the target area to be scaled down from the target global stretched grid model for downscaling, a seamless transition from large-scale to local high-precision wind resource fields is achieved.

[0061] In some optional implementations, step S101 above includes:

[0062] Step S1011: Obtain the preset non-hydrostatic global numerical model.

[0063] The preset non-hydrostatic global numerical model adopts a dynamic framework of finite volume cubic spherical grid (FV3). This dynamic framework approximates the three-dimensional compressible non-hydrostatic atmospheric equations on the cubic spherical grid using the finite volume method. At the same time, by dividing the sphere into six faces (each face is calculated separately, and boundary physical quantity information such as momentum, heat, and water vapor flux variables are exchanged between the faces), the topological structure of the six cubic spheres ensures the continuity and scalability of the global grid. This grid avoids the convergence and singularity problems of traditional latitude and longitude grids in polar regions, ensuring the numerical stability and efficient parallel scalability of the global simulation.

[0064] Step S1012: Configure a time integration scheme based on horizontal explicit and vertical semi-implicit methods in the preset non-static global numerical model dynamic framework to obtain the global numerical model.

[0065] Among them, the horizontal explicit and vertical semi-implicit time integration scheme represents a time discretization method used in atmospheric numerical models to solve atmospheric dynamic equations. Its core is to employ different time discretization methods to address the differences in atmospheric motion characteristics in the horizontal and vertical directions. The propagation speeds of physical processes differ significantly in different directions. For example, the propagation speed of sound waves in the vertical direction is much faster than in the horizontal direction. If an explicit algorithm is used uniformly for all directions, the rapid perturbations in the vertical direction will result in an extremely small time step for the entire simulation, thus greatly increasing computational costs.

[0066] Specifically, numerical models essentially simulate and predict weather processes by solving mathematical equations describing atmospheric motion. These equations typically exist in the form of spatiotemporal partial differential equations, therefore, they must be discretized spatially and temporally for solving on a computer during numerical simulations. Thus, to improve the model's stability and computational efficiency in high-resolution simulation regions, a horizontal explicit and vertical semi-implicit time integration scheme is configured for the dynamic framework of the pre-defined non-hydrostatic global numerical model. This balances the computational efficiency of horizontal small-scale explicit integration with the stability of vertical acoustic implicit constraints, thereby maintaining a reasonable time step even in high-resolution regions.

[0067] Furthermore, by processing acoustic waves in a semi-implicit manner vertically, the time step constraint can be alleviated, and the stability of numerical calculation can be improved; by processing small-scale perturbations explicitly horizontally, the spatial resolution can be improved.

[0068] Furthermore, based on the horizontal explicit and vertical semi-implicit time integration schemes configured in the solver of the preset non-hydrostatic global numerical model, when the user sets parameters such as time step, grid points / resolution in the model configuration file, the preset non-hydrostatic global numerical model will automatically activate its internal time integration framework for numerical calculation after startup.

[0069] Furthermore, by configuring a time integration scheme based on horizontal explicit and vertical semi-implicit methods, it is possible to balance the computational efficiency of horizontal small-scale explicit methods with the stability of vertical acoustic implicit constraints. This allows for maintaining a reasonable time step in high-resolution regions, thus solving the problem of extremely small time steps and drastically increased computational costs caused by rapid vertical perturbations in traditional explicit schemes.

[0070] Step S1013: Using a preset physics scheme package, configure the physical process parameters in the global numerical model to obtain the initial global model framework.

[0071] The preset physics scheme package represents a standardized and modular set of multiple physical process parameterization schemes provided for numerical models. These schemes are used for unified invocation, configuration, and accurate simulation of atmospheric physical processes within the numerical model. This embodiment employs the CCPP (Common Community Physics Package) physics scheme package, aiming to achieve standardization, modularization, and cross-model sharing of physics schemes.

[0072] Furthermore, the CCPP physics scheme package is a physics scheme package that integrates most parameterization schemes (including cloud microphysics, convection, radiation, boundary layer, land surface processes, etc.). Researchers can directly call it in the model according to the study area and simulation requirements. Specifically, when building downscaled cases, the CCPP_SUITES switch and the CCPP version to be called are set, and the selected parameterization scheme and the parameters in the parameterization scheme are set in configuration files such as namelist.

[0073] Specifically, by introducing a pre-defined physical scheme package, modular invocation of physical process parameters can be achieved. Furthermore, within this framework, the same physical scheme is used both globally and locally, ensuring consistency in physical processes.

[0074] Furthermore, the CCPP currently provides multiple physical process parameterization schemes, which can be automatically invoked in the first global numerical model. Combining the physical mechanisms of high-resolution wind resource simulation and the need for adaptability to complex terrain, this invention selects a representative combination of parameterization schemes, wherein: the current model uses the GFDL cloud microphysics parameterization scheme for cloud microphysics parameterization; the GFS sa-TKE-EDMF boundary layer scheme for boundary layer parameterization; the GFS sa-SAS deep convection parameterization scheme and the sa-MF shallow convection parameterization scheme for convection parameterization; the RRTMG shortwave / longwave radiation parameterization scheme for radiative transfer parameterization; the Noah land surface process scheme for land surface parameterization; and the GFS surface layer parameterization scheme for near-surface layer parameterization.

[0075] Furthermore, to improve the performance of wind field simulations in complex terrain areas, the Unified Gravity Wave physics scheme is used. This scheme handles gravity wave processes excited by sub-grid-scale terrain forcing that cannot be directly resolved by the model resolution, employing terrain-induced gravity wave drag parameterization at the sub-grid scale. In actual simulations, when the model resolution cannot directly resolve gravity waves excited by small-scale terrain forcing such as mountains and canyons, these unresolved processes are explicitly expressed through physical parameterization. The Unified Gravity Wave scheme estimates the momentum dissipation (dragging) effect of these sub-grid-scale gravity waves on the background wind field during their propagation in the troposphere and stratosphere, particularly impacting the simulation of near-surface wind speed structure and strong wind events. In practical applications, this parameterization scheme can be invoked in the model namelist configuration, and parameters such as drag coefficient and drag height can be set according to research needs.

[0076] Furthermore, modular invocation of physical process parameters was achieved by using a pre-defined physical scheme package. At the same time, the same set of physical scheme combinations was used globally and locally, ensuring physical consistency, reducing cross-scale discontinuities caused by differences in parameterization schemes, and mitigating the impact of global model error propagation to regional models to some extent.

[0077] In some optional implementations, step S102 above includes:

[0078] Step S1021: In the initial global model framework, based on the preset mesh stretching coefficient and the Schmidt transform method, the original cubic spherical uniform mesh is analytically deformed to obtain the target stretched mesh.

[0079] The six-sided topology of the target stretched mesh is consistent with the six-sided topology of the original cubic spherical uniform mesh.

[0080] To achieve high-resolution local simulations on a global model framework, the Schmidt transform concept is introduced into the original cubic spherical uniform grid in the initial global model framework to perform analytical deformation on the original uniform grid. Without changing the six-sided topology of the original spherical grid, the global uniform grid is gradually densified in the target region and correspondingly sparsed in the opposite region.

[0081] Specifically, step S1021 includes:

[0082] Step a1: Calculate the mesh coordinate transformation coefficient using the mesh stretching coefficient.

[0083] Step a1: In the initial global model framework, the latitude of the original cubic spherical uniform grid is stretched according to the grid coordinate transformation coefficient to obtain multiple stretched grid latitudes.

[0084] Step a3: Using rigid body rotation, rotate multiple original grid dimensions of the encrypted region corresponding to the original cubic spherical uniform grid to the multiple stretched grid dimensions to obtain the target stretched grid.

[0085] The encrypted region is the area near Antarctica in the original cubic spherical uniform grid that is concentrated due to the latitude transformation of the original grid.

[0086] Specifically, by comparing the spatial resolution of the original driving data with the target output resolution, the corresponding stretching factor, i.e., the mesh stretching coefficient s, can be calculated.

[0087] Furthermore, the mesh coordinate transformation coefficient can be calculated using the selected mesh stretching factor s, as shown in the following relationship (1):

[0088]

[0089] In the formula: CF represents the grid-converting factor; s represents the grid stretching factor, which mainly controls the strength and proportion of grid stretching (s>1).

[0090] Furthermore, the original latitude can be... Mapped to the stretching mesh latitude φ, as shown in equation (2):

[0091]

[0092] In the formula: φ represents the converted latitude; This represents the original grid latitude.

[0093] Furthermore, by mapping the latitude and longitude of the original cubic sphere grid to the South Pole, only the latitude direction of the original grid needs to be transformed (according to the above relationship (2)), while the longitude remains unchanged. Then, the entire latitude and longitude coordinate system is rotated in a rigid-body manner to move the South Pole to the latitude and longitude coordinates of the center point of the target area.

[0094] After the above steps, any original cubic spherical uniform grid can be converted into a grid with a spacing of approximately 1 / s in the target area.

[0095] For example, taking the original cubic spherical uniform grid C100 (each face has 100*100 grid points) as an example, the original resolution is approximately 100km. For a selected target area, taking the stretching factor s = 4, the average resolution of the stretched target surface (the sphere where the spatial densification area is located) is approximately... The minimum resolution of the target surface is approximately The coarsest resolution is approximately On the opposite side of the Earth, the grid spacing is magnified to approximately s times the original (about 400 km). The stretched grid produced by this transformation has the highest resolution at the center of the target and transitions smoothly and continuously outward to the coarse resolution of the background without abrupt boundary changes.

[0096] The resolution can be calculated using the following formula (3):

[0097]

[0098] In the formula: The average resolution (km) of the target surface is represented by N; N represents the number of grid points per sphere.

[0099] The stretching process described above maintains the consistency of the global grid without requiring the addition of new grid surfaces or manual specification of nested boundaries, while achieving a smooth transition from coarse to high resolution. As the stretching coefficient *s* changes, the spatial resolution and refinement range of the target area can be flexibly adjusted: the larger *s*, the more concentrated and finer the high-resolution area, while the surrounding coarse-resolution area expands accordingly. Therefore, this stretching process can precisely capture the impact of complex terrain and small-scale turbulence on the wind field while preserving global-scale climate background information, making it an effective means of obtaining refined wind resource information.

[0100] Furthermore, the aforementioned stretching mesh transformation method is incorporated into the initial global model framework, enabling the model to run directly on non-uniform resolution meshes under a unified global numerical model architecture. The resulting initial global stretching mesh model not only retains the dynamic framework, physical process parameterization, and time integration scheme of the initial global model framework, but also enables numerical integration on the new global stretching mesh during runtime, achieving high-resolution simulation of the target region while maintaining relatively low-resolution computation in non-target regions.

[0101] Furthermore, in this model system, local encrypted simulation no longer relies on independent nested regional models, but is performed as part of the global model for numerical calculation. This breaks through the limitations of the traditional "global model + regional model" nesting, and realizes the unified simulation of global background and local details under a single framework, effectively ensuring the continuity and consistency of large-scale circulation and local processes.

[0102] Step S1022: Based on the target stretched mesh, an initial global stretched mesh model is generated through a preset physical process parameterization scheme.

[0103] Among them, the preset physical process parameterization scheme has resolution-aware characteristics.

[0104] Specifically, in the physical process parameterization scheme, a parameterization method with resolution-aware characteristics is adopted for meshes of different horizontal resolutions, so that the parameterization scheme can be adaptively adjusted with the change of mesh scale, thereby ensuring the continuity of physical processes in the transition zone between coarse and fine meshes, and thus generating the corresponding initial global stretched mesh model.

[0105] For example, within a global-stretched grid model framework, when the resolution of a locally refined region approaches the analytical scale of convection, a resolution-aware convection scheme can be used to gradually weaken or even disable deep convection parameterization; while in regions with coarser resolution, convection parameterization is retained. In this way, consistent physical process descriptions can be achieved globally and locally within the same model framework, and the physical plausibility and numerical stability of simulation results can be maintained in the coarse-to-fine grid transition region.

[0106] Furthermore, the above process does not require adding new grid surfaces or manually setting nested boundaries. It preserves the continuity of the global grid and large-scale climate background information, while also improving the spatial resolution of the target area in a targeted manner. This solves the problems of boundary discontinuity and insufficient physical consistency in traditional nested downscaling.

[0107] In some optional implementations, the theoretical gust calculation model in step S103 above is obtained through the following steps:

[0108] Step b1: Obtain the initial gust calculation model according to the preset gust definition.

[0109] Step b2: Determine the initial gust factor based on the initial gust calculation model, peak value coefficient, wind speed standard deviation, and turbulence intensity index.

[0110] Step b3: Determine the target gust factor based on the initial gust factor and the influence factors of terrain height and surface roughness.

[0111] Step b4: Determine the theoretical gust calculation model based on the target gust factor and the initial gust calculation model.

[0112] Specifically, traditional numerical models typically output wind field information in multiple layers of three-dimensional wind vector information (including zonal wind, meridional wind, and wind speed) and near-surface wind speed products, based on vertical layer resolution. However, the impact of extreme gust events cannot be ignored during the planning, design, and operation phases of wind farms and photovoltaic power plants, especially regarding the structural safety and operational stability of the units.

[0113] In this embodiment, referring to the WMO definition of gusts, the gust is defined as the maximum instantaneous wind speed with a 3-second moving average within a certain time window T (T is generally taken as 1 hour in numerical models).

[0114] Furthermore, ideally, gust characteristics should be obtained based on high-frequency wind speed time series. However, due to the limited output frequency of traditional numerical models (usually hourly), directly obtaining 3-second wind speed sequences is computationally expensive and difficult to achieve in long-term, wide-area simulations. Therefore, it is necessary to use theoretical or empirical parameterized formulas to estimate gust wind speeds based on wind speed outputs with limited time resolution to meet engineering application requirements. Thus, the initial gust calculation model can be expressed as the following relationship (4):

[0115]

[0116] In the formula: WG(t,T) represents the gust wind speed at time t within the time window T; This represents the disturbance wind speed at time t within the time window T; This indicates the average wind speed.

[0117] Furthermore, the gust speed can be viewed as a function of the gust factor (GF) and the average wind speed, that is, the initial gust calculation model can also be expressed as the following relationship (5):

[0118]

[0119] Furthermore, based on the above relationships (4) and (5), the gust factor (GF) can be obtained, as shown in the following relationship (6):

[0120]

[0121] The initial gust factor (GF) is defined as the ratio of gust speed to average wind speed.

[0122] Furthermore, it will affect the wind speed. Defined as the product of the peak coefficient and the standard deviation of wind speed during time period T, it can be expressed as the following relationship (7):

[0123]

[0124] In the formula: σ represents the peak value coefficient; T This represents the standard deviation of wind speed during time period T.

[0125] Furthermore, based on the definition of the turbulence intensity index, the initial gust factor can be expressed as the following relationship (8):

[0126]

[0127] In the formula: This represents the turbulence intensity index.

[0128] Furthermore, considering the modulating effect of terrain and surface roughness on wind speed, and referring to the Parratt model and the Wieringa empirical model, after taking terrain and surface roughness into account, the initial gust factor shown in the above equation (8) can be converted into the target gust factor shown in the following equation (9):

[0129]

[0130] In the formula: z represents height; z0 represents surface roughness. Topography and roughness affect the gust factor by altering airflow motion (e.g., a rough surface increases friction, making wind speed fluctuations more complex; topography lifts airflow, which may enhance disturbances).

[0131] Furthermore, substituting the target gust factor shown in the above relation (9) into the above relation (5) yields the corresponding theoretical gust calculation model. This method estimates instantaneous gusts based on gust factors, average wind speed, and surface properties, making up for the shortcomings of traditional numerical models in simulating short-term strong winds and extreme wind speeds. It provides a reliable reference for the design of extreme wind speeds in wind farms and enhances the engineering practicality of wind resource information.

[0132] In some optional implementations, step S104 above includes:

[0133] Step S1041: Obtain the initial basic high-resolution terrain dataset and the preset driving dataset.

[0134] The initial high-resolution terrain dataset represents SRTM (Shuttle Radar Topography Mission) data with a spatial resolution of 90m from NASA, which serves as the original input for terrain information in the target global stretched grid model simulation.

[0135] Furthermore, the preset driving dataset represents the initial and boundary condition data selected according to research needs to drive the operation of the global stretched grid model. For wind resource assessment, reanalysis data such as ERA5 can be selected; for wind resource prediction, multi-scale prediction data such as IFS, GFS, and CFS can be selected.

[0136] Furthermore, the pre-defined driving dataset provides a large-scale meteorological background field (such as air pressure, temperature of multiple pressure layers, humidity, and wind field) for the global stretched grid model. It is a key input for realizing the downscaling simulation of high-resolution wind resource fields from coarse-resolution meteorological data, ensuring the consistency between the simulation process and the actual atmospheric circulation characteristics.

[0137] Step S1042: Using the center latitude and longitude of the target area as the center of the mesh stretching, the original cubic spherical uniform mesh is stretched using the Schmidt transform method to obtain the target stretched mesh of the target area.

[0138] Based on the actual application objectives, the geographical scope and the latitude and longitude coordinates of the center of the simulation area for high-resolution wind resource simulation are determined. The mesh stretching transformation is performed using the center latitude and longitude of the target area to be scaled down as the center of the mesh stretching. The original cubic spherical mesh is smoothly densified along the latitudinal direction in the target area and correspondingly sparsed on the opposite side, resulting in a set of regional high-resolution global variable-resolution meshes, i.e., the target stretched mesh.

[0139] Specifically, based on the Schmidt transform method, the mesh coordinate transformation coefficient CF is calculated using the mesh stretching coefficient through the above relation (1).

[0140] Furthermore, using the above relation (2), the latitude of the original cubic spherical uniform grid is stretched and transformed according to the grid coordinate transformation coefficient, and the latitude and longitude of the original cubic spherical grid are mapped to the South Pole. Only the latitude direction of the original grid needs to be transformed, while the longitude remains unchanged. Then, the entire latitude and longitude coordinate system is moved to the center point latitude and longitude coordinates of the target area by rigid body rotation, and finally the target stretched grid with the highest resolution and smooth outward transition in the target area is formed.

[0141] Step S1043: Process the initial basic high-resolution terrain dataset to obtain the target basic terrain dataset.

[0142] Specifically, the initial high-resolution terrain dataset (90m spatial resolution SRTM data from NASA) is projected from the traditional latitude and longitude grid coordinate system to a pre-defined cubic spherical coordinate system. Based on the new coordinates of the target stretched grid, the projected terrain data is resampled to ensure that the grid density of the terrain data matches the grid distribution of the target stretched grid, thus preserving the high-resolution terrain details of the target area. The outer areas are simplified according to the grid sparseness. Furthermore, an interpolation algorithm is used to map the resampled terrain data to each grid point of the target stretched grid, achieving a one-to-one correspondence between terrain height values ​​and grid nodes, ensuring a high degree of consistency between the terrain undulation characteristics and the spatial structure of the target stretched grid.

[0143] Furthermore, to avoid instability in model numerical calculations caused by abrupt changes or high-frequency noise in the original terrain data, the interpolated terrain data is processed by smoothing, terrain gradient processing, and filtering. While maintaining the main terrain features, small-scale noise (such as high-frequency terrain disturbances) is removed, taking into account both the authenticity of terrain details and the stability of dynamic solutions.

[0144] Furthermore, the above processing steps generate the corresponding target basic terrain dataset.

[0145] Step S1044: Based on the preset driving dataset, the target stretched grid and the target basic terrain dataset, the target global stretched grid model is downscaled to obtain a high-resolution wind resource information set of the target area after downscaling.

[0146] Specifically, corresponding data can be selected from the preset driving dataset according to research needs, and used as the initial and boundary conditions of the target global stretched grid model to ensure that the driving data matches the spatiotemporal scale of the original cubic spherical uniform grid.

[0147] Furthermore, set the key parameters of the simulation, including the simulation start time, integration duration (e.g., for seasonal prediction, it can be set to 92 days in summer, from June to August), and output frequency (e.g., output hourly). In the model configuration file, specify the paths of the target stretched mesh and the target base terrain dataset to complete the case initialization.

[0148] Optionally, for the characteristics of wind field in the boundary layer and near-surface layer, the vertical resolution settings are densified, specifically at heights of 10m, 50-150m (intervals of 10m), 180m and 200m, to adapt to the wind resource data requirements of wind turbines and photovoltaic power plants with different hub heights and improve the accuracy of near-surface wind field simulation.

[0149] Furthermore, the target global stretched grid model is invoked on the supercomputing platform. Based on the configured preset driving dataset, target stretched grid and target basic terrain dataset, the target global stretched grid model automatically solves atmospheric dynamic equations and physical processes during operation, and simultaneously runs the embedded target gust calculation model.

[0150] Furthermore, during the operation of the model, wind resource information of the target area is calculated and stored in real time, including three-dimensional wind vectors (zonal wind, meridional wind), wind speed, wind direction, wind power density, and instantaneous maximum gust wind speed generated by the target gust calculation model. Finally, a downscaled wind resource information set is formed, realizing a seamless transition from large-scale driving data to local high-precision wind resource field.

[0151] In one example, addressing the urgent need for high-resolution, refined wind resource simulation and forecasting in the wind power industry, a wind resource dynamics downscaling method based on a global stretched grid model with a finite-volume cubic spherical grid as the dynamic framework is proposed. Building upon the traditional global uniform grid model framework, a Schmidt transform operator is introduced to generate a global variable-resolution grid with regional spatial resolution refinement capabilities. Compared to traditional WRF-based regional nested dynamics downscaling, this method avoids boundary discontinuities and error propagation caused by multi-model coupling, while also possessing a complete response to the global climate background field and the ability to finely analyze local complex terrain / microscale turbulence. Furthermore, to meet the application requirements of wind / photovoltaic power plants for safe operation and extreme wind speed early warning, a gust calculation and output module is added to the traditional numerical model output variables. Figure 2 As shown, the specific steps include:

[0152] 1. Global stretched mesh pattern construction, such as Figure 3 As shown, it includes:

[0153] Step 1: Basic global model architecture and its configuration.

[0154] The UFS global model developed by NOAA was used for debugging and application. This model is a non-hydrostatic global numerical model based on a finite volume cubic spherical grid (FV3) as its dynamic framework. This dynamic framework approximates the three-dimensional compressible non-hydrostatic atmospheric equations on the cubic spherical grid using the finite volume method. Furthermore, by dividing the sphere into six faces (each face is calculated independently, and boundary physical quantity information such as momentum, heat, and water vapor fluxes are exchanged between the facets), it avoids the convergence and singularity problems of traditional latitude and longitude grids in polar regions, ensuring the numerical stability and efficient parallel scalability of the global simulation. To improve the model's stability and computational efficiency in high-resolution simulation regions, a horizontal explicit-vertical semi-implicit time integration scheme was adopted for the dynamic framework time integration scheme. This balances the computational efficiency of horizontal small-scale explicit integration with the stability of vertical acoustic implicit constraints, thus maintaining a reasonable time step in high-resolution regions.

[0155] Among them, (1) the essence of numerical models is to simulate and predict weather processes by solving mathematical equations that describe atmospheric motion. These equations usually exist in the form of a set of spatiotemporal partial differential equations. Therefore, when performing numerical simulations, they must be discretized in space and time respectively so that they can be solved on a computer.

[0156] The "horizontal explicit-vertical semi-implicit" time integration scheme is a time discretization method for solving atmospheric dynamic equations. The propagation speeds of physical processes vary significantly in different directions. For example, sound waves propagate much faster vertically than horizontally. If an explicit algorithm is used uniformly for all directions, the rapid perturbations in the vertical direction will result in an extremely small time step for the entire simulation, greatly increasing computational costs. The "vertical semi-implicit" approach handles sound waves, alleviating time step constraints and improving numerical computation stability, while the "horizontal explicit" approach handles small-scale perturbations, improving spatial resolution.

[0157] (2) The horizontal explicit-vertical semi-implicit time integration scheme is configured in the solver of this mode. When the user sets the time step, grid points / resolution and other parameters in the mode configuration file, the mode will automatically start its internal time integration framework for calculation.

[0158] In terms of physical process parameterization, the Common Community Physics Package (CCPP) is introduced, enabling modular invocation of physical process parameterization. Under this framework, the same set of physical schemes is used globally and locally, ensuring consistency of physical processes. Currently, CCPP provides multiple physical process parameterization schemes, which can be called in the model according to the study area and simulation requirements. Among them, this invention selects a set of representative parameterization scheme combinations, namely: the current model uses the GFDL cloud microphysics parameterization scheme for cloud microphysics parameterization, the GFS sa-TKE-EDMF boundary layer and free atmospheric turbulence schemes for boundary layer parameterization, the GFS sa-SAS deep convection parameterization scheme and the sa-MF shallow convection parameterization scheme for convection parameterization, the RRTMG shortwave / longwave radiation parameterization scheme for radiative transfer parameterization, the Noah land surface process scheme for land surface parameterization, and the GFS surface layer parameterization scheme for near-surface layer parameterization. To reduce the impact of gravity waves generated by complex terrain forcing on the wind field, the Unified Gravity Wave physics scheme is used. This scheme is used to handle the gravity wave process excited by terrain forcing at the sub-grid scale that cannot be directly resolved by the model resolution. Terrain gravity wave drag parameterization is applied to the sub-grid scale.

[0159] CCPP aims to standardize, modularize, and share physical schemes across models. The CCPP physical parameterization scheme package is a physical scheme package that integrates most parameterization schemes (including cloud microphysics, convection, radiation, boundary layer, land surface processes, etc.). Researchers can directly call it in the model according to the study area and simulation requirements. Specifically, when building downscaled cases, the CCPP_SUITES switch and the CCPP version to be called are set, and the selected parameterization scheme and the parameters in the parameterization scheme are set in configuration files such as namelist.

[0160] Furthermore, the Unified Gravity Wave physics scheme: (1) In actual simulations, when the model resolution cannot directly resolve gravity waves excited by small-scale terrain forcing such as mountains and canyons, these unresolved processes are explicitly expressed through physical parameterization. The Unified Gravity Wave scheme is used to estimate the momentum dissipation (dragging) effect of these subgrid-scale gravity waves on the background wind field during their propagation in the troposphere and stratosphere, especially having a significant impact on the simulation of near-surface wind speed structure and strong wind events. (2) Similarly, in practical applications, this parameterization scheme is mainly called in the model namelist configuration when constructing downscaled cases, and the drag coefficient, drag height, and other parameter values ​​are set according to the research needs.

[0161] Step 2: Introduce the stretched mesh transformation method into the global model architecture to form a global stretched mesh model.

[0162] To achieve high-resolution local simulations on a global model architecture, a Schmidt transform is introduced onto the original cubic spherical uniform mesh. This transforms the original uniform mesh analytically, gradually densifying it in the target region and correspondingly sparsening it in the opposite regions without altering the six-sided topology of the original spherical mesh. The specific steps are as follows:

[0163] The mesh coordinate transformation coefficient is calculated based on the selected mesh stretching coefficient, as shown in the above relationship (1).

[0164] Furthermore, the original latitude Mapped to the stretch grid latitude φ, as shown in equation (2) above.

[0165] Furthermore, by mapping the latitude and longitude of the original cubic sphere grid to the South Pole, only the latitude direction of the original grid needs to be transformed (according to the above relationship (2)), while the longitude remains unchanged. Then, the entire latitude and longitude coordinate system is rotated to move the South Pole to the center point latitude and longitude coordinates of the target area.

[0166] Furthermore, after the above two steps, any original cubic spherical uniform mesh can be converted to a mesh spacing of approximately 1 / s in the target region. Taking the original cubic spherical uniform mesh C100 (each face has 100*100 grid points) as an example, the original resolution is approximately 100km. For a selected target region, taking the stretching factor s = 4, the average resolution of the stretched target surface (the sphere where the spatial densification area is located) is approximately... The minimum resolution of the target surface is approximately The coarsest resolution is approximately On the opposite side of the Earth, the grid spacing is magnified to approximately s times the original (about 400 km). The stretched grid produced by this transformation has the highest resolution at the center of the target and transitions smoothly and continuously outward to the coarse resolution of the background without abrupt boundary changes.

[0167] Furthermore, in the physical process parameterization scheme, a resolution-aware parameterization method is employed for grids of different horizontal resolutions. This allows the parameterization scheme to adaptively adjust with changes in grid scale, thereby ensuring the continuity of physical processes in the coarse-to-fine grid transition zone. For example, within the global stretched grid model framework, when the resolution of the locally refined region approaches the analytical scale of convection, a resolution-aware convection scheme can be used to gradually weaken or even disable deep convection parameterization; while in coarser resolution regions, convection parameterization is retained. In this way, consistent physical process descriptions can be achieved globally and locally within the same model framework, maintaining the physical rationality and numerical stability of simulation results in the coarse-to-fine grid transition zone.

[0168] Furthermore, this method eliminates the need for adding new mesh surfaces or manually specifying nested boundaries, maintaining the consistency of the global mesh while achieving a smooth transition from coarse to high resolution. As the stretching coefficient *s* changes, the spatial resolution and refinement range of the target region can be flexibly adjusted: a larger *s* results in a more concentrated high-resolution region, a larger stretching ratio, and a finer spatial resolution, while the surrounding coarse-resolution region expands accordingly. Therefore, through this stretching process, the influence of complex terrain and small-scale turbulence on the wind field can be precisely captured, while global-scale climate background information can be preserved, making it an effective means of obtaining refined wind resource information.

[0169] 2. Construction and embedding of the theoretical gust calculation module, such as... Figure 4 As shown, it specifically includes:

[0170] Traditional numerical models typically output wind field information based on vertical layer resolution, providing multi-layered three-dimensional wind vector information (including zonal wind, meridional wind, and wind speed), as well as near-surface wind speed products. However, the impact of extreme gust events cannot be ignored during the planning, design, and operation phases of wind farms and photovoltaic power plants, especially for the structural safety and operational stability of the units. Referring to the WMO definition of gusts, the definition of a gust, i.e., a preset gust, is defined as the maximum instantaneous wind speed with a 3-second moving average within a certain time window T (T is generally 1 hour in numerical models). Ideally, gust characteristics should be obtained based on high-frequency wind speed time series. However, due to the limited output frequency of traditional numerical models (usually hourly), directly obtaining 3-second wind speed sequences is computationally expensive and difficult to achieve in long-term, wide-area simulations. Therefore, it is necessary to estimate gust wind speeds based on wind speed outputs with limited time resolution using theoretical or empirical formulas to meet engineering application requirements. Thus, gust wind speed can be expressed as the above relationship (4).

[0171] Furthermore, gust speed can be considered as a function of gust factor and average wind speed, as shown in equation (5) above. The gust factor GF is further defined as the ratio of gust speed to average wind speed, as shown in equation (6) above.

[0172] Further defining the disturbance wind speed as the product of the peak coefficient and the standard deviation of the wind speed over a period of time, can be expressed as the above relationship (7).

[0173] Further defining the turbulence intensity index, the gust factor can be expressed as the above relationship (8).

[0174] To further consider the modulating effect of topography and surface roughness on wind speed, referring to the Parratt model and the Wieringa empirical model, after taking topography and surface roughness into account, the above relationship (8) can be transformed into the above relationship (9).

[0175] Furthermore, the aforementioned gust calculation model is added to the diagnostic calculation module of the global stretched grid model, enabling data interaction through the model variable interface. During model operation, this gust calculation model automatically reads variables such as average wind speed, terrain height, and surface roughness from the model output. Based on these input variables, the output frequency and time window T of the gust speed are set (consistent with the main model output). In the diagnostic module, the instantaneous gust speed within the time window T is calculated online according to the target gust calculation model, and then generated as an independent output gust speed through the model I / O, thereby providing extreme wind speed design references for wind power and photovoltaic projects.

[0176] 3. Preparation for downscaling test, such as Figure 5 As shown, it specifically includes:

[0177] Step 1: Create the stretched mesh for the target area.

[0178] First, based on research requirements, an original cubic spherical global uniform grid matching the spatial resolution of the driving data is selected, and the target region to be scaled down is determined within it. Using the center latitude and longitude of this target region as the center of grid stretching, the corresponding stretching coefficient is calculated by comparing the original spatial resolution with the target output resolution. Then, a grid stretching transformation is performed, smoothly refining the original cubic spherical grid along the latitudinal direction in the target region and correspondingly sparsening it on the opposite side, resulting in a global variable-resolution grid with a high regional resolution. Subsequently, during runtime, the model configuration file is set to point to this stretched grid to enable subsequent simulation calls.

[0179] Step 2: Terrain adaptation.

[0180] SRTM data with a spatial resolution of 90m from NASA was used as the model's base terrain data. Considering that the model's dynamic framework uses a cubic spherical grid (typically divided into six cubic spheres), it is necessary to convert the terrain data from the traditional latitude and longitude grid to the cubic spherical coordinate system and resample the terrain data onto the cubic spherical stretched grid.

[0181] The specific process is as follows: First, the SRTM data is projected onto the original uniform cubic spherical grid. To match the effect of the stretched grid, the projected terrain data needs to be resampled according to the new coordinates of the target stretched grid and the grid density. This ensures that the grid density of the terrain data matches the grid distribution of the target stretched grid, preserving high-resolution terrain details in the target area, while simplifying the outer areas according to the grid sparseness. An interpolation algorithm is then used to map the resampled terrain data to each grid point of the target stretched grid, achieving a match between the terrain data and the target stretched grid. Simultaneously, to avoid numerical instability caused by abrupt changes in terrain data or high-frequency noise, the interpolated terrain data undergoes terrain smoothing, terrain gradient processing, and filtering to remove small-scale noise, balancing terrain detail with the stability of the dynamic solution.

[0182] 4. Downscaling experimental design and execution, such as Figure 6 As shown.

[0183] Based on research needs, driving data (e.g., for resource assessment, reanalysis data such as ERA5 can be selected; for resource prediction, multi-scale prediction data such as IFS, GFS, and CFS can be selected) are selected as the initial and boundary conditions for the target global stretched grid model. Case initialization is completed by creating downscaling cases, setting the simulation start time, integration duration, and output frequency, and specifying the paths to the target stretched grid and the target base terrain dataset in the model configuration file. To more accurately characterize the wind field features within the boundary layer, especially to meet the wind resource data requirements of photovoltaic power plants and wind turbines with different hub heights, the vertical resolution of the near-surface layer below the boundary layer is further set in the model's vertical layer. Specific vertical heights are set to 10m, 50-150m (intervals of 10m), 180m, and 200m, adaptable to the operating heights of various wind turbine types. Subsequently, the global stretched grid model is run for downscaling calculations. During the run, wind resource information, including wind speed, wind direction, wind power density, and gust speed, will be automatically solved based on the stretched grid, interpolated terrain field, and driving field data. The entire wind resource dynamics downscaling simulation process can be completed efficiently under a unified model framework, achieving a seamless transition from large-scale meteorological driving data to high-precision wind resource fields with strong terrain adaptability and high vertical resolution, providing refined support for wind energy development and utilization.

[0184] Furthermore, to verify the effectiveness of the model, two sets of experiments were designed and conducted. One set of experiments used the original cubic spherical uniform grid of 1°×1° as the global base grid (denoted as the original experiment). The other set of experiments took the complex terrain region of western my country as the representative, selected the eastern part of the Qinghai-Tibet Plateau as the target region, set the center of the stretched grid to 35°N and 95°E, and set the stretching coefficient to 8, thereby achieving a high-resolution simulation of 0.125°×0.125° in this region (denoted as the downscaling experiment).

[0185] Both sets of experiments used seasonal forecast data as the boundary conditions and driving fields for a global stretched grid model, simulating seasonal forecasts of summer wind speeds over the past 20 years. The simulation was conducted over 180 days each year, completing a downscaling simulation of wind resource dynamics for the region. Both sets of experiments were performed on the same supercomputing platform, using the same number of physical CPU cores. The first set ran for approximately 4 hours per year, while the second set ran for approximately 9 hours (WRF runs for approximately 20 hours under the same configuration).

[0186] From such Figure 7The simulation results show that, compared to the original experiment, the downscaling experiment exhibits a more refined spatial distribution of climatological wind speed within the target area, especially in the complex terrain of the Tibetan Plateau, where the simulation results are closer to actual observations, particularly in the Tibetan Plateau region. In areas far from the center of the target area, the resolution gradually decreases, and a distinct mosaic grid shape gradually appears north of 45°N and east of 130°E, but the spatial distribution is similar to the original results. This is also an advantage of downscaling a global stretched grid model; as a global model, it spatially refines the resolution of the area of ​​interest without significantly losing the distribution characteristics of other regions globally. Further quantitative evaluation shows that in East Asia, the spatial correlation coefficient between the downscaling experiment's summer climatological wind speed spatial distribution and the observed data reaches 0.95, better than the 0.92 of the original experiment; in terms of temporal evolution, for the western part of the Tibetan Plateau (… Figure 7 The time correlation coefficient between the summer wind speed anomaly and the time series of observed data in the downscaling experiment (e, 28°-34°N, 82°-92°E) was 0.62, significantly higher than the 0.35 in the original experiment. These results indicate that the wind resource dynamics downscaling model based on the global stretched grid model has significant advantages in improving the accuracy of wind resource simulation in key areas. In the figure, (a) and (b) represent the spatial distribution (unit: m / s) of the summer 10m wind speed climatology over the past 20 years in the output results of the original experiment (1°×1°) and the downscaling experiment (0.125°×0.125°), respectively; (ce) represents the temporal evolution of the regional average summer 10m wind speed anomaly over the past 20 years, with each region being the Northeast region (40°-50°N, 120°-130°E), the East region (30°-36°N, 115°-122°E), and the West region (28°-34°N, 82°-92°E). The black solid line represents the observation results, the light gray dotted line represents the original experiment results, and the dark gray dashed line represents the downscaling experiment results.

[0187] The wind resource dynamics downscaling method based on a global stretched grid model using a finite volume cubic spherical grid (FV3) dynamic framework presented in this example has the following effects:

[0188] 1. Within the global grid system, by introducing a stretched grid transformation method, the spatial resolution of the target area can be refined at any location. This preserves large-scale information while achieving a seamless transition from large-scale driving data to local high-precision wind resource fields. Simultaneously, a terrain preprocessing scheme matching the model's grid structure is constructed. Through terrain gradient adjustment and dynamic resolution refinement, the accuracy of wind resource simulation in complex terrain areas is improved, and the numerical stability of the model under high-resolution conditions is ensured.

[0189] 2. To address the issue that traditional numerical models only output average wind speeds and fail to reflect extreme gust characteristics, this example embeds a gust calculation module into a global stretched grid model. To further adapt to the wind resource characteristics of complex terrain, the influence of terrain height and ground roughness on gust characteristics is considered in the gust calculation module. This module is also embedded in the global stretched grid downscaling system, allowing for the direct output of maximum gust wind speeds based on the model's output time intervals. This significantly improves the model's ability to characterize short-duration strong wind events and provides technical support for applications such as new energy resource assessment under complex terrain conditions, wind farm extreme wind speed design, and operational safety assessment.

[0190] 3. Relying on the constraints of physical equations, this method can more accurately reflect the interaction between local wind resource distribution and large-scale circulation background. Compared with statistical downscaling methods that rely on historical statistical relationships, it has stronger extrapolation capabilities and adaptability to climate change conditions. Compared with traditional dynamic downscaling methods using regional models, this method introduces a stretched grid transformation method based on a global model framework, which can achieve progressive densification of spatial resolution in any target area, making the grid present a continuous transition from dense to sparse from the center outward. During the integration process, it maintains the smoothness and physical consistency of the difference calculation, taking into account both global coverage and local accuracy, improving the simulation accuracy of key areas while retaining large-scale information features. In addition, the finite volume cubic spherical grid used has good mass conservation properties, numerical stability, and scalability, supporting efficient large-scale parallel computing. Compared with traditional regional dynamic downscaling methods, the overall computational efficiency can be improved by about 2 to 3 times.

[0191] This embodiment also provides a wind resource dynamics downscaling device based on a global stretched grid model. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0192] This embodiment provides a wind resource dynamics downscaling device based on a global stretched grid model, such as... Figure 8 As shown, the device includes:

[0193] Module 801 is used to acquire the initial global model framework, which uses the original cubic spherical uniform grid.

[0194] The first generation module 802 is used to stretch the original cubic spherical uniform mesh using the Schmidt transform method in the initial global model framework and generate the initial global stretched mesh model.

[0195] The second generation module 803 is used to add a theoretical gust calculation model to the diagnostic calculation module of the initial global stretched mesh model and generate the target global stretched mesh model.

[0196] The processing module 804 is used to downscale the wind resources of the target area to be downscaled using the target global stretched grid model, so as to obtain a high-resolution wind resource information set.

[0197] Further functional descriptions of the above modules are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0198] In this embodiment, the wind resource dynamics downscaling device based on the global stretched grid model is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit), a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0199] This invention also provides a computer device having the above-described features. Figure 8 The wind resource dynamics downscaling device shown is based on a global stretched grid model.

[0200] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 9 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 9 Take a processor 10 as an example.

[0201] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GPA), or any combination thereof.

[0202] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0203] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0204] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0205] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0206] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0207] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0208] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A wind resource dynamics downscaling method based on a global stretched grid model, characterized in that, The method includes: Obtain an initial global model framework, which adopts an original cubic spherical uniform grid. In the initial global model framework, the original cubic spherical uniform mesh is stretched using the Schmidt transform method to generate the initial global stretched mesh model; Add a theoretical gust calculation model to the diagnostic calculation module of the initial global stretched grid model and generate a target global stretched grid model. The theoretical gust calculation model is used to simulate and calculate the instantaneous gust wind speed during the model operation by referring to the WMO definition of gusts and based on the parameters output by the global stretched grid model at a finite time resolution, using theoretical and empirical formulas. This can enhance the model's ability to output extreme characteristics of wind resources. The wind resources of the target region to be downscaled are processed using the target global stretched grid model to obtain a high-resolution wind resource information set. This includes obtaining the initial global model framework, including: A preset non-hydrostatic global numerical model is obtained. The preset non-hydrostatic global numerical model adopts a dynamic framework of finite volume cubic spherical mesh and is constructed by approximately solving the three-dimensional compressible non-hydrostatic atmospheric equations on the cubic spherical mesh using the finite volume method. In the preset non-hydrostatic global numerical model dynamic framework, a time integration scheme based on horizontal explicit and vertical semi-implicit is configured to obtain the global numerical model. The time integration scheme based on horizontal explicit and vertical semi-implicit represents a time discretization method used in atmospheric numerical models to solve atmospheric dynamic equations, which can take into account both the computational efficiency of horizontal small-scale explicit calculations and the stability of vertical acoustic implicit constraints. Using a pre-defined physics scheme package, the physical process parameters in the global numerical model are configured to obtain the initial global model framework. The pre-defined physics scheme package represents a standardized and modular set of multiple physical process parameterization schemes for the numerical model, which are used for unified invocation, configuration and accurate simulation of atmospheric physical processes in the numerical model. The theoretical gust calculation model is obtained through the following steps: Obtain the initial gust calculation model based on the preset gust definition; The initial gust factor is determined based on the initial gust calculation model, peak value coefficient, wind speed standard deviation, and turbulence intensity index. Considering the modulating effect of terrain and surface roughness on wind speed, the initial gust factor is transformed using the Parratt model and the Wieringa empirical model to obtain the target gust factor. The theoretical gust calculation model is determined based on the target gust factor, average wind speed, and the initial gust calculation model. The initial gust calculation model is expressed as the following relation: In the formula: express Moment, Time Window The wind speed inside; Indicates average wind speed; The gust factor is expressed as the following formula: In the formula: express Moment, Time Window The internal disturbance wind speed.

2. The method according to claim 1, characterized in that, In the initial global model framework, the original cubic spherical uniform mesh is stretched using the Schmidt transform method to generate an initial global stretched mesh model, including: In the initial global model framework, based on the preset mesh stretching coefficient and the Schmidt transform method, the original cubic spherical uniform mesh is analytically deformed to obtain the target stretched mesh. The six-sided topology of the target stretched mesh is consistent with the six-sided topology of the original cubic spherical uniform mesh. Based on the target stretched mesh, the initial global stretched mesh pattern is generated through a preset physical process parameterization scheme, which has resolution-aware characteristics.

3. The method according to claim 2, characterized in that, Within the initial global model framework, based on a preset mesh stretching coefficient and the Schmidt transform method, the original cubic spherical uniform mesh is analytically deformed to obtain the target stretched mesh, including: The mesh coordinate transformation coefficient is calculated using the mesh stretching coefficient. In the initial global model framework, the latitude of the original cubic spherical uniform grid is stretched according to the grid coordinate transformation coefficient to obtain multiple stretched grid latitudes; Using rigid body rotation, the latitudes of multiple original grids in the encrypted region corresponding to the original cubic spherical uniform grid are rotated to the latitudes of multiple stretched grids to obtain the target stretched grid. The encrypted region is the area in the original cubic spherical uniform grid that is concentrated near the South Pole due to the change in the latitude of the original grids.

4. The method according to claim 1, characterized in that, The wind resources of the target region to be downscaled in the target global stretched grid model are downscaled using the target global stretched grid model to obtain a high-resolution wind resource information set, including: Obtain the initial base high-resolution terrain dataset and the preset driving dataset; Using the center latitude and longitude of the target region as the center of the mesh stretching, the original cubic spherical uniform mesh is stretched using the Schmidt transform method to obtain the target stretched mesh of the target region. The initial high-resolution terrain dataset is processed to obtain the target terrain dataset; Based on the preset driving dataset, the target stretched grid, and the target basic terrain dataset, the target global stretched grid model is downscaled to obtain the target region's downscaled high-resolution wind resource information set.

5. A wind resource dynamics downscaling device based on a global stretched grid model, characterized in that, The device includes: The acquisition module is used to acquire an initial global model framework, which adopts an original cubic spherical uniform grid. The first generation module is used to stretch the original cubic spherical uniform mesh using the Schmidt transform method and generate an initial global stretched mesh pattern within the initial global pattern framework. The second generation module is used to add a theoretical gust calculation model to the diagnostic calculation module of the initial global stretched grid model and generate a target global stretched grid model. The theoretical gust calculation model is used to simulate and calculate the instantaneous gust wind speed during the model operation by referring to the WMO definition of gusts and based on the parameters output by the global stretched grid model at a finite time resolution, using theoretical and empirical formulas. This can enhance the model's ability to output extreme characteristics of wind resources. The processing module is used to downscale the wind resources of the target area to be downscaled using the target global stretched grid pattern, so as to obtain a high-resolution wind resource information set. The acquisition module is specifically used for: A preset non-hydrostatic global numerical model is obtained. The preset non-hydrostatic global numerical model adopts a dynamic framework of finite volume cubic spherical mesh and is constructed by approximately solving the three-dimensional compressible non-hydrostatic atmospheric equations on the cubic spherical mesh using the finite volume method. In the preset non-hydrostatic global numerical model dynamic framework, a time integration scheme based on horizontal explicit and vertical semi-implicit is configured to obtain the global numerical model. The time integration scheme based on horizontal explicit and vertical semi-implicit represents a time discretization method used in atmospheric numerical models to solve atmospheric dynamic equations, which can take into account both the computational efficiency of horizontal small-scale explicit calculations and the stability of vertical acoustic implicit constraints. Using a pre-defined physics scheme package, the physical process parameters in the global numerical model are configured to obtain the initial global model framework. The pre-defined physics scheme package represents a standardized and modular set of multiple physical process parameterization schemes for the numerical model, which are used for unified invocation, configuration and accurate simulation of atmospheric physical processes in the numerical model. The theoretical gust calculation model is obtained through the following steps: Obtain the initial gust calculation model based on the preset gust definition; The initial gust factor is determined based on the initial gust calculation model, peak value coefficient, wind speed standard deviation, and turbulence intensity index. Considering the modulating effect of terrain and surface roughness on wind speed, the initial gust factor is transformed using the Parratt model and the Wieringa empirical model to obtain the target gust factor. The theoretical gust calculation model is determined based on the target gust factor, average wind speed, and the initial gust calculation model. The initial gust calculation model is expressed as the following relation: In the formula: express Moment, Time Window The wind speed inside; Indicates average wind speed; The gust factor is expressed as the following formula: In the formula: express Moment, Time Window The internal disturbance wind speed.

6. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the wind resource dynamics downscaling method based on a global stretched grid model as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the wind resource dynamics downscaling method based on a global stretched grid model as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, Includes computer instructions for causing a computer to execute the wind resource dynamics downscaling method based on a global stretched grid model as described in any one of claims 1 to 4.

Images