Urban wind field prediction method, device, terminal and storage medium

Abstract

The present invention discloses a method, device, terminal, and storage medium for predicting urban wind fields. The method obtains numerical weather forecast data corresponding to a target area, determines an initial wind field image based on the numerical weather forecast data, determines a number of image channel components based on the initial wind field image, obtains a pre-trained wind field prediction model, and uses the wind field prediction model to learn the number of image channel components to determine a target wind field image. Because the present invention rapidly generates a rough initial wind field image using numerical weather forecast data and performs image-to-image conversion using the pre-trained wind field prediction model to obtain a more accurate target wind field image, the method significantly reduces computational time and reliance on expensive physical computing resources while maintaining high prediction accuracy. Therefore, the method effectively addresses the time-consuming and costly computational costs associated with using computational fluid dynamics to simulate urban wind fields in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to an urban wind field prediction method, device, terminal and storage medium. Background Art

[0002] Currently, computational fluid dynamics (CFD) methods are often used to simulate urban wind farms. Computational fluid dynamics simulation is a technology that uses data calculation methods to solve fluid flow. It can perform high-resolution spatial simulations of meteorological phenomena and present the meteorological characteristics and changes of local areas in detail. In addition, computational fluid dynamics also has the ability to handle complex physical processes, such as multiphase flow simulation and turbulence simulation. However, computational fluid dynamics is computationally complex and time-consuming, especially when dealing with high-dimensional complex fluid systems, the computational cost is extremely high.

[0003] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an urban wind field prediction method, device, terminal and storage medium in response to the above-mentioned defects of the existing technology, aiming to solve the problem that computational fluid dynamics is used in the existing technology to simulate urban wind fields, resulting in time-consuming and costly calculations.

[0005] The technical solutions adopted by the present invention to solve the problem are as follows:

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting an urban wind field, wherein the method includes:

[0007] Acquiring numerical weather forecast data corresponding to the target area, and determining an initial wind field image based on the numerical weather forecast data;

[0008] determining a plurality of image channel components according to the initial wind field image;

[0009] A pre-trained wind field prediction model is obtained, and the plurality of image channel components are learned using the wind field prediction model to determine a target wind field image.

[0010] In one implementation method, determining a plurality of image channel components based on the initial wind field image includes:

[0011] determining east-west wind speed data and north-south wind speed data according to the initial wind field image;

[0012] The east-west wind speed data and the north-south wind speed data are used as the image channel components.

[0013] In one implementation method, determining a plurality of image channel components based on the initial wind field image further includes:

[0014] determining wind speed data and wind direction data according to the initial wind field image;

[0015] Calculating the wind direction sine value and wind direction cosine value corresponding to the wind direction data;

[0016] The wind speed data, the wind direction sine value, and the wind direction cosine value are used as the image channel components.

[0017] In one implementation method, the method for constructing a training set of the wind field prediction model includes:

[0018] Obtaining a first altitude threshold and historical meteorological observation data, and performing interpolation calculation based on the first altitude threshold and the historical meteorological observation data to determine historical numerical weather forecast data;

[0019] The training set is constructed according to the historical numerical weather forecast data.

[0020] In one implementation method, constructing the training set based on the historical numerical weather forecast data includes:

[0021] Classifying the historical numerical weather forecast data to determine a number of data categories;

[0022] selecting from the historical numerical weather forecast data according to the data category to determine a number of historical numerical weather forecast data for training;

[0023] The training set is constructed according to the plurality of historical numerical weather forecast data used for training.

[0024] In one implementation method, constructing the training set based on the plurality of historical numerical weather forecast data for training includes:

[0025] Determining, based on each of the historical numerical weather forecast data used for training, a historical initial wind field image corresponding to each of the historical numerical weather forecast data used for training;

[0026] Using computational fluid dynamics to simulate each of the historical numerical weather forecast data used for training, and determining a historical simulated wind field image corresponding to each of the historical numerical weather forecast data used for training;

[0027] The training set is constructed according to the historical initial wind field image and the historical simulated wind field image.

[0028] In one implementation method, the wind field prediction model is a conditional generative adversarial network based on an image-to-image conversion model, the generator of the conditional generative adversarial network adopts a U-shaped network architecture, and the judge of the conditional generative adversarial network adopts a block generative adversarial network classifier.

[0029] In a second aspect, an embodiment of the present invention further provides an urban wind field prediction device, wherein the urban wind field prediction device includes:

[0030] A data acquisition module is used to obtain numerical weather forecast data corresponding to the target area and determine an initial wind field image based on the numerical weather forecast data;

[0031] a channel construction module, configured to determine a plurality of image channel components according to the initial wind field image;

[0032] The wind field prediction module is used to obtain a pre-trained wind field prediction model, learn the plurality of image channel components through the wind field prediction model, and determine a target wind field image.

[0033] In a third aspect, an embodiment of the present invention further provides a terminal comprising a memory and one or more processors; the memory stores one or more programs; the programs include instructions for executing any of the urban wind field prediction methods described above; and the processor is used to execute the programs.

[0034] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium on which a plurality of instructions are stored, wherein the instructions are suitable for being loaded and executed by a processor to implement any of the above-mentioned urban wind field prediction methods.

[0035] Beneficial effects of the present invention: The embodiments of the present invention obtain numerical weather forecast data corresponding to the target area, determine an initial wind field image based on the numerical weather forecast data, determine a number of image channel components based on the initial wind field image, obtain a pre-trained wind field prediction model, learn the number of image channel components through the wind field prediction model, and determine a target wind field image. Because the present invention quickly generates a rough initial wind field image using numerical weather forecast data, and performs image-to-image conversion using a pre-trained wind field prediction model to obtain a more accurate target wind field image, the computation time is greatly reduced, and the reliance on expensive physical computing resources is reduced while maintaining a high degree of prediction accuracy. Therefore, the problem of using computational fluid dynamics to simulate urban wind fields in the prior art, which results in time-consuming and costly computations, can be effectively solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1 It is a flow chart of the urban wind field prediction method provided by an embodiment of the present invention.

[0038] Figure 2 This is a flow chart of an embodiment of the training of the urban wind field prediction model provided by an embodiment of the present invention.

[0039] Figure 3 It is a schematic diagram of the internal modules of the urban wind field prediction device provided by an embodiment of the present invention.

[0040] Figure 4 This is a principle block diagram of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The present invention discloses a method, device, terminal, and storage medium for predicting urban wind fields. To clarify the objectives, technical solutions, and effects of the present invention, the present invention is further described below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are intended only to illustrate the present invention and are not intended to limit the present invention.

[0042] It will be understood by those skilled in the art that, unless expressly stated otherwise, the singular forms "a", "an", "said" and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the description of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, "connected" or "coupled" as used herein may include wireless connections or wireless couplings. The term "and / or" used herein includes all or any units and all combinations of one or more associated listed items.

[0043] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0044] Currently, computational fluid dynamics (CFD) methods are commonly used to simulate urban wind farms. CFD simulation is a technology that uses data calculation methods to solve fluid flow problems. It can perform high-resolution spatial simulations of meteorological phenomena, presenting detailed meteorological characteristics and changes in local areas. CFD also has the ability to handle complex physical processes, such as multiphase flow simulation and turbulence simulation. However, CFD is computationally complex and time-consuming, and the computational cost is extremely high when dealing with high-dimensional complex fluid systems.

[0045] In response to the above-mentioned shortcomings of the prior art, the present invention provides a method for predicting urban wind fields. The method obtains numerical weather forecast data corresponding to a target area, determines an initial wind field image based on the numerical weather forecast data, determines a number of image channel components based on the initial wind field image, obtains a pre-trained wind field prediction model, and uses the wind field prediction model to learn the number of image channel components to determine a target wind field image. Because the present invention rapidly generates a rough initial wind field image using numerical weather forecast data and performs image-to-image conversion using a pre-trained wind field prediction model to obtain a more accurate target wind field image, the method significantly reduces computational time and reliance on expensive physical computing resources while maintaining a high level of prediction accuracy. Therefore, the method effectively addresses the time-consuming and costly computational costs associated with using computational fluid dynamics to simulate urban wind fields in the prior art.

[0046] Exemplary methods:

[0047] like Figure 1 As shown, the method includes:

[0048] Step S100: Acquire numerical weather forecast data corresponding to the target area, and determine an initial wind field image according to the numerical weather forecast data.

[0049] Specifically, numerical weather forecast data (NWP) is data obtained by simulating and predicting atmospheric movement through numerical calculation methods. The data basis of numerical weather forecast is a large amount of meteorological observation data, including data collected by ground weather stations, high-altitude sounding balloons, meteorological satellites, radars and other observation equipment. Numerical weather forecast data contains the predicted values ​​of various meteorological elements, such as temperature, humidity, air pressure, wind direction, wind speed, precipitation, etc. The numerical weather forecast data corresponding to the target area is obtained, and the numerical weather forecast data is formatted into a rough two-dimensional initial wind field image. This embodiment breaks the limitations of traditional numerical simulation by obtaining numerical weather forecast data as initial data to perform wind field prediction, reduces the demand for detailed physical parameters and the dependence on expensive physical computing resources, and improves the efficiency of wind field prediction.

[0050] Step S200: determining a plurality of image channel components according to the initial wind field image.

[0051] Simply put, the initial wind image is converted into several image channel components, which can then be fed into the wind prediction model for learning, resulting in the target wind image. The following two methods for converting the initial wind image into several image channel components are provided. Depending on the number of image channel components obtained, the number of channels in the wind prediction model also varies.

[0052] In one implementation, determining a plurality of image channel components according to the initial wind field image includes:

[0053] determining east-west wind speed data and north-south wind speed data according to the initial wind field image;

[0054] The east-west wind speed data and the north-south wind speed data are used as the image channel components.

[0055] Simply put, two horizontal wind speed channels are constructed for the wind field prediction model: a north-south wind speed channel and an east-west wind speed channel. The east-west wind speed data u and the north-south wind speed data v are extracted from the initial wind field image and used as image channel components so that the wind field prediction model can process them and predict the target wind field image.

[0056] In one implementation, determining a plurality of image channel components according to the initial wind field image further includes:

[0057] determining wind speed data and wind direction data according to the initial wind field image;

[0058] Calculating the wind direction sine value and wind direction cosine value corresponding to the wind direction data;

[0059] The wind speed data, the wind direction sine value, and the wind direction cosine value are used as the image channel components.

[0060] Considering that using east-west wind speed data u and north-south wind speed data v as image channel components can lead to large wind direction errors in some cases, especially in areas with lower wind speeds. To address this issue, this embodiment improves the prediction performance of the wind field prediction model by setting three channels to extract more useful information from the initial wind field image.

[0061] Specifically, corresponding to the number of channels of the wind field prediction model, wind speed data and wind direction data are extracted from the initial wind field image, and the wind direction sine value and wind direction cosine value corresponding to the wind speed data and wind direction data are used as image channel components, which are specifically expressed as follows:

[0062] Wind speed data: used to indicate the overall strength of the wind.

[0063] Wind direction sine value (sin(θ), where θ is the wind direction data): represents the sine value of the wind direction data to smooth changes in wind direction and avoid directional discontinuities.

[0064] Wind direction cosine value (cos(θ), where θ is the wind direction data): represents the cosine value of the wind direction data, which is used to smooth wind direction changes and avoid directional discontinuities.

[0065] This embodiment avoids the errors that may be caused by directly using the inverse tangent function to calculate the wind direction when the wind speed is low by using sin(θ) and cos(θ), enhances the wind field prediction model's ability to capture wind field details, and enables the wind field prediction model to better handle the prediction problem of the local wind field.

[0066] Step S300: Obtain a pre-trained wind field prediction model, learn the plurality of image channel components through the wind field prediction model, and determine a target wind field image.

[0067] Simply put, by using the trained wind field prediction model to learn the image channel components, the target wind field image corresponding to the target area can be quickly obtained. The accuracy of the target wind field image is similar to the simulated wind field image of computational fluid dynamics simulation, realizing the mapping from the rough initial wind field image to the high-precision simulated wind field image, improving computational efficiency and reducing computational cost.

[0068] In one implementation, the method for constructing a training set of the wind field prediction model includes:

[0069] Obtaining a first altitude threshold and historical meteorological observation data, and performing interpolation calculation based on the first altitude threshold and the historical meteorological observation data to determine historical numerical weather forecast data;

[0070] The training set is constructed according to the historical numerical weather forecast data.

[0071] Simply put, historical meteorological observation data is data collected by observation equipment such as ground-based weather stations, high-altitude sounding balloons, meteorological satellites, and radar. Historical meteorological observation data covers a specific area over a period of time, which can be several years or months. The first altitude threshold is the height above the ground. Interpolation is performed based on the collected historical meteorological observation data and the first altitude threshold to obtain historical numerical weather forecast data at the location of the first altitude threshold. This allows the training set constructed based on historical numerical weather forecast data to provide more meteorological patterns.

[0072] In one implementation, constructing the training set based on the historical numerical weather forecast data includes:

[0073] Classifying the historical numerical weather forecast data to determine a number of data categories;

[0074] selecting from the historical numerical weather forecast data according to the data category to determine a number of historical numerical weather forecast data for training;

[0075] The training set is constructed according to the plurality of historical numerical weather forecast data used for training.

[0076] Specifically, historical numerical weather forecast data includes various types of meteorological conditions. This data is categorized based on the wind speed and wind direction combinations within the data. For example, in the summer, wind speeds in a particular region are generally low, with breezes or light winds being predominant. Under the influence of special weather systems, such as typhoons, higher wind speeds can occur, with winds primarily from the southwest. In winter, cold air frequently influences the region, resulting in relatively high wind speeds and northwesterly winds. Based on meteorological events occurring at different times and characteristics such as wind speed and direction, the data categories corresponding to the historical numerical weather forecast data can be categorized as: strong winds, northwesterly winds; strong winds, southwesterly winds; and light winds, southwesterly winds. Several historical numerical weather forecast data sets corresponding to each data category are randomly selected. These selected data sets are used as training sets, and a training set is constructed based on these training sets. Since the historical numerical weather forecast data used to construct the training set in this embodiment are obtained by screening according to data categories, these historical numerical weather forecast data used for training cover different wind speed and wind direction combinations and represent a wide range of meteorological conditions. Therefore, the wind field prediction model trained based on the training set constructed based on the historical numerical weather forecast data used for training has better generalization ability when predicting various types of scenarios.

[0077] In one implementation, constructing the training set based on the plurality of historical numerical weather forecast data for training includes:

[0078] Determining, based on each of the historical numerical weather forecast data used for training, a historical initial wind field image corresponding to each of the historical numerical weather forecast data used for training;

[0079] Using computational fluid dynamics to simulate each of the historical numerical weather forecast data used for training, and determining a historical simulated wind field image corresponding to each of the historical numerical weather forecast data used for training;

[0080] The training set is constructed according to the historical initial wind field image and the historical simulated wind field image.

[0081] Specifically, the historical numerical weather forecast data used for training is converted into historical initial images. The construction speed of the historical initial wind field image is relatively fast, but the accuracy of the historical initial wind field image is relatively low. Computational fluid dynamics is used to simulate the historical numerical weather forecast data used for training to obtain high-precision historical simulated wind field images. For example, computational fluid dynamics software is used to simulate the numerical weather forecast data used for training, and the data of the second height threshold after simulation is extracted to form a historical simulated wind field image. The historical initial wind field image and the historical simulated wind field image corresponding to the same historical numerical weather forecast data are combined into a training data pair. The mapping between the historical initial wind field image and the historical simulated wind field image is learned through the wind field prediction model, and the wind field simulation problem is combined with the image generation technology to reduce the physical computing resources required for wind field simulation.

[0082] In one implementation, the wind field prediction model is a conditional generative adversarial network based on an image-to-image conversion model, the generator of the conditional generative adversarial network adopts a U-shaped network architecture, and the judge of the conditional generative adversarial network adopts a block generative adversarial network classifier.

[0083] Specifically, since the conditional generative adversarial network (cGAN) can learn the mapping from input to output under given conditions and is particularly suitable for image-to-image conversion tasks, this embodiment adopts the conditional generative adversarial network as a wind field prediction model, and inputs the initial wind field image converted from rough numerical weather forecast data into the conditional generative adversarial network to obtain a target wind field image with similar accuracy to the simulated wind field image simulated by computational fluid dynamics, reducing the demand for detailed physical parameters and lowering the computational cost.

[0084] The architecture of the conditional generative adversarial network in this embodiment is based on an image-to-image conversion model (Pix2Pix model). The Pix2Pix model is a deep learning framework specifically designed for image-to-image translation, which is capable of mapping one image into another. The Pix2Pix model works by combining two adversarial modules, a generator and a recognizer: the generator is responsible for generating images similar to wind field images simulated by computational fluid dynamics, while the recognizer is responsible for distinguishing between the generated images (created by the generator) and wind field images simulated by computational fluid dynamics (from the training set). The task of the generator is to make the generated wind field images close to the real computational fluid dynamics results, while the judge evaluates the authenticity of these images.

[0085] Since this embodiment requires converting an initial wind field image, generated from rough numerical weather forecast data, into a target wind field image with similar accuracy to the wind field image simulated by computational fluid dynamics simulations, the Pix2Pix model is well-suited for this application scenario. It can output a detailed target wind field image given a rough initial wind field image. Secondly, the Pix2Pix model's loss function design combines adversarial loss and L1 loss.

[0086] Adversarial loss: The discriminator guides the generator by determining whether the generated target wind image is indistinguishable from the simulated wind image from the CFD simulation, ensuring the authenticity of the generated target wind image. L1 loss: This ensures that the generated image is structurally similar to the target image by calculating the pixel-by-pixel mean absolute error between the generated target wind image and the CFD simulation. This combination of loss functions ensures that the generated image not only looks realistic but also matches the accuracy of the CFD simulation.

[0087] In this embodiment, the generator of the conditional generative adversarial network uses a U-net architecture, including an encoder and decoder, to ensure that the generator can effectively capture and reconstruct the features of the input image. The judge uses a patch generative adversarial network classifier (PatchGAN classifier) ​​to evaluate the authenticity of the image based on local image regions (rather than the entire image), thereby improving the detail and stability of the generated image.

[0088] In one implementation, the U-Net architecture of the generator is adjusted to obtain a deeper encoder part to process higher resolution images, and the number and scale of the upsampling layer are adjusted to ensure that the final generated image matches the target resolution. At the same time, the discrimination range of the discriminator PatchGAN classifier is expanded to cover the entire image. Since the Pix2Pix model is for images with a resolution of 256×256, this embodiment expands the resolution of the image and adds resolutions of 128×128 and 512×512 to adapt to different resolution requirements. When the target wind farm image is at a resolution of 128×128, each pixel of the image represents an area of ​​7.8 meters × 7.8 meters, which is suitable for simulating urban wind farms. The change in image resolution can help users find the best balance between computational cost and accuracy. During training, the generator and discriminator of the wind farm prediction model update the network parameters according to images of different resolutions, with good flexibility and adaptability.

[0089] In one implementation, the wind farm prediction model can also incorporate other deep learning networks, such as Transformer networks, to enhance the model's ability to capture local wind farm details. Furthermore, other generative adversarial networks can be employed to improve the wind farm prediction model's generalization capabilities when handling unseen boundary conditions.

[0090] like Figure 2 As shown, the specific embodiments are as follows:

[0091] The training set is constructed using numerical weather forecast data from the Global Forecast System (GFS). The numerical weather forecast data is obtained by surface interpolation at a height of 20 meters above the ground (the first height threshold). The numerical weather forecast data includes 8760 hours of data for a certain area. The 8760 hours of numerical weather forecast data are classified according to wind speed and wind direction, and events are randomly selected from each data category. Finally, the numerical weather forecast data corresponding to 214 representative events are selected to construct the training set. These events represent a wide range of meteorological conditions to ensure that the training set includes different wind speed and wind direction combinations. Historical initial wind field images and historical simulated wind field images are generated respectively according to the numerical weather forecast data corresponding to the 214 representative events, wherein the historical simulated wind field images are extracted from the data at a height of 2 meters (the second height threshold) during the simulation process of computational fluid dynamics, and a training set is constructed according to the historical initial wind field images and the historical simulated wind field images, and the wind field prediction model is trained.

[0092] The images in the training set generate multiple image channel components before being input into the wind field prediction model, such as two image channel components corresponding to the east-west wind speed data u and the north-south wind speed data v, or three image channel components corresponding to the wind speed data, the sine value of the wind direction, and the cosine value of the wind direction.

[0093] The wind farm prediction model training cycle is set to 15,000 iterations. In each iteration, the generator and the discriminator use independent ADAM optimizers (Adaptive Moment Estimation) to update the network weights, and the learning rate is 2*10 -4 The use of ADAM optimizer can ensure the stability and efficiency of training.

[0094] In addition, numerical weather forecast data of 80 events (different from the training set data) were selected as the test set of the wind field prediction model to test the performance of the wind field prediction model.

[0095] Inputting numerical weather forecast data into a trained wind prediction model generates simulated wind field images with similar accuracy to those from computational fluid dynamics simulations. Using this trained and tested wind prediction model to perform wind field forecasts based on numerical weather forecast data reduces the computational time from eight hours to just a few seconds, compared to traditional computational fluid dynamics-based methods. Test results demonstrate that the wind prediction model can effectively predict local wind fields in urban areas, with an average absolute wind speed error of 0.35 m / s and a wind direction error of 27°.

[0096] Furthermore, tests using initial wind images of varying resolutions revealed that higher image resolutions did not significantly improve the performance of the wind prediction model. The wind prediction model maintained high accuracy even at low resolutions, enabling the system to run quickly with limited computing resources.

[0097] Based on the above embodiments, the present invention also provides an urban wind field prediction device, such as Figure 3 As shown, the device includes:

[0098] Data acquisition module 01 is used to obtain numerical weather forecast data corresponding to the target area and determine the initial wind field image based on the numerical weather forecast data;

[0099] A channel construction module 02 is configured to determine a plurality of image channel components based on the initial wind field image;

[0100] The wind field prediction module 03 is used to obtain a pre-trained wind field prediction model, learn the plurality of image channel components through the wind field prediction model, and determine a target wind field image.

[0101] Based on the above embodiment, the present invention further provides a terminal, whose principle block diagram can be shown as follows: Figure 4 As shown. The terminal includes a processor, a memory, a network interface, and a display screen connected via a system bus. The processor of the terminal is used to provide computing and control capabilities. The memory of the terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the terminal is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the urban wind field prediction method is implemented. The display screen of the terminal can be a liquid crystal display or an electronic ink display.

[0102] Those skilled in the art will understand that Figure 4 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the solution of the present invention, and does not constitute a limitation on the terminal to which the solution of the present invention is applied. The specific terminal may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0103] In one implementation, the terminal has one or more programs stored in its memory, and is configured to be executed by one or more processors. The one or more programs include instructions for performing an urban wind farm prediction method.

[0104] Those skilled in the art will appreciate that all or part of the processes in the above-described embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described embodiments. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0105] In summary, the present invention discloses a method, device, terminal, and storage medium for predicting urban wind fields. The method obtains numerical weather forecast data corresponding to a target area, determines an initial wind field image based on the numerical weather forecast data, determines a number of image channel components based on the initial wind field image, obtains a pre-trained wind field prediction model, and learns the number of image channel components through the wind field prediction model to determine a target wind field image. Since the present invention quickly generates a rough initial wind field image using numerical weather forecast data, and performs image-to-image conversion using a pre-trained wind field prediction model to obtain a more accurate target wind field image, the calculation time is greatly reduced while maintaining a high prediction accuracy. Therefore, it can effectively solve the problem of time-consuming and costly calculations in the prior art of using computational fluid dynamics to simulate urban wind fields.

[0106] It should be understood that the application of the present invention is not limited to the above examples. For those skilled in the art, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for predicting urban wind fields, characterized in that: The method comprises: Acquiring numerical weather forecast data corresponding to the target area, and determining an initial wind field image based on the numerical weather forecast data; determining a plurality of image channel components according to the initial wind field image; Obtaining a pre-trained wind field prediction model, and learning the plurality of image channel components using the wind field prediction model to determine a target wind field image; The method for constructing a training set of the wind field prediction model includes: Determining, based on each of the historical numerical weather forecast data used for training, a historical initial wind field image corresponding to each of the historical numerical weather forecast data used for training; Using computational fluid dynamics to simulate each of the historical numerical weather forecast data used for training, and determining a historical simulated wind field image corresponding to each of the historical numerical weather forecast data used for training; The training set is constructed according to the historical initial wind field image and the historical simulated wind field image.

2. The urban wind field prediction method according to claim 1, characterized in that: The determining of a plurality of image channel components according to the initial wind field image comprises: Determining east-west wind speed data and north-south wind speed data based on the initial wind field image; The east-west wind speed data and the north-south wind speed data are used as the image channel components.

3. The urban wind field prediction method according to claim 1, characterized in that: The determining of a plurality of image channel components according to the initial wind field image further includes: determining wind speed data and wind direction data according to the initial wind field image; Calculating the wind direction sine value and wind direction cosine value corresponding to the wind direction data; The wind speed data, the wind direction sine value, and the wind direction cosine value are used as the image channel components.

4. The urban wind field prediction method according to claim 1, characterized in that: The method for constructing a training set of the wind field prediction model includes: Obtaining a first altitude threshold and historical meteorological observation data, and performing interpolation calculation based on the first altitude threshold and the historical meteorological observation data to determine historical numerical weather forecast data; The training set is constructed according to the historical numerical weather forecast data.

5. The urban wind field prediction method according to claim 4, characterized in that: The step of constructing the training set based on the historical numerical weather forecast data includes: Classifying the historical numerical weather forecast data to determine a number of data categories; selecting from the historical numerical weather forecast data according to the data category to determine a number of historical numerical weather forecast data for training; The training set is constructed according to the plurality of historical numerical weather forecast data used for training.

6. The urban wind field prediction method according to claim 1, characterized in that: The wind field prediction model is a conditional generative adversarial network based on an image-to-image conversion model. The generator of the conditional generative adversarial network adopts a U-shaped network architecture, and the judge of the conditional generative adversarial network adopts a block generative adversarial network classifier.

7. An urban wind field prediction device, characterized in that: The device comprises: A data acquisition module is used to obtain numerical weather forecast data corresponding to the target area and determine an initial wind field image based on the numerical weather forecast data; a channel construction module, configured to determine a plurality of image channel components according to the initial wind field image; A wind field prediction module is used to obtain a pre-trained wind field prediction model, learn the plurality of image channel components through the wind field prediction model, and determine a target wind field image; The method for constructing a training set of the wind field prediction model includes: Determining, based on each of the historical numerical weather forecast data used for training, a historical initial wind field image corresponding to each of the historical numerical weather forecast data used for training; Using computational fluid dynamics to simulate each of the historical numerical weather forecast data used for training, and determining a historical simulated wind field image corresponding to each of the historical numerical weather forecast data used for training; The training set is constructed according to the historical initial wind field image and the historical simulated wind field image.

8. A terminal, characterized in that: The terminal includes a memory and one or more processors; the memory stores one or more programs; the programs include instructions for executing the urban wind field prediction method as described in any one of claims 1-6; and the processor is used to execute the programs.

9. A computer-readable storage medium having a plurality of instructions stored thereon, characterized in that: The instructions are suitable for being loaded and executed by a processor to implement the steps of the urban wind field prediction method described in any one of claims 1 to 6 above.

Images