Method, device, equipment, medium and product for predicting clear-air turbulence of aircraft

Abstract

The present application discloses an aircraft clear-air turbulence prediction method, device, equipment, medium and product, which relates to the field of aviation meteorology. The method includes: obtaining a sample set, where the sample set includes multiple samples, and one sample includes the turbulence observation data and the values of each meteorological data when a sample aircraft flies at the same time and in the same space; calculating the numerical values of each CAT diagnosis index according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnosis value sequence corresponding to the sample; constructing a total data set according to the CAT diagnosis value sequence corresponding to each sample and the turbulence observation data corresponding to each sample; performing data quality control, replacing the turbulence observation data with turbulence labels, and oversampling processing on the total data set in sequence to obtain a data set; training and validating a W-KNN model with the data set to obtain a clear-air turbulence prediction model, which is used to predict the clear-air turbulence of an aircraft to be predicted. The present application can improve the accuracy of the clear-air turbulence prediction result.

Description

Technical Field

[0001] The present application relates to the field of aviation meteorology, and in particular, to an aircraft clear-air turbulence prediction method, device, equipment, medium and product. Background Art

[0002] With the intensification of climate change, the probability of future clear-air turbulence (CAT) occurrence shows an increasing trend. CAT refers to the turbulence that occurs in the absence of significant cloud layers or weather systems, which is composed of vortex airflows of different scales and changes over time. The direction of its turbulence dissipation is from large scales to small scales. CAT is one of the dangerous meteorological phenomena in aviation flight, posing a major threat to the safety of aircraft and easily causing clear-air turbulence, which can easily lead to injuries to passengers and crew members. Therefore, predicting clear-air turbulence is very important, but the existing methods for predicting clear-air turbulence have inaccurate prediction results. Summary of the Invention

[0003] The purpose of the present application is to provide an aircraft clear-air turbulence prediction method, device, equipment, medium and product, which can improve the accuracy of clear-air turbulence prediction results.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] In the first aspect, the present application provides an aircraft clear-air turbulence prediction method, including:

[0006] Obtain a sample set, where the sample set includes multiple samples, and one sample includes the bump observation data and the values of each meteorological data when the sample aircraft flies at the same time and in the same space;

[0007] For any one sample, calculate the values of each CAT diagnostic index according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnostic value sequence corresponding to the sample;

[0008] Construct a total data set according to the CAT diagnostic value sequences corresponding to each sample and the bump observation data corresponding to each sample;

[0009] Perform preprocessing and oversampling processing on the total data set in sequence to obtain a data set; the preprocessing includes performing data quality control and replacing the bump observation data with turbulence labels in sequence;

[0010] Use the data set to train and verify a W-KNN model to obtain a clear-air turbulence prediction model, which is used to predict the clear-air turbulence of the aircraft to be predicted.

[0011] In the second aspect, the present application provides an aircraft clear-air turbulence prediction device, including:

[0012] An acquisition module, configured to acquire a sample set, where the sample set includes a plurality of samples, and one sample includes turbulence observation data and values of respective meteorological data when a sample aircraft flies at the same time and in the same space;

[0013] A CAT diagnosis value calculation module, configured to, for any one sample, calculate values of respective CAT diagnosis indices according to the values of the respective meteorological data corresponding to the sample to obtain a CAT diagnosis value sequence corresponding to the sample;

[0014] A total data set construction module, configured to construct a total data set according to the CAT diagnosis value sequence corresponding to each sample and the turbulence observation data corresponding to each sample;

[0015] A data set determination module, configured to perform preprocessing and oversampling processing on the total data set in sequence to obtain a data set; the preprocessing includes performing data quality control and replacing the turbulence observation data with turbulence labels in sequence;

[0016] A clear-air turbulence prediction model determination module, configured to train and validate a W-KNN model by using the data set to obtain a clear-air turbulence prediction model, where the clear-air turbulence prediction model is used to predict clear-air turbulence of an aircraft to be predicted.

[0017] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the aircraft clear-air turbulence prediction method according to any one of the above.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the aircraft clear-air turbulence prediction method according to any one of the above is implemented.

[0019] In a fifth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the aircraft clear-air turbulence prediction method according to any one of the above is implemented.

[0020] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application:

[0021] The present application provides an aircraft clear-air turbulence prediction method, device, equipment, medium and product. The numerical values of each CAT diagnosis index are calculated according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnosis value sequence corresponding to the sample; a total data set is constructed according to the CAT diagnosis value sequence corresponding to each sample and the turbulence observation data corresponding to each sample; data quality control, replacement of turbulence observation data with turbulence labels, and oversampling processing are successively performed on the total data set to obtain a data set; the data set is used to train and validate a W-KNN model to obtain a clear-air turbulence prediction model, which is used to predict the clear-air turbulence of the aircraft to be predicted. In the traditional method, a turbulence threshold needs to be defined before turbulence prediction, and the turbulence threshold is not unified, mainly depending on climate model data. Moreover, for different weight models, the turbulence intensity predicted by the traditional empirical model or its derived integrated CAT prediction model is also inconsistent. Therefore, the traditional turbulence prediction method is affected by the turbulence threshold and different models, resulting in inaccurate prediction results. For this reason, the International Civil Aviation Organization stipulates that the eddy dissipation rate (EDR) is used as the turbulence measurement index. However, with the challenges related to data quality control and the inherent limitations of physical models, the accuracy of the EDR estimate has also been affected. The clear-air turbulence prediction method proposed by the present invention does not require the application of turbulence thresholds and model data, so it will not be affected by the turbulence threshold and different models, improving the accuracy of the clear-air turbulence prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0023] Figure 1 It is an application environment diagram of an aircraft clear-air turbulence prediction method in an embodiment of the present application;

[0024] Figure 2 It is a flowchart of an aircraft clear-air turbulence prediction method provided in an embodiment of the present application;

[0025] Figure 3 It is a flowchart of predicting an A320 aircraft using the aircraft clear-air turbulence prediction method provided in an embodiment of the present application;

[0026] Figure 4 It is an ROC curve of the W-KNN model validating the A320 test data set;

[0027] Figure 5 It is a structural diagram of a computer device provided in an embodiment of the present application. Detailed implementation manners

[0028] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0029] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0030] Although a variety of CAT diagnostic indices have been developed currently, due to different indices focusing on different factors, the forecast effects of clear-air turbulence in different regions, different seasons, and different causes will also be different. Therefore, it is particularly important to improve the index forecasting algorithm to make it more suitable for predicting clear-air turbulence under different conditions. Such improvement will help improve flight safety and reduce the risks and losses brought by turbulence events. Based on this, the present application provides an aircraft clear-air turbulence prediction method. The aircraft clear-air turbulence prediction method provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 the following figure. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set separately, integrated on the server 104, placed in the cloud or on other servers. The terminal 102 can send the sample set to the server 104. After receiving the sample set, for any sample in the sample set, the server 104 calculates the numerical values of each CAT diagnostic index according to the values of the respective meteorological data corresponding to the sample to obtain the CAT diagnostic value sequence corresponding to the sample; constructs a total data set according to the CAT diagnostic value sequences corresponding to each sample and the bump observation data corresponding to each sample; performs preprocessing and oversampling processing on the total data set in sequence to obtain a data set; the preprocessing includes performing data quality control and replacing the bump observation data with turbulence labels in sequence; trains and validates the W-KNN model with the data set to obtain a clear-air turbulence prediction model. The server 104 can feedback the obtained clear-air turbulence prediction model to the terminal 102. In addition, in some embodiments, the aircraft clear-air turbulence prediction method can also be implemented by the server 104 or the terminal 102 alone. For example, the terminal 102 can directly perform aircraft clear-air turbulence prediction on the sample set, or the server 104 can obtain the sample set from the data storage system and perform aircraft clear-air turbulence prediction on the sample set.

[0031] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0032] In an exemplary embodiment, as Figure 2 shown, a method for predicting clear-air turbulence of an aircraft is provided. This method is executed by a computer device, and specifically can be executed alone by a computer device such as a terminal or a server, or can be jointly executed by a terminal and a server. In the embodiments of the present application, taking this method applied to Figure 1 the server 104 in it as an example for description, it includes the following steps. Among them:

[0033] Step 201, obtain a sample set. The sample set includes multiple samples. One sample includes the turbulence observation data and the values of each meteorological data of the sample aircraft flying at the same time and the same space. The meteorological data is the meteorological data with FL≥200, including meteorological factors such as temperature, meridional wind speed, zonal wind speed, vertical wind speed, and geopotential.

[0034] Step 202, for any one sample, calculate the values of each CAT diagnosis index according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnosis value sequence corresponding to the sample. The CAT diagnosis index includes 8 indexes: eddy dissipation rate (EDR), Richardson number (Ri), Ellrod turbulence index (Ellrod1), relative vorticity squared (VOSQ), non-linear balance equation residual (UBF), vertical wind shear (VWS), horizontal temperature gradient (TEMPG), and Brown energy dissipation index (Brown2). Table 1 shows the detailed information of the above 8 indexes, including the weights WG corresponding to each index s .

[0035] Table 1 8 CAT Diagnoses and Their Corresponding Weights

[0036]

[0037] In the EDR index, C kp is the Kolmogorov-Prandtl constant, ε is the eddy dissipation, ε(z) is the eddy dissipation of the aircraft at a given height z, TKE(z) is the turbulent kinetic energy of the aircraft at a given height z, TKE is the turbulent kinetic energy of the aircraft, and u' and w' are the perturbations of the zonal wind component and the vertical wind component at the location of the aircraft. Denotes the mean of -u'w'. U is the total wind speed, N is the Brunt-Vaisala frequency, x is the longitude of the aircraft's location, and y is the latitude of the aircraft's location. is the Laplace operator, Φ is the geopotential, J() is the Jacobian operator, f and β are the Coriolis parameter and its meridional derivative respectively, and ζ a is the absolute vorticity of the aircraft, and S V is the vertical wind shear at the aircraft's location, ζ is the relative vorticity of the aircraft, u is the meridional wind at the aircraft's location, v is the zonal wind at the aircraft's location, and T is the temperature at the aircraft's location.

[0038] Step 203: Construct a total dataset based on the CAT diagnostic value sequences corresponding to each sample and the bump observation data corresponding to each sample.

[0039] Step 204: Perform preprocessing and oversampling on the total dataset in sequence to obtain a dataset; the preprocessing includes performing data quality control (deleting duplicate values and missing values) and replacing the bump observation data with turbulence labels in sequence.

[0040] Step 205: Train and validate the W-KNN model using the dataset to obtain a clear-air turbulence prediction model, which is used to predict the clear-air turbulence of the aircraft to be predicted.

[0041] Implementing the above steps 201 to 205 can improve the accuracy of the clear-air turbulence prediction results.

[0042] In another exemplary embodiment of the present application, constructing a total dataset based on the CAT diagnostic value sequences corresponding to each sample and the bump observation data corresponding to each sample specifically includes:

[0043] Perform interpolation processing on the CAT diagnostic value sequences corresponding to each sample (implemented using the 'RegularGridInterpolator' function in python, which is essentially to use bilinear interpolation and vertical interpolation methods to unify the resolution of meteorological data and observation data) to obtain the numerical sequences of the interpolated CAT diagnostic indices corresponding to each sample.

[0044] Construct a total dataset based on the numerical sequences of the interpolated CAT diagnostic indices corresponding to each sample and the bump observation data corresponding to each sample.

[0045] In practical applications, the operation of replacing the bump observation data with turbulence labels specifically includes:

[0046] (1) Convert the turbulence intensity characters recorded in the bump observation data into digital tags using regular expressions. For example: NEG (negative) → 0, LGT (light) → 1, LGT-MOD → 2, MOD → 3, MOD-SEV → 4, SEV → 5, where MOD (moderate), SEV (severe).

[0047] (2) Define turbulence tags used to characterize different turbulence intensities. Turbulence tag 0 represents no turbulence event, that is, 0 in (1). Turbulence tag 1 represents a turbulence event of moderate or higher level, that is, 3-5 in (1), as shown in formula (1):

[0048]

[0049] (3) Convert the digital tags into turbulence tags according to the definition in (2).

[0050] In practical applications, the oversampling process is carried out using the Synthetic Minority Over-sampling Technique (SMOTE) algorithm to handle the imbalance of sample categories. The specific steps are as follows:

[0051] (1) For each sample point x (the CAT diagnosis value sequence and turbulence tag corresponding to the sample) in the minority class (turbulence tag is 0), calculate the distance between this sample point and other sample points y in the minority class to obtain K nearest neighbors. Usually, the Euclidean distance is used, and the formula is as follows:

[0052]

[0053] Among them, n represents the total number of features, F xj represents the value of x on the jth feature, F yj represents the value of y on the jth feature. Here, the feature refers to 8 types of CAT diagnoses.

[0054] (2) Set a sampling ratio according to the sample imbalance ratio to determine the sampling magnification. For x, randomly select several sample points from K' neighbors. Suppose the selected nearest neighbor is x*.

[0055] (3) For each randomly selected nearest neighbor x*, use the following formula to construct new sample points with x respectively:

[0056] x new = x + rand(0,1)*(x * - x) (3)

[0057] Among them, x newFor a new sample point, rand(0,1) generates a random number between 0 and 1, which determines the proportion of the step in the direction of x*.

[0058] In another exemplary embodiment of the present application, a dataset is used to train and validate the W-KNN model to obtain a clear-air turbulence prediction model, which specifically includes:

[0059] The dataset is divided into a training set and a test set. Set a random seed number, and divide the dataset into a training set and a test set according to the ratio of 8:2 based on the random seed number. Setting the same random seed number can ensure that the data is consistent every time the dataset is divided, so that the evaluation results of the model are reproducible.

[0060] The test set and the training set are used to train and validate the W-KNN model to obtain a clear-air turbulence prediction model, which is used to predict the clear-air turbulence of the aircraft to be predicted.

[0061] In another exemplary embodiment of the present application, the test set and the training set are used to train and validate the W-KNN model to obtain a clear-air turbulence prediction model, which specifically includes:

[0062] The training set is standardized to obtain a standardized training set; only the training set is standardized to avoid the problem of data information leakage caused by directly standardizing all datasets; the purpose of standardization is to eliminate the dimensional difference between different features, scale the data features, and improve the performance of the W-KNN algorithm. The following formula is used for data Z-score standardization: Where z is the value after standardization; x is the original data value; μ is the mean of the data; σ is the standard deviation of the data.

[0063] 10-fold cross-validation is used to process the standardized training set to estimate the hyperparameter values of the W-KNN model. The hyperparameters include: "n_neighbors" (i.e., the K value), "weights" (weight function), "algorithm" (neighbor search algorithm), "leaf_size" (leaf node size), "p" (distance metric parameter).

[0064] The standardized training set and the test set are used to validate the W-KNN model with determined hyperparameter values to obtain a clear-air turbulence prediction model.

[0065] In another exemplary embodiment of the present application, 10-fold cross-validation is used to process the standardized training set to estimate the hyperparameter values of the W-KNN model, which specifically includes:

[0066] Set 10-fold cross-validation in the random search algorithm to search for the optimal hyperparameter values: Preset a list of hyperparameter values, such as the number of neighbors "n_neighbors": range(100); the weight function "weights": ['uniform', 'distance']; the KNN search algorithm "algorithm" for accelerating the nearest neighbor search: ['auto', 'ball_tree', 'kd_tree']; the leaf size "leaf_size" for controlling the construction of KD trees and ball trees, which can affect the query speed and memory consumption: [10, 20, 30, 40]; the distance metric parameter "p": [1, 2, 3, 4, 5], etc.; then divide the standardized training set into 10 subsets of equal size, where 9 subsets are used to train the model and one subset is used to test the model. Repeat this process 10 times, each time selecting a different subset as the 10-fold cross-validation test set and the rest as the 10-fold cross-validation training set, and find the hyperparameter values that can make the model perform optimally through network search.

[0067] In another exemplary embodiment of the present application, the standardized training set and the test set are used to verify the W-KNN model determined by the hyperparameter values to obtain a clear air turbulence prediction model, specifically including:

[0068] Use the standardized training set to predict the turbulence labels of each sample in the test set to obtain the predicted turbulence labels of each sample in the test set.

[0069] Use the predicted turbulence labels of each sample in the test set and the turbulence labels of each sample in the test set to verify the W-KNN model determined by the hyperparameter values to obtain a clear air turbulence prediction model.

[0070] In another exemplary embodiment of the present application, using the standardized training set to predict the turbulence labels of each sample in the test set to obtain the predicted turbulence labels of each sample in the test set, specifically including:

[0071] For any sample in the test set, calculate the Manhattan distance between the CAT diagnosis value sequence corresponding to the sample and the CAT diagnosis value sequences corresponding to each sample in the standardized training set to obtain a distance sequence.

[0072] Sort all the Manhattan distances in the distance sequence from small to large to obtain a sorted distance sequence.

[0073] Select the turbulence labels of the samples corresponding to the first K Manhattan distances in the sorted distance sequence in the standardized training set to obtain a turbulence label set; K is the value of K in the W-KNN model determined by the hyperparameter values.

[0074] Use the weighted voting method with weights to process the turbulence label set to obtain the predicted turbulence label of the sample.

[0075] This application can effectively capture clear-air turbulence by taking advantage of the non-parametric and feature-weight consideration advantages of the weighted K-nearest neighbor (W-KNN) algorithm, improving the prediction accuracy.

[0076] This application also provides an embodiment in which the aircraft clear-air turbulence prediction method provided in the above embodiment is applied to an A320 aircraft, as Figure 3 shown, and the specific steps are as follows:

[0077] Step A1: Obtain the A320 aircraft turbulence observation data for consecutive years in a certain area, such as pilot reports or aircraft onboard data.

[0078] Step A2: Screen and download the meteorological data with the same spatio-temporal characteristics as the data in Step A1, and then calculate the numerical values of the CAT diagnostic index through the above meteorological data to obtain the CAT diagnostic value sequence.

[0079] In Step A3, the above CAT diagnostic value sequence uses the bilinear interpolation and vertical interpolation methods to take the time, pressure altitude layer, longitude and latitude in the pilot report data as the target interpolation points, and uses the RegularGridInterpolator function to obtain the numerical values of 8 types of CAT diagnostic indexes at each point to unify the spatio-temporal characteristics. Then, the turbulence intensity field in the turbulence observation data and the numerical sequence of the CAT diagnostic index after interpolation processing form the total data set R. R is a two-dimensional matrix, containing 13 columns of data, that is, the set jointly composed of the time, altitude, longitude and latitude, the numerical sequence of the CAT diagnostic index after interpolation processing, and the turbulence intensity field obtained through the above method.

[0080] Step A4: Perform quality control on the total data set (delete duplicate values and missing values) to avoid affecting the prediction performance of the model, and then convert the turbulence intensity into numbers using regular expressions. The specific method is to use the're' library in python to convert the characters recording the turbulence intensity level in the pilot report data into digital labels, for example: 'NEG' → 0, 'LGT' → 1, 'LGT-MOD' → 2, 'MOD' → 3,'MOD-SEV' → 4, 'SEV' → 5, 'EXT' → 6.

[0081] Step A5: Define the turbulence label used to characterize different turbulence intensities. The turbulence label 0 represents no turbulence event, that is, 0 in Step A4; the turbulence label 1 represents a turbulence event of moderate or greater level, that is, 3-6 in Step A4, as shown in formula (1).

[0082] Step A6: Convert the digital label into a turbulence label according to the definition in Step A5.

[0083] Step A7: Use the SMOTE algorithm to perform oversampling on the processed total dataset obtained in Step A6 to obtain a dataset.

[0084] Step A8: Subsequently, set the ratio of the training set to the test set to 8:2, and divide the dataset into a training set and a test set according to this ratio.

[0085] Step A9: Use a standardization method to normalize the training set to obtain a standardized training set.

[0086] Step A10: Use 10-fold cross-validation. Divide the standardized training set into 10 subsets of equal size, where 9 subsets are used to train the model and 1 subset is used to test the model. Repeat this process 10 times. Each time, select a different subset as the 10-fold cross-validation test set, and the rest as the 10-fold cross-validation training set.

[0087] Step A11: Establish a clear-air turbulence prediction model based on the W-KNN algorithm.

[0088] Step A12: Preset a list of parameter values. For example, the number of neighbors "n_neighbors": range(100); the weight function "weights": ['uniform', 'distance']; the KNN search algorithm "algorithm" for accelerating the nearest neighbor search: ['auto', 'ball_tree', 'kd_tree']; the leaf size "leaf_size" for controlling the construction of the KD tree and the ball tree, which can affect the query speed and memory consumption: [10, 20, 30, 40]; the distance metric parameter "p": [1, 2, 3, 4, 5], etc. Then, according to the data of the 10-fold cross-validation training set in Step A10, train the model, and then use the data of the 10-fold cross-validation test set in Step A10 to test the trained model; find the parameters that can make the model perform optimally through network search; it is found through research that when n_neighbors (i.e., the K value) = 27, weights = distance, algorithm = kd_tree, leaf_size = 20, p = 1, the prediction effect of the model is better.

[0089] Step A13: Use the test set in Step A7 to evaluate the model obtained in Step A12.

[0090] (1) Let L be the training set of L = {(Xi, yi), i = 1,..., n}, where Xi is the numerical sequence of the interpolated CAT diagnostic index corresponding to the i-th sample, and yi is the turbulence label corresponding to the i-th sample.

[0091] (2) Let Y be the numerical sequence of the interpolated CAT diagnostic index corresponding to any sample in the test set. Calculate the Manhattan distances from Y to each Xi in (1) using the following formula:

[0092]

[0093] where N represents the total number of CAT diagnostic indices in the numerical sequence of the interpolated CAT diagnostic index, Y s represents the value of the sth CAT diagnostic index in Y, and X is represents the value of the sth CAT diagnostic index in Xi.

[0094] (3) Sort the distances obtained in (2) from smallest to largest. Select the first K Manhattan distances and their corresponding turbulence labels yi to obtain a set of turbulence labels, and find the K neighborhoods P k :

[0095]

[0096] (4) Use well-known steps to construct the Receiver Operating Characteristic (ROC) curve, and then calculate the corresponding weight WG of the sth CAT diagnostic index using formula (7): s :

[0097]

[0098] where AUC s represents the area under the sth CAT diagnostic index corresponding to the Receiver Operating Characteristic (ROC) curve.

[0099] The Receiver Operating Characteristic (ROC) curve consists of 1 - PODn (x - axis) and PODy (y - axis), and is calculated by statistically analyzing the detection probability (PODy) of predicting the turbulence label as 1 and the detection probability (PODn) of predicting the turbulence label as 0 in the test set. When the model outputs a probability, different judgment thresholds will result in different PODy and 1 - PODn. By plotting the PODy and 1 - PODn points at multiple judgment thresholds and connecting these points, the ROC curve can be obtained.

[0100] (5) Use the weighted voting method with weights according to formula (8) to process the set of turbulence labels to obtain the predicted turbulence label y corresponding to Y:

[0101]

[0102] where I c (s i ) represents the category to which the ith sample belongs, as follows:

[0103]

[0104] Among them, c1 represents moderate or above turbulence events, and c2 represents no turbulence events.

[0105] (6) Model performance evaluation. The AUC value under the ROC curve is selected as the evaluation index for the clear-air turbulence prediction model obtained in step A12.

[0106] The working principle of ROC is that as the decision threshold changes, the ROC curve graph shows the relationship between PODy and 1 - PODn. The closer the curve is to the upper left corner (the closer to 1, with 0.5 being a random guess), the more perfect the prediction effect. The AUC value of the ROC curve is ultimately used as the performance skill score for turbulence prediction. For each threshold, PODy and PODn can be defined as:

[0107]

[0108] Where a: the number of samples with the predicted turbulence label being the positive class and the true turbulence label being the positive class, b: the number of samples with the predicted turbulence label being the positive class but the true turbulence label being the negative class, c: the number of samples with the predicted turbulence label being the negative class but the true turbulence label being the positive class, d: the number of samples with the predicted turbulence label being the negative class and the true turbulence label being the negative class.

[0109] Figure 4 is the ROC curve for validating the A320 test data set by the W - KNN model. The dashed line in it represents a random guess, and the AUC value of the area under the ROC curve reaches 0.97, indicating that the classification prediction ability of this model is good. [[ID=()]]

[0110] Furthermore, the maximum true skill statistic (TSS) scores of each model are statistically analyzed, also known as the Hanssen - Kuipers (H - K) discriminant, defined as TSS = PODy + PODn - 1. Its value range is from - 1 to 1. A value of 1 indicates that the diagnostic model has perfect skills in classification prediction, and a value less than 0 represents a negative discrimination skill (0 indicates no skill). The area under the ROC curve (AUC) and the true skill score (MaxTSS) for different models predicting CAT are shown in Table 2.

[0111] Table 2 Area under the ROC curve (AUC) and true skill score (MaxTSS) for different models predicting CAT

[0112]

[0113] Table 2 shows the prediction evaluation of each forecasting model for the selected clear-air turbulence events recorded in the A320. Weighted K-nearest neighbor (W-KNN), random forest (RF), and the integrated forecasting model (CATavg) represent different forecasting methods. This table compares the Max TSS values obtained from different diagnostic model forecasting methods. Among the single diagnostic models, the Ellrod1 index and the UBF index have the highest (0.272) and lowest (0.095) Max TSS values, respectively.

[0114] It is worth noting that compared with the single diagnostic model, the integrated forecasting model has a higher TSS value in turbulence prediction. This is because the integrated forecasting improves the statistical performance by combining multiple turbulence generation mechanisms. Compared with the random forest and the integrated forecasting model, the weighted K-nearest neighbor (W-KNN) forecasting model shows the highest TSS value in diagnosing moderate or greater turbulence events and non-turbulence events. Specifically, the TSS value of the W-KNN forecasting model increases from 0.315 of the integrated forecasting model to 0.825.

[0115] Compared with the integrated forecasting model, the W-KNN forecasting model has a smaller computational cost, and the prediction running time efficiency is increased by 1.5 to 1.7 times. Through experimental comparison, the computational time of the integrated forecasting method increases linearly with the number of single diagnoses used.

[0116] Generally speaking, the W-KNN forecasting model not only ensures high prediction accuracy but also significantly improves the computational efficiency, providing an effective technical means for the real-time forecasting of clear-air turbulence. In summary, the aircraft clear-air turbulence prediction method provided in this embodiment greatly improves the ability to capture CAT, providing safety guarantees for coping with the increasing clear-air turbulence in the future. It can also provide practical references for pilots flying A320 aircraft before takeoff, thereby improving the aviation safety factor and reducing economic costs. This has important practical significance for ensuring the safety and efficiency of civil aviation transportation.

[0117] This application comprehensively considers the effects of factors such as the horizontal and vertical shear of wind, the horizontal and vertical gradients of temperature, and the local changes of temperature on the CAT intensity, increasing the accuracy of turbulence prediction. At the same time, the W-KNN algorithm after parameter tuning is used for model training, reducing the model complexity and training time, and being able to obtain the prediction results of aircraft clear-air turbulence faster. The oversampling technique of the SMOTE algorithm is used to process the data set, avoiding the influence of the data sample class imbalance problem on the performance of the model, and improving the accuracy and reliability of the model.

[0118] Based on the same inventive concept, an embodiment of the present application further provides an aircraft clear-air turbulence prediction device for implementing the aircraft clear-air turbulence prediction method involved above. The implementation solutions provided by this device to solve problems are similar to those described in the above method. Therefore, the specific limitations in one or more embodiments of the aircraft clear-air turbulence prediction device provided below can refer to the limitations on the aircraft clear-air turbulence prediction method in the above text, and will not be repeated here.

[0119] In an exemplary embodiment, an aircraft clear-air turbulence prediction device is provided, including:

[0120] An acquisition module, configured to acquire a sample set, where the sample set includes multiple samples, and one sample includes the turbulence observation data and the values of each meteorological data when a sample aircraft flies at the same time and in the same space.

[0121] A CAT diagnosis value calculation module, configured to, for any one sample, calculate the numerical values of each CAT diagnosis index according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnosis value sequence corresponding to the sample;

[0122] A total data set construction module, configured to construct a total data set according to the CAT diagnosis value sequence corresponding to each sample and the turbulence observation data corresponding to each sample.

[0123] A data set determination module, configured to perform preprocessing and oversampling processing on the total data set in sequence to obtain a data set; the preprocessing includes performing data quality control and replacing the turbulence observation data with turbulence labels in sequence.

[0124] A clear-air turbulence prediction model determination module, configured to train and verify a W-KNN model using the data set to obtain a clear-air turbulence prediction model, where the clear-air turbulence prediction model is used to predict the clear-air turbulence of an aircraft to be predicted.

[0125] In an exemplary embodiment, a computer device is provided. This computer device can be a server or a terminal, and its internal structure diagram can be as Figure 5As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store aircraft clear-air turbulence prediction data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements an aircraft clear-air turbulence prediction method.

[0126] Those skilled in the art can understand that Figure 5 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the above method embodiments are implemented.

[0127] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by the processor, the above method embodiments are implemented.

[0128] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by the processor, the above method embodiments are implemented.

[0129] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0130] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the various embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random-access memories (ReRAM), magnetoresistive random-access memories (MRAM), ferroelectric random-access memories (FRAM), phase change memories (PCM), graphene memories, and the like. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0131] The databases involved in the various embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the various embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0132] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0133] Specific examples are used in this article to elaborate on the principles and implementation methods of this application. The descriptions of the above embodiments are only used to help understand the methods and core ideas of this application; at the same time, for those of ordinary skill in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. An aircraft clear-air turbulence prediction method, characterized in that, The aircraft clear-air turbulence prediction method includes: Obtaining a sample set, where the sample set includes multiple samples, and one sample includes the turbulence observation data and the values of each meteorological data when the sample aircraft flies at the same time and in the same space; For any one sample, calculating the values of each CAT diagnosis index according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnosis value sequence corresponding to the sample; CAT is future clear-air turbulence; Constructing a total data set according to the CAT diagnosis value sequences corresponding to each sample and the turbulence observation data corresponding to each sample; Performing preprocessing and oversampling processing on the total data set in sequence to obtain a data set; the preprocessing includes performing data quality control and replacing the turbulence observation data with turbulence labels in sequence; Dividing the data set into a training set and a test set; Performing standardization processing on the training set to obtain a standardized training set; Using 10-fold cross-validation to process the standardized training set to estimate the hyperparameter values of the W-KNN model; For any one sample in the test set, calculating the Manhattan distance between the CAT diagnosis value sequence corresponding to the sample and the CAT diagnosis value sequences corresponding to each sample in the standardized training set to obtain a distance sequence; Sorting all the Manhattan distances in the distance sequence from small to large to obtain a sorted distance sequence; Selecting the turbulence labels of each sample corresponding to the first K Manhattan distances in the sorted distance sequence in the standardized training set to obtain a turbulence label set; K is the value of K in the W-KNN model determined by the hyperparameter value; W-KNN represents weighted K-nearest neighbor; Processing the turbulence label set by using the weighted voting method to obtain the predicted turbulence label of the sample; Using the predicted turbulence labels of each sample in the test set and the turbulence labels of each sample in the test set to verify the W-KNN model determined by the hyperparameter value to obtain a clear-air turbulence prediction model. The clear-air turbulence prediction model is used to predict the clear-air turbulence of the aircraft to be predicted.

2. The aircraft clear-air turbulence prediction method according to claim 1, wherein, Constructing a total data set according to the CAT diagnosis value sequences corresponding to each sample and the turbulence observation data corresponding to each sample, specifically including: Performing interpolation processing on the CAT diagnosis value sequences corresponding to each sample to obtain a numerical sequence of the interpolated CAT diagnosis index corresponding to each sample; Constructing a total data set according to the numerical sequence of the interpolated CAT diagnosis index corresponding to each sample and the turbulence observation data corresponding to each sample.

3. An aircraft clear-air turbulence prediction device, characterized in that, The aircraft clear-air turbulence prediction device includes: An acquisition module, configured to acquire a sample set, where the sample set includes multiple samples, and one sample includes the turbulence observation data and the values of each meteorological data when the sample aircraft flies at the same time and in the same space; A CAT diagnosis value calculation module, configured to, for any one sample, calculate the values of each CAT diagnosis index according to the values of each meteorological data corresponding to the sample to obtain the CAT diagnosis value sequence corresponding to the sample; CAT is future clear-air turbulence; A total data set construction module, configured to construct a total data set according to the CAT diagnosis value sequences corresponding to each sample and the turbulence observation data corresponding to each sample; A dataset determination module, configured to perform preprocessing and oversampling processing on the total dataset in sequence to obtain a dataset; the preprocessing includes performing data quality control and replacing bump observation data with turbulence labels in sequence; A clear-air turbulence prediction model determination module, configured to divide the dataset into a training set and a test set; perform standardization processing on the training set to obtain a standardized training set; use 10-fold cross-validation to process the standardized training set to estimate the hyperparameter values of the W-KNN model; for any sample in the test set, calculate the Manhattan distance between the CAT diagnosis value sequence corresponding to the sample and the CAT diagnosis value sequences corresponding to the samples in the standardized training set to obtain a distance sequence; sort all the Manhattan distances in the distance sequence from smallest to largest to obtain a sorted distance sequence; select the turbulence labels of the samples corresponding to the first K Manhattan distances in the sorted distance sequence in the standardized training set to obtain a turbulence label set; K is the value of K in the W-KNN model determined by the hyperparameter value; W-KNN represents weighted K-nearest neighbor; use the weighted voting method to process the turbulence label set to obtain the predicted turbulence label of the sample; use the predicted turbulence labels of the samples in the test set and the turbulence labels of the samples in the test set to verify the W-KNN model determined by the hyperparameter value to obtain a clear-air turbulence prediction model. The clear-air turbulence prediction model is used to predict the clear-air turbulence of an aircraft to be predicted.

4. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the aircraft clear-air turbulence prediction method according to any one of claims 1-2.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the aircraft clear-air turbulence prediction method according to any one of claims 1-2.

6. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the aircraft clear-air turbulence prediction method according to any one of claims 1-2.

Images