A low-altitude turbulence field reconstruction and adaptive wind field prediction method and related device

Abstract

The application discloses a low-altitude turbulent field reconstruction and adaptive wind field prediction method and related device, comprising: acquiring multi-source meteorological data, preprocessing the multi-source meteorological data, and then performing feature extraction; inputting the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and a turbulent field; inputting the three-dimensional wind speed field into a sequence prediction model based on the combination of a physical information neural network and a long short-term memory network to predict a three-dimensional wind speed field at a future time, so that the method and the related device can efficiently fuse multi-source observation data, accurately reconstruct a three-dimensional turbulent field, and predict a wind speed field by using a physical information machine learning model.

Description

Technical Field

[0001] This invention belongs to the field of low-altitude meteorological observation and prediction technology, and relates to a method and related device for low-altitude turbulent flow field reconstruction and adaptive wind field prediction. Background Technology

[0002] With the continuous growth of installed capacity of new energy sources such as wind power and the booming development of the low-altitude economy, the demand for refined and timely low-altitude meteorological information, especially real-time monitoring and short-term forecasting of three-dimensional wind and turbulence fields, is becoming increasingly urgent. Accurate wind field forecasting is crucial for decision support in fields such as wind farm operation, drone operations, and general aviation safety. However, existing low-altitude wind field forecasting technologies still face many challenges: First, at the data input level, most existing wind field analysis and forecasting systems rely on a single data source or do not adequately integrate multi-source data. These multi-source data differ in spatiotemporal resolution, observation principles, and accuracy, lacking an effective collaborative preprocessing framework, leading to inconsistencies and uncertainties in the reconstructed initial wind field, thus limiting the accuracy of the initial forecast. Second, at the forecasting model level, many methods fail to fully combine the advantages of physical laws and data-driven models. Purely data-driven models may produce physically unreasonable predictions outside the training data distribution; while traditional numerical models based on physical equations, limited by computational efficiency and parameterization schemes, struggle to efficiently achieve rapid, high-frequency update forecasts at scales from hundreds of meters to kilometers. Furthermore, existing methods are relatively weak in adaptive model updates and response mechanisms to rapid changes in weather systems. When encountering rapidly changing weather processes such as severe convection and intense turbulence, the performance of fixed prediction models will rapidly decline. The lack of effective online performance monitoring and model-triggered update mechanisms makes it impossible to guarantee the continuous reliability of prediction results.

[0003] In summary, existing technologies fall short in achieving high-precision, high-resolution, physically consistent, and adaptively updated real-time predictions of low-altitude wind fields. Therefore, there is an urgent need for an innovative method that can efficiently fuse multi-source observation data, accurately reconstruct the three-dimensional turbulent field, utilize physical information machine learning models for wind field prediction, and possess online monitoring and adaptive retraining capabilities. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and related device for low-altitude turbulent field reconstruction and adaptive wind field prediction. This method and related device can efficiently fuse multi-source observation data, accurately reconstruct the three-dimensional turbulent field, and use physical information machine learning models to predict wind speed fields.

[0005] To achieve the above objectives, this invention discloses a method for low-altitude turbulent field reconstruction and adaptive wind field prediction, comprising: Acquire multi-source meteorological data, preprocess the multi-source meteorological data, and then extract features; The extracted features are input into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field. The three-dimensional wind speed field is input into a sequence prediction model based on a combination of physical information neural network and long short-term memory network to predict the three-dimensional wind speed field at future times.

[0006] Furthermore, multi-source meteorological data are acquired through wind profiler radar, automatic weather stations, and Doppler weather radar.

[0007] Furthermore, the preprocessing process for the multi-source meteorological data is as follows: For the multi-source meteorological data, outlier detection is performed based on the 3σ criterion, missing value imputation is performed using Kriging interpolation, and spatiotemporal alignment is performed considering the sampling characteristics of the equipment.

[0008] Furthermore, the multi-source data fusion framework based on multi-output Gaussian processes uses the Matern kernel function to capture spatial correlations, where the kernel function length scale l is set to 0.1-10 km, and the process noise covariance is set to 1×10⁻⁶. -6 -1×10 -4 The measurement noise covariance was set to 0.01-0.1; the three-dimensional wind speed components were inverted by solving the Doppler weather radar data using at least three devices simultaneously.

[0009] Furthermore, the sequence prediction model based on the combination of physical information neural network and long short-term memory network is a PINN-LSTM architecture, in which the physical information neural network introduces three-dimensional incompressible Navier-Stokes equations and continuity equations as constraints.

[0010] Furthermore, PINN enhances the rationality and extrapolation stability of the model by introducing physical constraints. The PINN physical equations are set as three-dimensional incompressible Navier-Stokes equations and continuity equations. No-slip boundary conditions are used near the ground. The loss function consists of three parts: data loss, physical loss, and boundary loss. The weights are dynamically adjusted using adaptive weights based on the proportion of loss.

[0011] Furthermore, it also includes: calculating the vertical profile of turbulence intensity based on physical models and machine learning.

[0012] This invention discloses a low-altitude turbulent field reconstruction and adaptive wind field prediction system, comprising: The acquisition module is used to acquire multi-source meteorological data, preprocess the multi-source meteorological data, and then extract features. The fusion module is used to input the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field. The prediction module is used to input the three-dimensional wind speed field into a sequence prediction model based on a combination of physical information neural network and long short-term memory network to predict the three-dimensional wind speed field at future times.

[0013] The present invention discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the low-altitude turbulence field reconstruction and adaptive wind field prediction method.

[0014] The present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the low-altitude turbulence field reconstruction and adaptive wind field prediction method.

[0015] The present invention has the following beneficial effects: In practical operation, the low-altitude turbulence field reconstruction and adaptive wind field prediction method and related device of the present invention acquire multi-source meteorological data, preprocess the multi-source meteorological data, extract features, and input the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field. This efficiently fuses multi-source observation data, accurately reconstructs the three-dimensional turbulence field, and inputs the three-dimensional wind speed field into a sequence prediction model based on a combination of physical information neural network and long short-term memory network to predict the three-dimensional wind speed field at future times. Thus, wind speed field prediction is performed using a physical information machine learning model, which is highly practical. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] 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, not all, of the embodiments of the present invention. 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.

[0019] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0020] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0021] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0022] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0023] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0024] 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0026] Example 1 refer to Figure 1 The low-altitude turbulent field reconstruction and adaptive wind field prediction method of the present invention includes the following steps: 1) Acquire multi-source meteorological data, preprocess the multi-source meteorological data, and then extract features; Multi-source meteorological data acquired from wind profiler radar, automatic weather stations, and Doppler weather radar were collaboratively preprocessed and feature extracted. Specifically, a strict data quality control process was first implemented for raw data in different formats (including CSV and Excel formats), identifying and removing common radar anomalies (such as invalid values ​​like -9999 and -9998), and then based on 3D... σ Outlier detection is performed to ensure data reliability. For missing values, Kriging interpolation is used for imputation. Considering the sampling characteristics and spatiotemporal resolution differences of different devices, a three-dimensional linear interpolation algorithm is used to achieve spatiotemporal alignment of the data. The specific interpolation formula is as follows:

[0027] in, w i = 1 / d i 2 , d iThis represents the spatial distance from the interpolation point to the observation point. The interpolation search radius can be set to 5-10 kilometers, and the minimum number of valid points is at least 3.

[0028] To ensure the spatial validity of the data, the observation data is limited to latitude and longitude, and the extraction of key meteorological elements covers the three-dimensional wind speed components at different altitudes provided by the wind profiler radar. u , v , w The time resolution is set to 6 minutes, and the vertical resolution can be set to 60-240 meters; the ground automatic weather station provides ground wind speed, wind direction, temperature and air pressure, with a time resolution of 10 minutes and a radial coverage range of 5-10 kilometers; the Doppler weather radar provides radial wind speed and spectral width, with a time resolution of 6 minutes and a spatial coverage range of hundreds of kilometers.

[0029] 2) Calculate the vertical profile of turbulence intensity based on physical models and machine learning; A physics-guided machine learning architecture is constructed to calculate the vertical profile of turbulent kinetic energy (TKE). First, three-dimensional velocity vectors at different altitudes are obtained using wind profiler radar, and the turbulent kinetic energy at discrete points at different altitudes under the same latitude and longitude is calculated. Based on this, the XGBoost machine learning algorithm is introduced to construct a turbulence intensity prediction model. The TKE calculated by the physical model is used as the training objective, and the altitude, three-dimensional wind speed components, and their vertical gradients from the wind profiler radar data are used as inputs. The recommended settings for key hyperparameters are: maximum tree depth of 3-10; learning rate of 0.01-0.4; and subtree sampling ratio of 0.5-1.0. Automatic tuning is performed using a Bayesian optimization algorithm, with an iteration count typically set to 50-100. K-fold cross-validation is used during training, and the mean squared error and coefficient of determination are used as evaluation metrics to ensure the model's generalization ability. Based on the real-time calculated vertical TKE profile, it is monitored whether there are local TKE values ​​exceeding a preset threshold. σ If the condition is exceeded (e.g., a severe convective weather event or drastic wind field change), it is considered a localized severe convective weather event, triggering the system retraining process (at fixed intervals during periods of continuous TKE exceedance). T (0 triggers) to enable the model to better adapt to drastic changes in weather systems.

[0030] 3) Reconstruct the three-dimensional wind field based on multi-source data fusion; A multi-source data fusion framework based on multi-output Gaussian processes unifies wind speed observation data from different sources and at different resolutions into a single computational system. Wind profiler radar provides precise vertical three-dimensional wind speeds, automatic weather stations provide near-surface horizontal wind speed constraints, and Doppler weather radar obtains the three-dimensional wind speed components through simultaneous solutions from multiple devices (at least three). Wind field inversion from the Doppler radar is achieved by solving the following system of linear equations:

[0031] in, Radial wind speed, It is the azimuth angle. The angle of elevation.

[0032] In multi-output Gaussian process modeling, data source identifiers are introduced as an additional input dimension, and the Matern kernel function is used to capture spatial correlations. A data reshaping method is employed to transform the multi-output problem into a single-output problem, i.e., the Gaussian process model is trained separately for each output dimension. Recommended settings for key parameters are: kernel function length scale. The range can be set to 0.1-10 km; the process noise covariance can be set to 1×10. -6 -1×10 -4 The measurement noise covariance can be set to 0.01-0.1. Based on the reconstructed three-dimensional wind speed field, turbulent kinetic energy (TKE) is used as the turbulence intensity index. The regional average wind speed is calculated using a moving average, and the moving window size is set to the order of hundreds of meters as needed. This method can effectively fuse multi-source heterogeneous data to generate a continuous and consistent three-dimensional turbulent field, providing reliable three-dimensional wind field distribution information for subsequent wind field time extrapolation.

[0033] 4) Temporal extrapolation prediction of wind fields based on physical information neural networks; A sequence prediction model based on a combination of physical information neural network and long short-term memory network is constructed to learn and extrapolate the spatiotemporal evolution of the three-dimensional wind field. The historical three-dimensional wind speed field reconstructed in step 3) is used as input, and the forward-predicted three-dimensional wind speed field at future moments is used as output.

[0034] The sequence prediction model employs a PINN-LSTM architecture. The LSTM part learns the spatiotemporal historical patterns of wind speed field evolution, while the PINN part enhances the model's rationality and extrapolation stability by introducing physical constraints. The PINN physical equations are set as three-dimensional incompressible Navier-Stokes equations and continuity equations, with no-slip boundary conditions applied near the ground. The loss function consists of three parts: data loss, physical loss, and boundary loss, with weights dynamically adjusted using adaptive weights based on the loss proportion. An attention mechanism is introduced into the network structure, enabling the model to adaptively focus on historical moments most relevant to the prediction target. Recommended settings for key hyperparameters are: hidden layer dimension 64-512; number of network layers 3-6; dropout rate 0.1-0.3. These parameters need to be determined through cross-validation to achieve optimal configuration. During the training phase, a time sliding window method is used to construct training samples, with the input sequence length set to... n For each time step, a sequence length of 0.3 is recommended. nEach time step. To improve the model's generalization ability and robustness, data augmentation techniques are introduced during training, namely adding Gaussian noise (noise level recommended to be set to 0.01) and time warp (warp factor recommended to be set to 0.1), thereby avoiding overfitting to some extent.

[0035] Based on the predicted future wind speed field, the corresponding predicted turbulent kinetic energy (TKE) is calculated and stored. After time has passed, this predicted TKE is compared with the actual TKE value calculated from the observed data in step 2), and its mean square error is calculated. If this mean square error exceeds a set threshold... σ 2. At a certain time step, it is indicated that the current wind field evolution pattern has deviated significantly from the historical pattern learned by the model, and the prediction reliability has decreased. The system triggers the model retraining process starting from step 1) to update the model parameters using the latest data to adapt to the changes in wind field characteristics.

[0036] 5) Real-time updates and continuous monitoring; The system periodically repeats the entire process from step 1) to step 4). The update frequency is set to 5-30 minutes according to the actual needs of the low-altitude aircraft. The sliding time window mechanism maintains the timeliness of the data and ensures that the system can respond to rapid changes in weather conditions in a timely manner.

[0037] System status monitoring is divided into two aspects: data quality assessment: the reliability of input data is ensured by calculating the data validity ratio and consistency check. When the data quality is lower than the predetermined threshold, data re-collection or supplementary interpolation processing is triggered; anomaly handling mechanism: when data anomalies or training failures are detected, the system automatically switches to historical wind fields or prompts manual intervention. This closed-loop architecture enables the system to have good adaptability and robustness, and can provide reliable and high-precision wind field predictions for common application scenarios.

[0038] Example 2 The low-altitude turbulence field reconstruction and adaptive wind field prediction system of the present invention includes: The acquisition module is used to acquire multi-source meteorological data, preprocess the multi-source meteorological data, and then extract features. The fusion module is used to input the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field. The prediction module is used to input the three-dimensional wind speed field into a sequence prediction model based on a combination of physical information neural network and long short-term memory network to predict the three-dimensional wind speed field at future times.

[0039] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0040] Example 3 A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a low-altitude turbulence field reconstruction and adaptive wind field prediction method. For example, the method includes: acquiring multi-source meteorological data; preprocessing the multi-source meteorological data; extracting features; inputting the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and a turbulence field; and inputting the three-dimensional wind speed field into a sequence prediction model based on a combination of a physical information neural network and a long short-term memory network to predict the three-dimensional wind speed field at future times. The memory may include main memory, such as high-speed random access memory, or it may also include non-volatile memory, such as at least one disk storage device. The processor, network interface, and memory are interconnected via an internal bus, which can be an industry standard architecture bus, a peripheral component interconnection standard bus, an extended industry standard architecture bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory stores the program; specifically, the program may include program code, which includes computer operation instructions. Memory can include main memory and non-volatile memory, and provides instructions and data to the processor.

[0041] Example 4 A computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the low-altitude turbulence field reconstruction and adaptive wind field prediction method. For example, the method includes: acquiring multi-source meteorological data; preprocessing the multi-source meteorological data; extracting features; inputting the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field; and inputting the three-dimensional wind speed field into a sequence prediction model based on a combination of a physical information neural network and a long short-term memory network to predict the three-dimensional wind speed field at future times. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0042] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0044] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0045] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0046] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0047] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0048] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields, characterized in that, include: Acquire multi-source meteorological data, preprocess the multi-source meteorological data, and then extract features; The extracted features are input into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field. The three-dimensional wind speed field is input into a sequence prediction model based on a combination of physical information neural network and long short-term memory network to predict the three-dimensional wind speed field at future times.

2. The method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields according to claim 1, characterized in that, Multi-source meteorological data are acquired through wind profiler radar, automatic weather stations, and Doppler weather radar.

3. The method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields according to claim 1, characterized in that, The preprocessing process for the multi-source meteorological data is as follows: For the multi-source meteorological data, outlier detection is performed based on the 3σ criterion, missing value imputation is performed using Kriging interpolation, and spatiotemporal alignment is performed considering the sampling characteristics of the equipment.

4. The method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields according to claim 1, characterized in that, The multi-source data fusion framework based on a multi-output Gaussian process uses the Matern kernel function to capture spatial correlations, where the kernel function length scale l is set to 0.1-10 km and the process noise covariance is set to 1×10⁻⁶. -6 -1×10 -4 The measurement noise covariance was set to 0.01-0.1; the three-dimensional wind speed components were inverted by solving the Doppler weather radar data using at least three devices simultaneously.

5. The method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields according to claim 1, characterized in that, The sequence prediction model based on the combination of physical information neural network and long short-term memory network is a PINN-LSTM architecture, in which the physical information neural network introduces three-dimensional incompressible Navier-Stokes equations and continuity equations as constraints.

6. The method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields according to claim 5, characterized in that, PINN enhances the rationality and extrapolation stability of the model by introducing physical constraints. The PINN physical equations are set as three-dimensional incompressible Navier-Stokes equations and continuity equations. No-slip boundary conditions are used near the ground. The loss function consists of three parts: data loss, physical loss, and boundary loss. The weights are dynamically adjusted using adaptive weights based on the proportion of loss.

7. The method for reconstructing low-altitude turbulent flow fields and predicting adaptive wind fields according to claim 1, characterized in that, Also includes: The vertical profile of turbulence intensity is calculated based on physical models and machine learning.

8. A low-altitude turbulent field reconstruction and adaptive wind field prediction system, characterized in that, include: The acquisition module is used to acquire multi-source meteorological data, preprocess the multi-source meteorological data, and then extract features. The fusion module is used to input the extracted features into a multi-source data fusion framework based on a multi-output Gaussian process to obtain a three-dimensional wind speed field and turbulence field. The prediction module is used to input the three-dimensional wind speed field into a sequence prediction model based on a combination of physical information neural network and long short-term memory network to predict the three-dimensional wind speed field at future times.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the low-altitude turbulent field reconstruction and adaptive wind field prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the low-altitude turbulent field reconstruction and adaptive wind field prediction method as described in any one of claims 1-7.

Images