A method and device for predicting turbulence type based on wind measurement radar

Abstract

The application provides a turbulence type prediction method and device based on a wind detection radar. The turbulence type prediction method comprises the following steps: obtaining a pre-trained turbulence prediction model and circulation prediction data. The turbulence prediction model is trained based on historical turbulence kinetic energy budget parameters which are inversely calculated based on historical wind speed data and historical circulation data of a circulation database. The circulation prediction data comprises several circulation factors, such as sea level pressure, 850-hPa potential height, 925-hPa zonal wind, 925-hPa meridional wind and 850-hPa air temperature. The turbulence type prediction result is determined according to the predicted turbulence kinetic energy budget parameters and classification rules. The predicted turbulence kinetic energy budget parameters comprise buoyancy production, shear production and dissipation rate. Compared with a single index, the predicted turbulence kinetic energy budget parameters are more conducive to reflecting the balance between turbulence generation and dissipation, and the difference influence of the boundary layer at different times.

Description

Technical Field

[0001] This application relates to the field of meteorology, and in particular to a method and apparatus for predicting turbulence types based on wind-measuring radar. Background Technology

[0002] Currently, the application of wind-measuring radar mainly focuses on real-time observation and post-event identification to monitor and warn of ongoing turbulence events. However, it has the following shortcomings: First, it lacks the ability to provide advance forecasting, only monitoring turbulence events that have already occurred, making it difficult to meet the pre-event prevention and control needs of scenarios such as low-altitude flight and wind power operation and maintenance; Second, when classifying turbulence events by comparing them with fixed thresholds based on single indicators such as wind speed, it is easily affected by the physical balance between turbulence generation and dissipation, and the boundary layer differences between daytime and nighttime turbulence-dominant factors. The classification results are coarse, with low matching degree with actual low-altitude turbulence risks and low accuracy. Summary of the Invention

[0003] This application provides a method and apparatus for predicting turbulence type based on wind-measuring radar, to solve at least one problem existing in related technologies. The technical solution is as follows: In a first aspect, embodiments of this application provide a method for predicting turbulence type based on wind-measuring radar, including: A pre-trained turbulence prediction model is obtained. The turbulence prediction model is trained based on the historical turbulence kinetic energy budget parameters inverted from the historical wind speed data of the wind measuring radar and the historical circulation data of the circulation database. It represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation data. Obtain circulation forecast data and input the circulation forecast data into the turbulence prediction model to obtain the predicted turbulence kinetic energy balance parameters; Based on the predicted turbulence kinetic energy balance parameters and classification rules, the turbulence type prediction results are determined. The circulation forecast data includes several types of circulation factors, such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature; the predicted turbulent kinetic energy balance parameters include buoyancy generation, shear generation, and dissipation rate. The prediction results for determining the turbulence type based on the predicted turbulence kinetic energy balance parameters and classification rules include: Determine the time period to which the circulation forecast data corresponds to the predicted turbulent kinetic energy balance parameters; If the time period is daytime, the buoyancy generation, shear generation, and dissipation rate are classified using the daytime classification threshold to determine the turbulence type prediction result; If the time period is nighttime, the buoyancy generation, shear generation, and dissipation rate are classified using the nighttime classification threshold to determine the turbulence type prediction result.

[0004] In one implementation, if the time period is daytime, the buoyancy generation items, shear generation items, and dissipation rate are classified using a daytime classification threshold to determine the turbulence type prediction result, including: If the time period is daytime, the shear generation term is compared with the first threshold and the second threshold to obtain the first comparison result, and the absolute value of the buoyancy generation term is compared with the third threshold and the fourth threshold to obtain the second comparison result; The summation result is determined by the sum of the absolute values ​​of the shear generation term and the buoyancy generation term. The third comparison result is obtained based on the dissipation rate and the summation result. Based on the first comparison result, the second comparison result, and the third comparison result, the turbulence type prediction result is determined.

[0005] In one implementation, determining the turbulence type prediction result based on the first comparison result, the second comparison result, and the third comparison result includes: When the first comparison result is that the shear generation term is less than the first threshold, the second comparison result is that the absolute value of the buoyancy generation term is less than the third threshold, and the third comparison result is that the dissipation rate is greater than or equal to the summation result, the turbulence type prediction result is determined to be weak turbulence. When the first comparison result is that the shear generation term is greater than or equal to the first threshold and less than the second threshold, the second comparison result is that the absolute value of the buoyancy generation term is greater than or equal to the third threshold and less than the fourth threshold, and the third comparison result is that the dissipation rate is greater than the sum of the preset ratio and less than the sum of the results, the turbulence type prediction result is determined to be medium turbulence. When the first comparison result is that the shear generation term is greater than or equal to the second threshold, the second comparison result is that the absolute value of the buoyancy generation term is greater than or equal to the fourth threshold, and the third comparison result is that the dissipation rate is less than or equal to the summation result of the preset ratio, the predicted turbulence type is determined to be strong turbulence.

[0006] In one implementation, the pre-trained turbulence prediction model is obtained through the following steps: Historical wind speed data from the wind-measuring radar and historical circulation data from the circulation database are acquired. Based on the historical wind speed data, historical turbulent kinetic energy budget parameters are calculated. These parameters include buoyancy generation, shear generation, dissipation rate, and turbulent transport. Factor extraction is performed on historical circulation data to obtain historical circulation factors. The types of historical circulation factors include those in the circulation forecast data. The historical turbulent kinetic energy budget parameters and historical circulation factors are normalized. Using a preset time matching window, the normalized historical turbulent kinetic energy budget parameters and the normalized historical circulation factors are matched and bound to obtain several bound datasets. Based on several bound datasets, a pre-trained turbulence prediction model is trained to obtain a pre-trained turbulence prediction model.

[0007] In one implementation, the step of extracting factors from historical circulation data to obtain historical circulation factors includes: Based on the preset time matching window, the historical circulation data is divided into several circulation data subsets, and the sea level pressure, 850hPa geopotential height, 925hPa zonal wind, 925hPa meridional wind and 850hPa air temperature of each circulation data subset are extracted as the first circulation factor. By using the factor contribution rate algorithm, the first circulation factor of each circulation data subset is screened to obtain the second circulation factor retained for each circulation data subset; Multicollinearity removal is performed on the second circulation factor retained in each circulation data subset to obtain the historical circulation factor retained in each circulation data subset.

[0008] In one implementation, the first circulation factor of each circulation data subset is filtered using a factor contribution rate algorithm to obtain the second circulation factor retained for each circulation data subset: Using the factor contribution rate algorithm, principal component analysis is performed on each first circulation factor in each circulation data subset to determine the variance contribution rate corresponding to each first circulation factor in each circulation data subset. The variance contribution rate is compared with the contribution rate threshold. The first circulation factor with a variance contribution rate less than the contribution rate threshold is removed, and the second circulation factor retained in each circulation data subset is obtained.

[0009] In one implementation, the process of performing multicollinearity removal on the second circulation factor retained in each subset of circulation data to obtain the historical circulation factor retained in each subset of circulation data includes: Based on the second circulation factors retained in the subset of circulation data, determine the correlation coefficients between each second circulation factor and the variance inflation factor for each second circulation factor; The variance inflation factor is compared with the inflation threshold, and the second circulation factor whose variance inflation factor is less than or equal to the inflation threshold is retained to obtain the candidate circulation factor. The absolute value of the correlation coefficient corresponding to the candidate circulation factor is compared with the correlation threshold. Candidate circulation factors whose absolute value of the correlation coefficient is less than or equal to the correlation threshold are retained. When the absolute value of the correlation coefficient corresponding to the candidate circulation factor is greater than the correlation threshold, the candidate circulation factors to be retained are determined based on the priority of the circulation factor type or the data quality of the candidate circulation factors. Based on all the retained candidate circulation factors, the historical circulation factors retained for each circulation data subset are obtained.

[0010] In one implementation, the preset model includes several original models; the preset model is trained based on several bound datasets to obtain a pre-trained turbulence prediction model. Based on several bound datasets, each original model is trained to obtain several candidate models; Each candidate model is assigned a pre-defined evaluation index, and a pre-trained turbulence prediction model is selected from several candidate models based on the priority of the pre-defined evaluation index. or, Each candidate model is assigned a normalized preset evaluation index. The normalized preset evaluation index and its corresponding weight are weighted and summed to obtain the model score of each candidate model. Based on the model scores of the candidate models, a pre-trained turbulence prediction model is selected from several candidate models.

[0011] Secondly, embodiments of this application provide a turbulence type prediction device based on wind-measuring radar, comprising: The acquisition module is used to acquire the pre-trained turbulence prediction model. The turbulence prediction model is trained based on the historical turbulence kinetic energy budget parameters inverted from the historical wind speed data of the wind measuring radar and the historical circulation data of the circulation database. It represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation data. The prediction module is used to acquire circulation forecast data, input the circulation forecast data into the turbulence prediction model, and obtain the predicted turbulence kinetic energy balance parameters. The classification module is used to determine the turbulence type prediction result based on the predicted turbulence kinetic energy balance parameters and classification rules; The circulation forecast data includes several types of circulation factors, such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature; the predicted turbulent kinetic energy balance parameters include buoyancy generation, shear generation, and dissipation rate. The prediction results for determining the turbulence type based on the predicted turbulence kinetic energy balance parameters and classification rules include: Determine the time period to which the circulation forecast data corresponds to the predicted turbulent kinetic energy balance parameters; If the time period is daytime, the buoyancy generation, shear generation, and dissipation rate are classified using the daytime classification threshold to determine the turbulence type prediction result; If the time period is nighttime, the buoyancy generation, shear generation, and dissipation rate are classified using the nighttime classification threshold to determine the turbulence type prediction result.

[0012] In one embodiment, the turbulence type prediction device based on wind-measuring radar further includes a training module for: Historical wind speed data from the wind-measuring radar and historical circulation data from the circulation database are acquired. Based on the historical wind speed data, historical turbulent kinetic energy budget parameters are calculated. These parameters include buoyancy generation, shear generation, dissipation rate, and turbulent transport. Factor extraction is performed on historical circulation data to obtain historical circulation factors. The types of historical circulation factors include those in the circulation forecast data. The historical turbulent kinetic energy budget parameters and historical circulation factors are normalized. Using a preset time matching window, the normalized historical turbulent kinetic energy budget parameters and the normalized historical circulation factors are matched and bound to obtain several bound datasets. Based on several bound datasets, a pre-trained turbulence prediction model is trained to obtain a pre-trained turbulence prediction model.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a processor and a memory, wherein the memory stores instructions that are loaded and executed by the processor to implement the methods in any of the above-described embodiments.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed, implements the methods in any of the above-described embodiments.

[0015] The beneficial effects of the above technical solution include at least the following: By acquiring a pre-trained turbulence prediction model and circulation forecast data, the turbulence prediction model is trained based on historical turbulence kinetic energy budget parameters inverted from historical wind speed data from a wind-measuring radar and historical circulation data from a circulation database. It characterizes the mapping relationship between historical turbulence kinetic energy budget parameters and historical circulation data. Therefore, only circulation forecast data needs to be input to predict and determine the turbulence kinetic energy budget parameters in advance. The circulation forecast data includes several types of parameters such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature. The circulation factor helps improve prediction accuracy. Based on the predicted turbulent kinetic energy budget parameters and classification rules, the turbulence type prediction results are determined. The predicted turbulence kinetic energy budget parameters include buoyancy generation, shear generation, and dissipation rate. Compared with a single index, these parameters are more effective in reflecting the balance between turbulence generation and dissipation, as well as the influence of boundary layer differences at different times, thus improving prediction accuracy. By distinguishing between daytime and nighttime periods, and using targeted daytime and nighttime classification thresholds respectively, the buoyancy generation, shear generation, and dissipation rate are classified, which helps improve the accuracy of turbulence type prediction results.

[0016] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, these aspects, embodiments, and features will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0017] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0018] Figure 1 This is a schematic flowchart illustrating the steps of a turbulence type prediction method based on wind-measuring radar according to an embodiment of this application; Figure 2 This is a structural block diagram of a turbulence type prediction device based on wind-measuring radar according to an embodiment of this application; Figure 3 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0019] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0020] Reference Figure 1 The flowchart illustrates a method for predicting turbulence type based on wind-measuring radar according to an embodiment of this application. This method may include at least steps S100-S400: S100: Obtain the pre-trained turbulence prediction model.

[0021] Among them, the turbulence prediction model is trained based on the historical turbulence kinetic energy budget parameters inverted from the historical wind speed data of the wind measuring radar and the historical circulation data of the circulation database. It represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation data, and more specifically represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation factors of the historical circulation data, so as to realize the continuous and refined characterization of the turbulence state.

[0022] Alternatively, the wind-measuring radar can be a Doppler radar or a wind-measuring lidar, or other radar equipment that can be used to measure wind fields.

[0023] S200. Obtain circulation forecast data and input the circulation forecast data into the turbulence prediction model to obtain the predicted turbulence kinetic energy balance parameters.

[0024] S300. Based on the predicted turbulent kinetic energy balance parameters and classification rules, determine the turbulence type prediction results.

[0025] The circulation forecast data includes several types of circulation factors among sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature; the predicted turbulent kinetic energy balance parameters include buoyancy generation, shear generation, dissipation rate, and turbulent transport, with the turbulent transport term not used for classification in this embodiment.

[0026] The technical solution of this application embodiment obtains a pre-trained turbulence prediction model and circulation forecast data. The turbulence prediction model is trained based on historical turbulence kinetic energy balance parameters inverted from historical wind speed data from a wind-measuring radar and historical circulation data from a circulation database. It represents the mapping relationship between historical turbulence kinetic energy balance parameters and historical circulation data. Therefore, only circulation forecast data needs to be input to predict and determine the predicted turbulence kinetic energy balance parameters in advance. The circulation forecast data includes sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature. The inclusion of several circulation factors helps improve prediction accuracy. Based on the predicted turbulent kinetic energy budget parameters and classification rules, the turbulence type prediction results are determined. The predicted turbulence kinetic energy budget parameters include buoyancy generation, shear generation, and dissipation rate, which are more effective than single indicators in reflecting the balance between turbulence generation and dissipation, as well as the influence of boundary layer differences at different times, thus improving prediction accuracy. By distinguishing between daytime and nighttime periods, and using targeted daytime and nighttime classification thresholds respectively, the buoyancy generation, shear generation, and dissipation rate are classified, which helps improve the accuracy of turbulence type prediction results.

[0027] In one implementation, the pre-trained turbulence prediction model in step S100 is trained through the following steps: S101. Obtain historical wind speed data from the wind-measuring radar and historical circulation data from the circulation database, and calculate the historical turbulent kinetic energy balance parameters based on the historical wind speed data.

[0028] Optionally, historical wind speed data (height based on actual settings, e.g., 0-1000m) is measured using a wind-measuring radar over a past historical time interval, and corresponding historical circulation data for the same time interval is obtained from a circulation database. This covers low-level circulation characteristics and corresponding turbulence states under different seasons, day and night times, and various weather circulation backgrounds. This data is then used to allow the model to autonomously learn the nonlinear mapping relationship between historical turbulent kinetic energy budget parameters and historical circulation data. In one implementation, an effective height threshold can be set for quality control, for example, 500m. When the effective detection height of the wind-measuring radar in a single instance is less than 500m, that portion of the historical wind speed data is deleted. This is because precipitation, dense fog, and hardware failures can all cause attenuation of the laser detection signal, leading to a decrease in the effective detection height, thus avoiding impacting subsequent training results.

[0029] For example, the circulation database is ERA5, the fifth-generation global atmospheric reanalysis dataset released by the European Centre for Medium-Range Weather Forecasts (ECMWF). It should be noted that historical turbulent kinetic energy budget parameters, including buoyancy generation, can be calculated from historical wind speed data. Shearing generation items Dissipation rate and turbulent transport terms Satisfying the turbulence trend term The turbulent kinetic energy budget equation can be inverted using existing publicly available methods without specific limitations. For example, the method described in the patent with authorization publication number CN119575516B can be employed. Specifically, the turbulent kinetic energy budget equation is used to constrain the relationship between historical turbulent kinetic energy budget parameters during the training process, as follows:

[0030] S102. Extract factors from historical circulation data to obtain historical circulation factors.

[0031] The types of historical circulation factors include those found in the circulation forecast data. For example, if the circulation forecast data includes five types of circulation factors—sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa temperature—then the types of historical circulation factors will include all five types. If the circulation forecast data includes four types of circulation factors—850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa temperature—then the types of historical circulation factors will include all four types. This ensures that the types of circulation factors in the input circulation forecast data are the same as those included in the historical circulation factors. The turbulence prediction model is trained using these types of circulation factors.

[0032] Optionally, S102 includes S1021-S1022: S1021. According to the preset time matching window, the historical circulation data is divided into several circulation data subsets, and the sea level pressure, 850hPa geopotential height, 925hPa zonal wind, 925hPa meridional wind and 850hPa air temperature of each circulation data subset are extracted as the first circulation factor.

[0033] For example, the preset time matching window is 15 minutes, and the historical circulation data is divided into several circulation data subsets at 15-minute intervals.

[0034] It should be noted that, in this embodiment of the application, five types of circulation factors are used as examples, including sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature. Therefore, the sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature of each circulation data subset are extracted as the first circulation factor. In other embodiments, there can be two, three, or four types, and in this case, the circulation factors of two, three, or four types are extracted as the first circulation factor.

[0035] The selection of the 850hPa and 925hPa isobaric surfaces was made to improve accuracy, taking into account the 0-1000m detection range, atmospheric vertical structure, and turbulence formation mechanism. The specific basis for this selection is as follows: (1) High matching: In the standard atmosphere, the 925hPa isobaric surface corresponds to an altitude of about 700-800m, and the 850hPa isobaric surface corresponds to an altitude of about 1400-1500m. The two isobaric surfaces just cover the upper and adjacent atmospheric layers within the 0-1000m detection range. They are key transitional layers between the low-altitude boundary layer and the free atmosphere, and can fully characterize the circulation background that affects low-altitude turbulence. (2) Turbulent dynamics source: 925 hPa is located near the top of the boundary layer. The wind field in this layer is the main layer for the occurrence of low-level jet streams and horizontal wind shear, and wind shear is the core driving force of mechanical turbulence; 850 hPa is the core layer of low-level weather system activity. The geopotential height of this layer reflects the position and intensity of large-scale systems such as low pressure, high pressure, troughs and ridges. The temperature of this layer directly determines the distribution of large-scale cold and warm air and the stability of atmospheric stratification, thereby controlling the strength of buoyancy turbulence; (3) Data practicality: 850hPa and 925hPa are standard isobaric surfaces for routine meteorological observation and reanalysis data, with high data integrity and good continuity.

[0036] Furthermore, for the 0-1000m detection range, the accuracy is relatively lower compared to other isobaric surfaces such as 850hPa and 925hPa because: the near-surface 1000hPa isobaric surface is greatly affected by topography, surface friction, and local thermal disturbances, and cannot reflect large-scale circulation; the 700hPa and above high-altitude isobaric surfaces (such as 300hPa and 500hPa) correspond to altitudes exceeding 3000m, far from the 0-1000m detection range, and have a negligible direct impact on low-altitude turbulence, so their introduction would only add redundant information. Therefore, wind speed and temperature from other isobaric surfaces cannot replace the relevant factors of 850hPa and 925hPa to achieve the technical effect of this scheme.

[0037] S1022. Using the factor contribution rate algorithm, the first circulation factor of each circulation data subset is screened to obtain the second circulation factor retained in each circulation data subset.

[0038] First, using the factor contribution rate algorithm, principal component analysis (PCA) is performed on each first circulation factor in each circulation data subset to determine the variance contribution rate corresponding to each first circulation factor in each circulation data subset, namely, the variance contribution rates corresponding to the five first circulation factors: sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature.

[0039] Secondly, the variance contribution rate of each subset of circulation data is compared with the contribution rate threshold, and each subset of circulation data is independently screened. For example, if the contribution rate threshold is 5%, the first circulation factors with variance contribution rates less than the contribution rate threshold are removed. These first circulation factors are redundant features, and the second circulation factors retained in each subset of circulation data are obtained.

[0040] S1023. Perform multicollinearity removal on the second circulation factor retained in each circulation data subset to obtain the historical circulation factor retained in each circulation data subset.

[0041] Multicollinearity refers to a stable linear correlation between two or more circulation factors, which can cause distortion in model parameter estimation and a decrease in generalization ability.

[0042] First, based on the second circulation factors retained in the circulation data subset, the correlation coefficients between each second circulation factor are determined. For example, the correlation coefficient is the Pearson correlation coefficient. At this point, the pairwise correlation coefficients between the retained second circulation factors can be determined. Furthermore, based on the second circulation factors retained in the circulation data subset, the variance inflation factor (VIF) of each second circulation factor can be determined.

[0043] Secondly, the variance inflation factor is compared with the inflation threshold, for example, the inflation threshold is 10 and the correlation threshold is 0.8; the second circulation factors whose variance inflation factor is less than or equal to the inflation threshold of 10 are retained to obtain candidate circulation factors; while the second circulation factors whose variance inflation factor is greater than the inflation threshold of 10 are considered to have serious multicollinearity and need to be removed.

[0044] Then, the absolute values ​​of the correlation coefficients corresponding to the candidate circulation factors are compared with the correlation threshold of 0.8. Candidate circulation factors whose absolute values ​​of correlation coefficients are less than or equal to the correlation threshold of 0.8 are retained. Conversely, those with correlation coefficients greater than the correlation threshold of 0.8 are considered to have strong linear correlations and need to be eliminated. However, since the correlation coefficients are determined based on pairwise relationships between circulation factors, it is necessary to determine which two to eliminate. Specifically, when the absolute value of the correlation coefficient corresponding to a candidate circulation factor is greater than the correlation threshold, two factors can be considered: (1) Priority of circulation factors: For example, the highest priority is set to 925hPa zonal wind and 925hPa meridional wind. The priority of other factors is reduced in turn to 850hPa air temperature, 850hPa geopotential height and sea level pressure. According to this priority, the candidate circulation factors to be retained are determined so as to retain the dominant factor.

[0045] (2) The data quality of candidate circulation factors, for example, the 925hPa zonal wind and the 925hPa meridional wind are both of the highest priority. At this time, the data quality of candidate circulation factors can be determined. The method of determining the data quality can be based on the actual selection and is not specifically limited. For example, the effective data volume (non-empty, non-wild values, etc.) can be divided by the total data volume to determine the data quality. Then, candidate circulation factors with better data quality are retained to determine the retained candidate circulation factors.

[0046] Finally, based on all the retained candidate circulation factors, the historical circulation factors retained for each subset of circulation data are obtained.

[0047] S103. Normalize the historical turbulent kinetic energy budget parameters and historical circulation factors. Using a preset time matching window, match and bind the normalized historical turbulent kinetic energy budget parameters with the normalized historical circulation factors to obtain several bound datasets.

[0048] Optionally, the historical turbulent kinetic energy budget parameters and historical circulation factors are normalized, for example, using the Min-Max normalization method, so that each parameter in the normalized historical turbulent kinetic energy budget parameters and the normalized historical circulation factors is mapped to the [0,1] interval, eliminating the dimensional differences between different physical quantities. Then, the normalized historical turbulent kinetic energy budget parameters and the normalized historical circulation factors are matched and bound to obtain several bound datasets; it can be understood that the bound datasets are divided at 15-minute time intervals.

[0049] S104. Based on several bound datasets, train the preset model to obtain a pre-trained turbulence prediction model.

[0050] Optionally, it includes S1041, S1042, or S1043: S1041. Based on several bound datasets, train each original model to obtain several candidate models.

[0051] Optionally, several bound datasets can be divided into a certain proportion, such as a training set, validation set, and test set ratio of 7:2:1. Stratified sampling ensures that the proportion of samples from different meteorological scenarios in the three sets of data is consistent with that of the original dataset, avoiding model bias caused by uneven sample distribution and improving the model's generalization ability. During training, the historical turbulent kinetic energy budget parameters in the training set are used as output labels, and the historical circulation factors are used as inputs for training.

[0052] In one implementation, the preset model includes several original models. For example, four original models constructed using four different algorithms—Gradient Boosting Tree (XGBoost, LightGBM), Random Forest, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU)—are used. Other embodiments are not limited to these examples. Tree models excel at uncovering nonlinear relationships between elements, while LSTM and GRU adapt to the evolution patterns of meteorological time-series data. Subsequent step-by-step training and optimization are beneficial for fully fitting the turbulent thermodynamic and dynamic changes corresponding to circulation variations. The model stability, forecast accuracy, and scene generalization ability are far superior to existing technologies. In this embodiment, these four original models are trained according to the training set to obtain four candidate models. It is understood that during training, the hyperparameters of the model are continuously adjusted until the training termination conditions are met, such as the number of iterations reaching a preset number, the loss value of the loss function verified based on the validation set being less than the error threshold, or the loss value no longer decreasing after several consecutive iterations. Based on the hyperparameters that meet the training termination conditions, candidate models are determined.

[0053] S1042. Determine the preset evaluation indicators for each candidate model, and determine the pre-trained turbulence prediction model from several candidate models based on the priority of the preset evaluation indicators.

[0054] Optionally, the preset evaluation metrics include, but are not limited to, at least two of the following: mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R²), and inference time. This application uses all types as an example. Then, the priorities of the preset evaluation metrics can be set in advance, such as MAE, RMSE, R², and inference time from high to low. The optimal candidate model is selected based on the priority. If the preset evaluation metrics of the same priority are the same, the next priority is compared until the optimal candidate model is determined. The optimal candidate model is the pre-trained turbulence prediction model.

[0055] S1043. Determine the normalized preset evaluation indexes corresponding to each candidate model, and perform weighted summation based on the normalized preset evaluation indexes and their corresponding weights to obtain the model score of each candidate model. Based on the model scores of the candidate models, determine the pre-trained turbulence prediction model from several candidate models.

[0056] Optionally, each preset evaluation index is normalized, and a weighted sum is performed based on the normalized preset evaluation index and its corresponding weight (preset separately) to obtain the model score for each candidate model. Then, based on the model scores of the candidate models, the candidate model with the lowest score is selected from several candidate models to be determined as the pre-trained turbulence prediction model. Therefore, this ultimately helps to ensure the prediction accuracy and adaptability of the pre-trained turbulence prediction model.

[0057] In this embodiment, historical wind speed data and historical circulation data are stored locally after acquisition. The local storage can be updated daily based on incremental data, allowing for continuous training and adjustment of the pre-trained turbulence prediction model. For example, when the amount of newly added data (sample size) reaches 5% of the total historical samples, the system automatically initiates an incremental training mechanism, merging the new data into the total sample dataset and re-executing steps S100-S300 to generate an iteratively updated version of the pre-trained turbulence prediction model. Simultaneously, a full-data retraining is conducted quarterly, integrating all historical samples to remodel and replace the original online inference model. This periodic full-data retraining continuously corrects model biases, constantly improving the model's long-term forecast stability and accuracy.

[0058] Optionally, it is also equipped with regular quality spot checks, which start an automated quality inspection program every month to randomly select data samples from local storage and compare them with radiosonde, gradient tower, and conventional ground meteorological data. When the deviation of parameters B, S, D, and Tt exceeds the preset error threshold, the problematic samples are marked, isolated, and removed to continuously ensure the overall quality of data in local storage.

[0059] The method described in this application creatively proposes a dual-source data standardization quality control and time-series matching system. It binds historical turbulent kinetic energy budget parameters derived from historical wind speed data with circulation factors from historical circulation data to enable model training, significantly improving dataset standardization and model trainability. Simultaneously, a 15-minute precise time-series pairing mechanism is established to achieve standardized matching of wind speed and circulation data. Compared to traditional unordered sampling methods, the constructed sample set has a high signal-to-noise ratio, low redundancy interference, and clear physical correspondences, providing a reliable data foundation for high-precision turbulence forecasting modeling. Furthermore, by introducing a factor contribution rate algorithm and multicollinearity removal, standardized feature selection is achieved, ensuring feature effectiveness.

[0060] In one implementation, step S200 involves obtaining circulation forecast data from the same circulation database. This data includes several circulation factors, such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature. The circulation forecast data is then input into a turbulence prediction model to obtain predicted turbulence kinetic energy balance parameters. In some embodiments, these parameters can be further refined using turbulence kinetic energy balance equations. This refinement can also correct extreme values ​​and abrupt changes that violate atmospheric physics laws, and smoothing can be achieved by combining adjacent height profiles, ensuring the physical rationality of the turbulence forecast profile.

[0061] In some implementations, when inputting circulation forecast data, corresponding preprocessing steps such as normalization, factor contribution rate algorithms, and multicollinearity removal can be performed to ensure the accuracy of the prediction. Furthermore, the ERA5 circulation database typically runs the Integrated Forecasting System (IFS) global numerical weather prediction model daily. Using the current global observation data assimilation analysis field as the initial state, it calculates forward by solving atmospheric physical equations (Navier-Stokes equations, thermodynamic equations, radiative transfer equations, etc.) and outputs global gridded meteorological element forecast fields for multiple timeframes (e.g., hourly, up to 10-15 days). Therefore, it includes future circulation forecast data.

[0062] In one implementation, step S300 includes steps S310-S330: S310. Determine the time period to which the circulation forecast data corresponding to the predicted turbulent kinetic energy balance parameters belong.

[0063] Optionally, when acquiring circulation forecast data, the data will include a timestamp, allowing the determination of the time period to which the predicted turbulent kinetic energy balance parameters belong. This time period includes the daytime (08:00–20:00) and the nighttime (20:00–08:00 the following day). It should be noted that when acquiring circulation forecast data, wind speed data for the same time period can be subsequently acquired and the results derived, used for operational verification of the prediction and assessment of prediction accuracy.

[0064] S320. If the time period is daytime, use the daytime classification threshold to classify the buoyancy generation items, shear generation items, and dissipation rate to determine the turbulence type prediction results.

[0065] It should be noted that since buoyancy dominates turbulent energy changes during the daytime and shear dominates turbulent energy changes at night, it is necessary to distinguish specific classification criteria and configure different classification thresholds for the daytime and nighttime periods to ensure the rationality and accuracy of the classification.

[0066] Optionally, if the time period is during the daytime, the buoyancy generation item is classified using the daytime classification threshold. Shearing generation items and dissipation rate The prediction results are classified to determine the turbulence type.

[0067] For example, S320 includes S3201-S3203: S3201. If the time period is daytime, compare the shear generation term with the first threshold and the second threshold to obtain the first comparison result, and compare the absolute value of the buoyancy generation term with the third threshold and the fourth threshold to obtain the second comparison result.

[0068] Optionally, the daytime classification threshold includes a first threshold of 1.0 × 10⁻⁶. -4 The second threshold is 5.0 × 10 -4 The third threshold is 8.0 × 10 -5 The fourth threshold is 3.0 × 10 -4 In other implementations, the specific values ​​can be different and are not specifically limited. If the time period is daytime, the generated items will be cut. Compared with the first threshold of 1.0 × 10⁻⁴ and the second threshold of 5.0 × 10⁻⁴ -4 The comparison is performed to obtain the first comparison result; and the buoyancy generation term is then included. absolute value Compared with the third threshold of 8.0×10 -5 The fourth threshold is 3.0 × 10 -4 By comparing the results, we obtain the second comparison result.

[0069] S3202. Determine the summation result based on the sum of the absolute values ​​of the shear generation term and the buoyancy generation term. Obtain the third comparison result based on the dissipation rate and the summation result.

[0070] Optionally, based on the shearing generated items With buoyancy generation term absolute value The sum of the terms determines the result of the summation. According to the dissipation rate and the summation result The third comparison result was obtained.

[0071] S3203. Based on the first comparison result, the second comparison result, and the third comparison result, determine the turbulence type prediction result.

[0072] 1. When the first comparison result is a shearing term. Less than the first threshold of 1.0 × 10 -4 The second comparison result is the buoyancy generation term. absolute value Less than the third threshold of 8.0 × 10 -5 And the third comparison result is the dissipation rate. Greater than or equal to the summation result: The predicted turbulence type was determined to be weak turbulence, indicating that turbulent energy generation is weak, dissipation is dominant, and the airflow is stable with no obvious vertical disturbance.

[0073] 2. When the first comparison result is a shearing term. Greater than or equal to the first threshold of 1.0 × 10 -4 And less than the second threshold of 5.0 × 10 -4 (i.e., 1.0 × 10) -4 ≤ <5.0×10 -4 The second comparison result is the absolute value of the buoyancy-generating term. Greater than or equal to the third threshold of 8.0 × 10 -5 And less than the fourth threshold of 3.0 × 10 -4 That is (8.0×10 -5 ≤ <3.0×10 -4 The third comparison result is that the dissipation rate is greater than the sum of the preset proportions but less than the sum of the preset proportions (for example, the preset proportion is 0.5, i.e.) < < The predicted turbulence type was determined to be medium turbulence, indicating a dynamic balance between turbulent kinetic energy generation and dissipation, and the presence of persistent small-amplitude airflow disturbances.

[0074] 3. When the first comparison result is a shearing term. Greater than or equal to the second threshold of 5.0 × 10 -4 The second comparison result is the absolute value of the buoyancy-generating term. Greater than or equal to the fourth threshold of 3.0 × 10 -4 Furthermore, the third comparison result is the sum of results where the dissipation rate is less than or equal to a preset ratio, i.e. ≤ The predicted turbulence type is strong turbulence, indicating that shear or buoyancy generates a large amount of turbulent energy, which cannot be dissipated in balance, resulting in severe airflow disturbance and easy low-altitude turbulence.

[0075] It should be noted that if the conditions do not fall under any of the above three categories, it may be due to the complex and changeable weather, making it impossible to determine a specific type. In this case, accuracy is prioritized and no specific classification is output. The system waits for new circulation forecast data to be input for prediction. Once it is determined that one of the above three categories is met, the turbulence type prediction result is output.

[0076] S330. If the time period is nighttime, use the nighttime classification threshold to classify the buoyancy generation items, shear generation items, and dissipation rate to determine the turbulence type prediction result.

[0077] Optionally, if the time period is nighttime, the buoyancy generation item is classified using a nighttime classification threshold. Shearing generation items and dissipation rate The prediction results are classified to determine the turbulence type.

[0078] Similarly, the nighttime classification thresholds also include a first threshold, a second threshold, a third threshold, and a fourth threshold, following the same logic as the daytime thresholds, which will not be repeated here. It can also output predictions for weak, moderate, and strong turbulence types, the difference being the different threshold values. For example, the first threshold is 6.0 × 10⁻⁶. -5 Second threshold 2.0 × 10 -4 The third threshold is 4.0 × 10 -5 The fourth threshold is 1.2 × 10⁻⁶. -4 The values ​​may vary, and this value is just an example and does not constitute a specific limitation.

[0079] The classification rules in this application, based on complete turbulent kinetic energy balance parameters, provide a more scientific basis for judgment and results that better reflect real-world low-altitude conditions. This avoids ignoring the physical balance of turbulence generation, transport, and dissipation, and eliminates deviations caused by day-night differences in threshold values. It relies on buoyancy generation... Shearing generation items and dissipation rate The classification method distinguishes the boundary layer differences between daytime thermal turbulence and nighttime mechanical turbulence, and establishes a quantitative classification standard. The classification results have complete physical mechanisms, clear levels, and fit the actual turbulence characteristics of urban low-altitude airspace, effectively solving the problems of crude traditional classification and high risk misjudgment rate.

[0080] The method described in this application overcomes the technical bottleneck of traditional methods that can only monitor but not predict, enabling advanced prediction driven by the physical mechanism of low-altitude turbulence. This is beneficial for early prevention and standardization in business scenarios such as low-altitude flight, wind power operation and maintenance, and regional environmental monitoring. At the same time, it no longer relies on real-time radar observation of turbulence, but instead extrapolates the evolution of low-altitude turbulence through future circulation fields, truly achieving early prediction of turbulence type and intensity. This fundamentally solves the industry pain point that traditional technologies can only provide post-event warnings and cannot prevent and control events in advance.

[0081] In one implementation, since wind speed data at different altitudes can determine the corresponding turbulent kinetic energy balance parameters, predicted turbulent kinetic energy balance parameters at different altitudes can be predicted. This enables the output and automatic classification of predicted turbulent kinetic energy balance parameters at each altitude and time interval, generating standardized turbulence risk classification products. Furthermore, differentiated alarm prompts and operational guidelines are configured for different turbulence classifications and application scenarios, adapting to business scenarios such as low-altitude flight, wind power operation and maintenance, and regional environmental monitoring. For example: I. Weak Turbulence Alarm level: No alarm General description: Turbulent disturbances are weak, the overall airflow is stable, and the atmospheric vertical mixing capacity is relatively weak.

[0082] 1. Low-altitude flight: Regular weather warnings will be issued, and drones and low-altitude aircraft may fly normally along predetermined routes and at predetermined altitudes.

[0083] 2. Wind power operation and maintenance: Wind turbines maintain normal operation mode, and inspection and maintenance operations are carried out normally.

[0084] 3. Regional environmental monitoring: Atmospheric diffusion conditions are generally normal; monitoring work should be carried out according to the standard procedure.

[0085] II. Medium Turbulence Alert Level: Blue (Caution Warning) General description: There are persistent, minor airflow disturbances, and the vertical mixing of the atmosphere is moderate. There is no substantial safety threat, but monitoring needs to be strengthened.

[0086] 1. Low-altitude flight: Issue flight warnings, recommend reducing flight speed, avoiding the altitude layer where disturbances are concentrated, and prohibiting fine hovering and aerobatic maneuvers.

[0087] 2. Wind power operation and maintenance: The wind turbines are operating normally, and the equipment status is monitored in real time. High-altitude hoisting and high-risk outdoor maintenance operations are suspended.

[0088] 3. Regional environmental monitoring: The atmospheric diffusion conditions are favorable, with a focus on the short-distance diffusion trend of pollutants.

[0089] III. Strong Turbulence Alarm Level: Yellow Risk Alarm General description: Severe airflow disturbances, with obvious turbulence and strong vertical movement, posing a high safety risk.

[0090] 1. Low-altitude flight: Issue a flight risk warning and prohibit aircraft from taking off; equipment that has already taken off should immediately land at the nearest lander or detour around the turbulent area.

[0091] 2. Wind power operation and maintenance: The wind turbines are switched to the reduced load operation mode, all outdoor operation and maintenance work is completely stopped, and personnel are evacuated from high-altitude areas.

[0092] 3. Regional environmental monitoring: The atmospheric diffusion conditions are excellent, so we should increase the frequency of monitoring and assess the range and duration of cross-border transport of pollutants.

[0093] Therefore, this embodiment can automatically generate a complete set of forecast products, including a single-time, full-altitude turbulent kinetic energy budget profile (composed of predicted turbulent kinetic energy budget parameters at various altitudes from 0 to 1000m), turbulence type (weak turbulence, moderate turbulence, strong turbulence), risk level, and corresponding alarm prompts. Each forecast result is labeled with the corresponding turbulence type. Simultaneously, all real-time observation data from wind measuring radars, circulation forecast data, historical circulation data, intermediate model extrapolation results, and final forecast products are backed up locally and in the cloud, and automatically pushed to operational terminals such as low-altitude safety management, atmospheric environment monitoring, and wind energy assessment. This enables the practical application of the entire method, demonstrating strong practicality and scalability. The entire method is standardized and can be automated, widely adaptable to various industry scenarios such as low-altitude flight safety, wind power operation and maintenance, and atmospheric environment monitoring, significantly improving its engineering applicability and operational versatility.

[0094] Reference Figure 2 The diagram illustrates a structural block diagram of a turbulence type prediction device based on wind-measuring radar according to an embodiment of this application. The device may include: The acquisition module is used to acquire the pre-trained turbulence prediction model. The turbulence prediction model is trained based on the historical turbulence kinetic energy budget parameters inverted from the historical wind speed data of the wind measuring radar and the historical circulation data of the circulation database. It represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation data. The prediction module is used to acquire circulation forecast data, input the circulation forecast data into the turbulence prediction model, and obtain the predicted turbulence kinetic energy balance parameters. The classification module is used to determine the turbulence type prediction result based on the predicted turbulence kinetic energy balance parameters and classification rules; The circulation forecast data includes several types of circulation factors, such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature; the predicted turbulent kinetic energy balance parameters include buoyancy generation, shear generation, and dissipation rate. The prediction results for determining the turbulence type based on the predicted turbulence kinetic energy balance parameters and classification rules include: Determine the time period to which the circulation forecast data corresponds to the predicted turbulent kinetic energy balance parameters; If the time period is daytime, the buoyancy generation, shear generation, and dissipation rate are classified using the daytime classification threshold to determine the turbulence type prediction result; If the time period is nighttime, the buoyancy generation, shear generation, and dissipation rate are classified using the nighttime classification threshold to determine the turbulence type prediction result.

[0095] In one embodiment, the turbulence type prediction device based on wind-measuring radar further includes a training module for: Historical wind speed data from the wind-measuring radar and historical circulation data from the circulation database are acquired. Based on the historical wind speed data, historical turbulent kinetic energy budget parameters are calculated. These parameters include buoyancy generation, shear generation, dissipation rate, and turbulent transport. Factor extraction is performed on historical circulation data to obtain historical circulation factors. The types of historical circulation factors include those in the circulation forecast data. The historical turbulent kinetic energy budget parameters and historical circulation factors are normalized. Using a preset time matching window, the normalized historical turbulent kinetic energy budget parameters and the normalized historical circulation factors are matched and bound to obtain several bound datasets. Based on several bound datasets, a pre-trained turbulence prediction model is trained to obtain a pre-trained turbulence prediction model.

[0096] The functions of each module in the device of this application embodiment can be found in the corresponding description in the above method, and will not be repeated here.

[0097] Reference Figure 3 The diagram illustrates a structural block diagram of an electronic device according to an embodiment of this application. The electronic device includes a memory 310 and a processor 320. The memory 310 stores instructions that can be executed on the processor 320. The processor 320 loads and executes these instructions to implement the turbulence type prediction method based on wind-measuring radar described in the above embodiment. The number of memories 310 and processors 320 can be one or more.

[0098] In one embodiment, the electronic device further includes a communication interface 330 for communicating with external devices and exchanging data. If the memory 310, processor 320, and communication interface 330 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0099] Optionally, in a specific implementation, if the memory 310, processor 320 and communication interface 330 are integrated on a single chip, the memory 310, processor 320 and communication interface 330 can communicate with each other through an internal interface.

[0100] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the turbulence type prediction method based on wind-measuring radar provided in the above embodiments.

[0101] This application also provides a chip, which includes a processor for calling and executing instructions stored in a memory, causing a communication device on which the chip is installed to perform the method provided in this application.

[0102] This application also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in the application embodiment.

[0103] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting the Advanced Reduced Instruction Set Computing (RISC) machine (ARM) architecture.

[0104] Further, optionally, the aforementioned memory may include read-only memory and random access memory, and may also include non-volatile random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0105] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0106] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0107] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0108] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0109] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0110] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0111] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0112] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting turbulence type based on wind-measuring radar, characterized in that, include: A pre-trained turbulence prediction model is obtained. The turbulence prediction model is trained based on the historical turbulence kinetic energy budget parameters inverted from the historical wind speed data of the wind measuring radar and the historical circulation data of the circulation database. It represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation data. Obtain circulation forecast data and input the circulation forecast data into the turbulence prediction model to obtain the predicted turbulence kinetic energy balance parameters; Based on the predicted turbulence kinetic energy balance parameters and classification rules, the turbulence type prediction results are determined. The circulation forecast data includes several types of circulation factors, such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature; the predicted turbulent kinetic energy balance parameters include buoyancy generation, shear generation, and dissipation rate. The prediction results for determining the turbulence type based on the predicted turbulence kinetic energy balance parameters and classification rules include: Determine the time period to which the circulation forecast data corresponds to the predicted turbulent kinetic energy balance parameters; If the time period is daytime, the buoyancy generation, shear generation, and dissipation rate are classified using the daytime classification threshold to determine the turbulence type prediction result; If the time period is nighttime, the buoyancy generation, shear generation, and dissipation rate are classified using the nighttime classification threshold to determine the turbulence type prediction result.

2. The turbulence type prediction method based on wind-measuring radar according to claim 1, characterized in that: If the time period is daytime, the buoyancy generation, shear generation, and dissipation rate are classified using a daytime classification threshold to determine the turbulence type prediction results, including: If the time period is daytime, the shear generation term is compared with the first threshold and the second threshold to obtain the first comparison result, and the absolute value of the buoyancy generation term is compared with the third threshold and the fourth threshold to obtain the second comparison result; The summation result is determined by the sum of the absolute values ​​of the shear generation term and the buoyancy generation term. The third comparison result is obtained based on the dissipation rate and the summation result. Based on the first comparison result, the second comparison result, and the third comparison result, the turbulence type prediction result is determined.

3. The turbulence type prediction method based on wind-measuring radar according to claim 2, characterized in that: The determination of the turbulence type prediction result based on the first comparison result, the second comparison result, and the third comparison result includes: When the first comparison result is that the shear generation term is less than the first threshold, the second comparison result is that the absolute value of the buoyancy generation term is less than the third threshold, and the third comparison result is that the dissipation rate is greater than or equal to the summation result, the turbulence type prediction result is determined to be weak turbulence. When the first comparison result is that the shear generation term is greater than or equal to the first threshold and less than the second threshold, the second comparison result is that the absolute value of the buoyancy generation term is greater than or equal to the third threshold and less than the fourth threshold, and the third comparison result is that the dissipation rate is greater than the sum of the preset ratio and less than the sum of the results, the turbulence type prediction result is determined to be medium turbulence. When the first comparison result is that the shear generation term is greater than or equal to the second threshold, the second comparison result is that the absolute value of the buoyancy generation term is greater than or equal to the fourth threshold, and the third comparison result is that the dissipation rate is less than or equal to the summation result of the preset ratio, the predicted turbulence type is determined to be strong turbulence.

4. The turbulence type prediction method based on wind-measuring radar according to any one of claims 1-3, characterized in that: The pre-trained turbulence prediction model is obtained through the following steps: Historical wind speed data from the wind-measuring radar and historical circulation data from the circulation database are acquired. Based on the historical wind speed data, historical turbulent kinetic energy budget parameters are calculated. These parameters include buoyancy generation, shear generation, dissipation rate, and turbulent transport. Factor extraction is performed on historical circulation data to obtain historical circulation factors. The types of historical circulation factors include those in the circulation forecast data. The historical turbulent kinetic energy budget parameters and historical circulation factors are normalized. Using a preset time matching window, the normalized historical turbulent kinetic energy budget parameters and the normalized historical circulation factors are matched and bound to obtain several bound datasets. Based on several bound datasets, a pre-trained turbulence prediction model is trained to obtain a pre-trained turbulence prediction model.

5. The turbulence type prediction method based on wind-measuring radar according to claim 4, characterized in that: The factor extraction from the historical circulation data to obtain historical circulation factors includes: Based on the preset time matching window, the historical circulation data is divided into several circulation data subsets, and the sea level pressure, 850hPa geopotential height, 925hPa zonal wind, 925hPa meridional wind and 850hPa air temperature of each circulation data subset are extracted as the first circulation factor. By using the factor contribution rate algorithm, the first circulation factor of each circulation data subset is screened to obtain the second circulation factor retained for each circulation data subset; Multicollinearity removal is performed on the second circulation factor retained in each circulation data subset to obtain the historical circulation factor retained in each circulation data subset.

6. The turbulence type prediction method based on wind-measuring radar according to claim 5, characterized in that: The first circulation factor of each circulation data subset is filtered using the factor contribution rate algorithm to obtain the second circulation factor retained for each circulation data subset: Using the factor contribution rate algorithm, principal component analysis is performed on each first circulation factor in each circulation data subset to determine the variance contribution rate corresponding to each first circulation factor in each circulation data subset. The variance contribution rate is compared with the contribution rate threshold. The first circulation factor with a variance contribution rate less than the contribution rate threshold is removed, and the second circulation factor retained in each circulation data subset is obtained.

7. The turbulence type prediction method based on wind-measuring radar according to claim 5, characterized in that: The process of removing multicollinearity from the second circulation factor retained in each subset of circulation data yields the historical circulation factor retained in each subset of circulation data, including: Based on the second circulation factors retained in the subset of circulation data, determine the correlation coefficients between each second circulation factor and the variance inflation factor for each second circulation factor; The variance inflation factor is compared with the inflation threshold, and the second circulation factor whose variance inflation factor is less than or equal to the inflation threshold is retained to obtain the candidate circulation factor. The absolute value of the correlation coefficient corresponding to the candidate circulation factor is compared with the correlation threshold. Candidate circulation factors whose absolute value of the correlation coefficient is less than or equal to the correlation threshold are retained. When the absolute value of the correlation coefficient corresponding to the candidate circulation factor is greater than the correlation threshold, the candidate circulation factors to be retained are determined based on the priority of the circulation factor type or the data quality of the candidate circulation factors. Based on all the retained candidate circulation factors, the historical circulation factors retained for each circulation data subset are obtained.

8. The turbulence type prediction method based on wind-measuring radar according to claim 4, characterized in that: The preset model includes several original models; the preset model is trained based on several bound datasets to obtain a pre-trained turbulence prediction model. Based on several bound datasets, each original model is trained to obtain several candidate models; Each candidate model is assigned a pre-defined evaluation index, and a pre-trained turbulence prediction model is selected from several candidate models based on the priority of the pre-defined evaluation index. or, Each candidate model is assigned a normalized preset evaluation index. The normalized preset evaluation index and its corresponding weight are weighted and summed to obtain the model score of each candidate model. Based on the model scores of the candidate models, a pre-trained turbulence prediction model is selected from several candidate models.

9. A turbulence type prediction device based on wind-measuring radar, characterized in that, include: The acquisition module is used to acquire the pre-trained turbulence prediction model. The turbulence prediction model is trained based on the historical turbulence kinetic energy budget parameters inverted from the historical wind speed data of the wind measuring radar and the historical circulation data of the circulation database. It represents the mapping relationship between the historical turbulence kinetic energy budget parameters and the historical circulation data. The prediction module is used to acquire circulation forecast data, input the circulation forecast data into the turbulence prediction model, and obtain the predicted turbulence kinetic energy balance parameters. The classification module is used to determine the turbulence type prediction result based on the predicted turbulence kinetic energy balance parameters and classification rules; The circulation forecast data includes several types of circulation factors, such as sea level pressure, 850 hPa geopotential height, 925 hPa zonal wind, 925 hPa meridional wind, and 850 hPa air temperature; the predicted turbulent kinetic energy balance parameters include buoyancy generation, shear generation, and dissipation rate. The prediction results for determining the turbulence type based on the predicted turbulence kinetic energy balance parameters and classification rules include: Determine the time period to which the circulation forecast data corresponds to the predicted turbulent kinetic energy balance parameters; If the time period is daytime, the buoyancy generation, shear generation, and dissipation rate are classified using the daytime classification threshold to determine the turbulence type prediction result; If the time period is nighttime, the buoyancy generation, shear generation, and dissipation rate are classified using the nighttime classification threshold to determine the turbulence type prediction result.

Images