Aviation flight meteorological analysis method and system
Through machine learning-based meteorological forecast optimization method and meteorological factor interpolation technology, the prediction problem of visual range obstacle weather phenomena in aviation flight is solved, accurate forecasting of route meteorological conditions and flexible route design are achieved, and the safety of aviation flight is improved and efficiency.
Patent Information
- Application Number
- CN202411054747.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
AI Technical Summary
The prior art is difficult to accurately and reliably predict weather phenomena such as horizontal visibility and low clouds during aviation flights, and it is impossible to flexibly design flight routes to cope with meteorological conditions in different spaces and times.
Using machine learning-based meteorological forecasting optimization method, we can build a large data of weather forecasting optimized by aviation service, combine meteorological factor interpolation technology and numerical forecasting tools to generate three-dimensional grid forecasting field products, provide customized meteorological factor forecasting data, and use LightGBM and XGBoost models perform prediction optimization of visibility and cloud high volume.
It realizes accurate and reliable forecasts of meteorological elements at all route points and forecast moments throughout the entire route, and can flexibly design flight routes in different spaces and times, provide timely and accurate route meteorological information, and improves the safety of aviation flights. Sex and efficiency.
Smart Images

Figure CN118981616A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of aviation meteorological technology, and in particular to an aviation flight meteorological analysis method and system. Background Art
[0002] Aviation meteorology is crucial to aviation flight safety. Aviation flight needs to obtain accurate, reliable and comprehensive meteorological information in a timely manner to reduce losses caused by weather and avoid various flight accidents caused by weather factors. With the advancement of new construction and expansion of air traffic control projects at airports across the country, the aviation meteorological professional service system has developed rapidly, which can provide targeted and customizable meteorological services to local air traffic control, airports, base airlines and other meteorological users, greatly helping the aviation transportation industry to improve service quality and efficiency. At the same time, with the increasing maturity and popularization of drone technology and the pilot construction of low-altitude short-distance aviation business, aviation flights have further increased their requirements for refined services of aviation meteorological big data.
[0003] On the other hand, whether it is civil aviation or low-altitude flight, weather phenomena such as obstructed visibility have an important impact on flight safety, because they will significantly reduce the pilot's line of sight, increase the difficulty and risk of flight. Weather phenomena refer to some physical processes that occur in the atmosphere and on the ground. They are produced under certain weather conditions and are the comprehensive result of changes in various meteorological elements. At present, weather phenomena such as obstructed visibility mainly rely on the comprehensive judgment of various meteorological elements. The judgment is relatively complex and the accuracy is not enough. How to improve the accurate and reliable forecast of horizontal visibility and low clouds in aviation flight is a technical problem that urgently needs to be solved. At the same time, in order to meet the development needs of aviation meteorology under the new situation, and to provide precise meteorological services for flexible routes at low altitude and short distances in the future, how to provide a route planning and analysis method that can flexibly design flight routes in different spaces and times, and provide accurate and reliable route meteorological information in a timely manner, is also a technical problem that urgently needs to be solved. Summary of the invention
[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide an aviation flight meteorological analysis method and system. Based on the needs of aviation flight meteorological analysis and according to the meteorological elements of concern to aviation meteorology, the present invention provides a targeted aviation meteorological big data refinement service, which can forecast the aviation meteorological elements of each waypoint and each forecast time during the entire route. In particular, it can accurately and reliably forecast the meteorological elements of weather phenomena such as visibility obstruction (such as horizontal visibility and low clouds). Furthermore, according to the development needs of aviation meteorology under the new situation, it provides a method and system that can flexibly design flight routes in different spaces and times, and provide accurate and reliable route flight meteorological information in a timely manner.
[0005] To achieve the above objectives, the present invention provides the following technical solutions:
[0006] An aviation flight meteorological analysis method, the method comprising the following steps:
[0007] Based on the meteorological elements of concern to aviation meteorology, meteorological forecast big data for aviation service optimization is constructed; wherein the meteorological elements include meteorological elements related to weather phenomena such as visibility obstruction, and when constructing the meteorological forecast big data for aviation service optimization, forecast optimization based on machine learning is performed on meteorological elements related to weather phenomena such as visibility obstruction;
[0008] Acquire planned route information, the route information including waypoint information on the route and flight time information of the route;
[0009] Through the interpolation technology of meteorological elements and the meteorological forecast big data optimized by the aforementioned aviation services, the aviation meteorological element information of each waypoint and each forecast time on the planned route is analyzed to obtain the meteorological element forecast data of any waypoint in the entire route process, and the aviation meteorological element forecast data is visualized according to the needs of aviation flight.
[0010] Further, based on the model forecast data provided by the numerical forecast tool, the aforementioned meteorological forecast big data for aviation service optimization is constructed. At this time, after obtaining the model forecast data provided by the preset numerical forecast tool, a three-dimensional grid forecast field product of meteorological elements of concern to aviation meteorology is generated based on the model forecast data. The three-dimensional grid forecast field product sets the vertical resolution according to the aviation flight requirements, sets a smaller height interval at the low layer below the preset height threshold, and sets a larger height interval at the high layer above the preset height threshold;
[0011] The three-dimensional grid forecast field product is configured to provide altitude layer data, customized meteorological element forecast data of aviation meteorological concern, aviation index forecast data of icing index and turbulence index, and full-plane visualization display information of airports, waypoints and control areas.
[0012] Furthermore, the meteorological elements of concern to aviation meteorology also include one or more of temperature, precipitation, snowfall, relative humidity, wind direction and speed, 0°C altitude and low-altitude wind shear.
[0013] Furthermore, the meteorological elements related to the visibility obstruction weather phenomenon include visibility and cloud height and cloud amount, and corresponding to the forecast optimization of visibility and cloud height and cloud amount, a visibility forecast optimization module and a cloud height and cloud amount forecast optimization module are respectively configured;
[0014] The visibility forecast optimization module is configured to: develop a visibility site optimization forecast model algorithm based on the model forecast data provided by the numerical forecast tool, using a preset machine learning model, supplemented by observation data from airport meteorological observation points, and external spatiotemporal background feature information, and provide visibility forecast products with a preset time resolution;
[0015] The cloud height and cloud amount forecast optimization module is configured as follows: based on the three-dimensional grid forecast field products of meteorological elements generated by numerical forecast tools and the Metar messages of airport observations, using a preset machine learning model, developing optimized cloud amount classification forecast model algorithms and cloud base height forecast model algorithms, and providing cloud amount classification forecast products and cloud height regression products with a preset time resolution.
[0016] Furthermore, in the visibility forecast optimization module, the machine learning model adopts the LightGBM model. At this time, the visibility forecast optimization module is configured to perform the following steps:
[0017] The input feature information of the configuration model includes model forecast data, site meteorological observation data and pollutant concentration data of the airport meteorological observation point, and external time feature information and space feature information. The time change of visibility and different features of visibility change at different sites are identified by configuring the external time feature information and space feature information in the input feature; wherein the time feature information includes the forecast step, month and hour information of the model, and the space feature information includes the station number and the latitude and longitude information of each site;
[0018] Train the LightGBM model based on the input features and analyze the importance of each feature; configure the dataset information of the training set, validation set, and test set during model training;
[0019] The model parameters are tuned through evaluation and analysis of the test set to optimize the forecast model; the final optimized forecast model is run in real time, real-time observation data and model forecast data are input, and hourly site visibility forecast products are output.
[0020] Furthermore, the visibility forecast optimization module is provided with a data preprocessing unit and a data cleaning unit; the data preprocessing unit is used to focus on low visibility weather during the model training process; the data cleaning unit is used to alleviate the problem of data imbalance, and remove high visibility samples that meet the screening conditions in the training set and the validation set to reduce the proportion of high visibility samples in the sample;
[0021] The data preprocessing unit is configured to: divide visibility into multiple levels according to visibility distribution and business needs, configure different weight coefficients for data samples of different visibility levels, wherein a larger weight coefficient is configured for low visibility samples than for high visibility samples to increase the proportion of low visibility samples in the loss function, so that the model training is inclined to low visibility events; and divide multi-year observation data and model forecast data into relatively independent training sets, validation sets and test sets, wherein the data of the most recent year is used as the test set, and the data of the previous year is sampled according to a preset ratio using a segmented interval method to divide the training set and the validation set;
[0022] The data cleaning unit is configured as follows: first, according to the designed daily start time, only data samples in which low visibility events occurred at any station on that day are retained; then, according to the observation data and model forecast data, high visibility samples with correct model forecasts are eliminated, wherein high visibility samples in the training set and the validation set are eliminated according to a preset ratio.
[0023] Furthermore, in the cloud height and cloud amount forecast optimization module, the machine learning model adopts an XGBoost model and a 4-layer fully connected neural network model. At this time, the cloud height and cloud amount forecast optimization module is configured to perform the following steps:
[0024] The input feature information of the configuration model includes the three-dimensional grid forecast field product of meteorological elements, the airport observation cloud height and cloud amount information provided by the Metar message, and the external time feature information and spatial feature information. The time change of cloud height and cloud amount, as well as the different characteristics of cloud height and cloud amount changes at different stations are identified by configuring the external time feature information and spatial feature information in the input feature; wherein the time feature information is the forecast step of the model, the day is the day of the year, and the hour is the hour of the day; the spatial feature information includes latitude, longitude and altitude information;
[0025] Based on the input features, the XGBoost model is trained. During model training, the dataset information of the training set, validation set, and test set is configured. The model parameters are tuned through evaluation and analysis of the test set. The optimized forecast model finally tuned is run in real time, and the pattern forecast data is input daily to output hourly cloud classification forecast products. Also, based on the input features, a 4-layer fully connected neural network model is trained. During model training, the dataset information of the training set, validation set, and test set is configured. During model training, the model parameters are tuned through evaluation and analysis of the test set. The optimized forecast model finally tuned is run in real time, and the pattern forecast data is input daily to output hourly cloud height regression products.
[0026] Furthermore, in the visibility forecast optimization module, the model forecast data used include the model forecast data of the East China regional business model CMA-SH9 and the weather-atmosphere environment coupling model;
[0027] In the cloud height and cloud amount forecast optimization module, the model forecast data used is the three-dimensional grid forecast field product of meteorological elements obtained by analyzing and processing the East China regional business model CMA-SH9; the Metar message records cloud shape, cloud height and cloud amount data, and there are multiple layers of records about cloud shape, cloud height and cloud amount. The bottom layer of record information is selected, and multi-level encoding is performed according to the cloud amount ratio; and, according to the set top layer height, the cloud amount observation and prediction values of the height layer greater than the top layer height are set to correspond to the top layer height, and normalized according to the top layer height.
[0028] Furthermore, it includes a route planning spatial analysis module and a route planning time analysis module, which are used to respectively perform route planning spatial analysis and route planning time analysis to obtain the weather conditions of any waypoint on the planned route at any time; the meteorological element interpolation technology includes meteorological element spatial interpolation and meteorological element temporal interpolation;
[0029] The route planning spatial analysis module is configured to: obtain route information of each waypoint during the route flight, and interpolate the grid point data of the forecast data to the required route position; wherein, the humidity, ice accumulation index and turbulence index at the waypoint are interpolated by using the inverse distance weighted method, and the four points closest to the required position are selected as sample points of the inverse distance weighted method; in the route planning analysis, the grid data of the pattern is first interpolated to the waypoint selected by the route, and then the time analysis is performed;
[0030] The route planning time analysis module is configured as follows: according to the continuity of flight conditions, a time interpolation method is used to interpolate meteorological elements with longer time intervals to obtain data at required times; wherein, for humidity, icing index and turbulence index during route flight, a quadratic spline interpolation method is used to obtain data at required times.
[0031] Furthermore, it also includes a custom route analysis module, which is configured to perform the following steps:
[0032] Enter the longitude and latitude information of each waypoint on the route into the system by coordinate input or map selection to complete the route preset;
[0033] After selecting a preset route, enter the route start flight time, and then obtain the time of each waypoint passing through the route based on the route start flight time; the time of each waypoint passing through the route is calculated by default according to the average reference speed of domestic civil aviation and the distance between waypoints, or the time of each waypoint passing through the route can be modified separately according to the user's settings;
[0034] The humidity, turbulence index and ice accumulation index of aviation weather forecast data are interpolated to each waypoint through the route planning spatial analysis module, and the aviation flight conditions when the route passes through each waypoint are obtained using the route planning time analysis module;
[0035] The vertical information of weather conditions at each waypoint in the route is displayed in the form of a profile.
[0036] The present invention also provides an aviation flight meteorological analysis system, the system comprising:
[0037] A weather forecast big data device for aviation service optimization is used to construct weather forecast big data for aviation service optimization based on meteorological elements of aviation meteorology concern; wherein the meteorological elements include meteorological elements related to weather phenomena such as visibility obstruction, and when constructing the weather forecast big data for aviation service optimization, forecast optimization based on machine learning is performed for meteorological elements related to weather phenomena such as visibility obstruction;
[0038] A route planning and analysis device is used to obtain planned route information, the route information including waypoint information on the route and flight time information of the route; and, through meteorological element interpolation technology and the aforementioned aviation service optimized weather forecast big data, analyze the aviation meteorological element information of each waypoint on the planned route and each forecast time to obtain the meteorological element forecast data of any waypoint in the entire route process, and visualize the aviation meteorological element forecast data according to the needs of aviation flight.
[0039] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art, as an example: Based on the needs of aviation flight meteorological analysis and according to the meteorological elements of aviation meteorology concern, the present invention provides targeted aviation meteorological big data refinement services, which can forecast the aviation meteorological elements of each waypoint and each forecast time during the entire route, especially the meteorological elements of weather phenomena such as visibility obstruction (such as horizontal visibility and low clouds) can be accurately and reliably forecasted. Furthermore, according to the development needs of aviation meteorology under the new situation, it provides the ability to flexibly design flight routes in different spaces and times, and to provide accurate and reliable route flight meteorological information in a timely manner.
[0040] The aviation flight weather analysis solution provided by the present invention includes weather forecast big data technology for optimizing aviation services and route planning analysis technology.
[0041] In the weather forecast big data technology for aviation service optimization, the three-dimensional grid forecast field products of meteorological elements of aviation meteorology concern are mainly generated based on the model forecast data of the preset numerical forecast tool (such as CMA-SH9 weather model forecast big data, etc.). The three-dimensional grid forecast field products can provide accurate and reliable weather forecast big data and provide data support for aviation meteorological route planning. In specific implementation, for example, the product can provide altitude layer data according to the specific needs of aviation flight, and provide customized meteorological element forecast data such as temperature, precipitation, snowfall, relative humidity, wind direction and speed, 0℃ altitude, low-altitude wind shear, visibility, low cloud cover, cloud base height, etc. of aviation meteorology concern; and, according to the needs of aviation meteorological services, it can also provide aviation index forecast data such as ice accumulation index and turbulence index; at the same time, it can also perform full-plane visualization of airports, waypoints and control areas according to needs.
[0042] In particular, for weather elements related to visibility obstruction weather phenomena, a forecast optimization based on machine learning is designed. Based on the model forecast data of the numerical forecast tool, the LightGBM machine learning algorithm is used, supplemented by the meteorological observation data and pollutant concentrations of the airport observation points, as well as the external spatiotemporal background feature information, to develop a visibility site optimization correction forecast algorithm that can provide visibility forecast products with a time resolution of 1 hour. In addition, based on the three-dimensional grid forecast field products and Metar messages of meteorological elements generated by the numerical forecast tool, XGBoost is used for cloud classification forecasting, and a 4-layer neural network is used for cloud base height forecasting.
[0043] In the route planning and analysis technology, a custom route planning function is provided - for example, the route planning can be carried out conveniently and quickly through the coordinate point route planning method; at the same time, through the meteorological element interpolation technology and the aforementioned aviation service optimized meteorological forecast big data technology, the aviation meteorological elements of each waypoint and each forecast time on the planned route are visualized, and the aviation conditions forecast display of each altitude layer throughout the route is provided to assist route analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 A flowchart of an aviation flight weather analysis method provided in an embodiment of the present invention.
[0045] Figure 2 A visibility data graph displayed by a three-dimensional grid forecast field product provided in an embodiment of the present invention.
[0046] Figure 3 A cloud base height data graph displayed by a three-dimensional grid forecast field product provided in an embodiment of the present invention.
[0047] Figure 4 This is an example diagram of the operation of presetting a route through coordinate points provided in an embodiment of the present invention.
[0048] Figure 5 An example of a coordinate point route provided in an embodiment of the present invention - a route turbulence profile.
[0049] Figure 6 An example of airport weather information forecast provided by an embodiment of the present invention - temperature map. DETAILED DESCRIPTION
[0050] The aviation flight meteorological analysis method and system disclosed in the present invention are further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated, and they can be combined with each other to achieve better technical effects. In the drawings of the following embodiments, the same reference numerals appearing in each drawing represent the same features or components, which can be applied to different embodiments. Therefore, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.
[0051] It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the invention. Any modification of the structure, change of the proportion relationship or adjustment of the size should fall within the scope of the technical content disclosed by the invention without affecting the effects and purposes that can be achieved by the invention. The scope of the preferred embodiments of the present invention includes other implementations, in which the functions can be performed in a substantially simultaneous manner or in a reverse order according to the functions involved, which should be understood by those skilled in the art of the technical field to which the embodiments of the present invention belong.
[0052] Techniques, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the techniques, methods, and devices should be considered as part of the specification. In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.
[0053] Example
[0054] See also Figure 1 As shown, an aviation flight meteorological analysis method is provided in this embodiment, and the method includes the following steps.
[0055] S100, constructing meteorological forecast big data for optimizing aviation services based on meteorological elements of concern to aviation meteorology. The meteorological elements include meteorological elements related to weather phenomena such as visibility obstruction. When constructing meteorological forecast big data for optimizing aviation services, forecast optimization based on machine learning is performed on meteorological elements related to weather phenomena such as visibility obstruction.
[0056] S200, obtaining planned route information, wherein the route information includes waypoint information on the route and flight time information of the route.
[0057] S300, through the meteorological element interpolation technology and the meteorological forecast big data optimized by the aforementioned aviation services, analyzes the aviation meteorological element information of each waypoint and each forecast time on the planned route, and displays it visually.
[0058] Specifically, based on the aviation meteorological element information of each waypoint on the route and each forecast time, the aviation meteorological element forecast data of any waypoint in the entire route can be obtained, and the aviation meteorological element forecast data can be visualized according to the needs of aviation flight. The aviation meteorological elements include customized meteorological element data such as temperature, precipitation, snowfall, relative humidity, wind direction and speed, 0°C altitude, low cloud cover, low-altitude wind shear, visibility, cloud base height, etc. that are of concern to aviation meteorology. Among them, the meteorological elements related to visibility obstruction weather phenomena include visibility and cloud height and cloud cover. As an example, see Figure 2 As shown in the figure, an example of the forecast data of visibility at a certain moment is shown. Figure 3 As shown in the figure, an example of the forecast data of cloud base height at a certain moment is shown.
[0059] In this embodiment, the aforementioned aviation service optimization weather forecast big data is constructed based on the model forecast data provided by the numerical forecast tool. At this time, after obtaining the model forecast data provided by the preset numerical forecast tool, a three-dimensional grid forecast field product of the meteorological elements of concern to aviation meteorology is generated based on the model forecast data (at this time, the constructed aviation service optimization weather forecast big data is the three-dimensional grid forecast field product of the meteorological elements). The three-dimensional grid forecast field product can set the vertical resolution according to the aviation flight requirements, set a smaller height interval at the low layer below the preset height threshold, and set a larger height interval at the high layer above the preset height threshold.
[0060] The three-dimensional grid forecast field product is configured to provide altitude layer data, customized meteorological element forecast data of aviation meteorological concern, aviation index forecast data of icing index and turbulence index, and full-plane visualization display information of airports, waypoints and control areas.
[0061] For example, based on the meteorological model forecast big data of the numerical forecast tool CMA-SH9, a three-dimensional grid forecast field product of meteorological elements of concern to aviation meteorology can be generated. The CMA-SH9 East China regional mesoscale regional model is built based on the ADAS data assimilation system and the mesoscale model WRF. It is mainly for short-term weather forecasts with a horizontal resolution of 9km. The forecast system reports twice a day at 08:00 and 20:00 Beijing time, and the forecast time is 72h. The generated three-dimensional grid forecast field product, according to aviation flight requirements, when the set height threshold is 3000 meters, the vertical resolution is set as follows: 300-meter intervals are used for low layers below 3000 meters (focus on low layers below 3000 meters), and 600-meter intervals are used for high layers above 3000 meters. Elements can include customized data such as temperature, precipitation, snowfall, relative humidity, wind direction and speed, 0℃ altitude, low-altitude wind shear, visibility, low cloud cover, and cloud base height that are of concern to aviation meteorology. At the same time, airports, waypoints and control areas can be fully visualized according to needs.
[0062] In this embodiment, preferably, a visibility forecast optimization module and a cloud height and cloud amount forecast optimization module are respectively configured corresponding to the forecast optimization of visibility and cloud height and cloud amount.
[0063] The visibility forecast optimization module is configured to: develop a visibility site optimization forecast model algorithm based on the model forecast data provided by the numerical forecast tool, using a preset machine learning model, supplemented by observation data from airport meteorological observation points, and external spatiotemporal background feature information, and provide visibility forecast products with a preset time resolution.
[0064] In the visibility forecast optimization module, the machine learning model preferably adopts the LightGBM model. The main idea of the LightGBM model is to use weak classifiers (decision trees) for iterative training to obtain the optimal model. The feature of the model is that it uses a histogram-based decision tree algorithm and a layer-by-layer leaf growth strategy optimization algorithm with depth restrictions. It has the advantages of good training effect, support for efficient parallelism, strong robustness, and not easy to overfit, and is very suitable for the multi-classification problem in the present invention.
[0065] The input features of the model include model forecast data, observation data from airport meteorological observation points, and external spatiotemporal background feature information.
[0066] The observation data mainly include meteorological observation data at the airport site and pollution concentration data observed at sites near the airport, where the observed meteorological elements may include visibility, temperature, air pressure, relative humidity, wind speed, wind direction, etc., and pollutant concentrations may include PM10 and PM2.5; wind speed and wind direction are converted into meridional wind (V) and latitudinal wind (U) in actual applications.
[0067] The model forecast data may include the East China regional business model CMA-SH9 and weather-atmospheric environment coupled model forecast products. Among them, CMA-SH9 includes ground forecast and high-altitude forecast. Ground forecast variables include visibility, 2m and ground temperature, sea level and ground pressure, precipitation, relative humidity, meridional wind (V) and zonal wind (U), cloud cover, dew point temperature and albedo, etc.; high-altitude forecast mainly includes relative humidity, meridional wind (V) and zonal wind (U), temperature and dew point temperature of the six pressure layers below 500hPa. Weather-atmospheric environment coupled model forecast products mainly include visibility, liquid water content and the concentrations of six pollutants: PM10, CO, NO2, O3, PM2.5 and SO2.
[0068] Except for the CMA-SH9 high-altitude forecast, which has a time interval of 3 hours, the other forecast products of the numerical model are hourly forecasts. Therefore, in order to ensure the consistency between the data, the high-altitude forecast products can be interpolated to hourly intervals. In addition, for the airport, the present invention preferably selects a 3×3 grid forecast product centered on the observation point as the input feature of the model. The detailed information of the observation and model variables is shown in the table below, which illustrates the variable information of the observation and model forecast products as the input features of the machine learning model.
[0069]
[0070] Since visibility has strong spatial locality and time-varying characteristics such as daily and seasonal changes, in order to improve the forecasting skills of the model, external spatiotemporal background feature information is added to the input features of this embodiment, including external time feature information and space feature information. By configuring external time feature information and space feature information in the input features, the time variation of visibility and the different characteristics of visibility changes at different sites can be identified. Specifically, the time feature information can include the forecast step, month and hour information of the model, and the space feature information can include the station number and the longitude and latitude information of each site.
[0071] At this time, the visibility forecast optimization module is configured to perform the following steps: configure the input feature information of the model; train the LightGBM model based on the input features, and analyze the importance of each feature; configure the data set information of the training set, validation set and test set during model training; tune the model parameters through evaluation and analysis of the test set to optimize the forecast model; run the optimized forecast model (or optimal forecast model) obtained by the final tuning in real time, input real-time observation data and model forecast data, and output hourly airport site visibility forecast products.
[0072] In this embodiment, a data preprocessing unit and a data cleaning unit are provided in the visibility forecast optimization module.
[0073] The data preprocessing unit is used to focus on low-visibility weather during the model training process, and it mainly adopts classification learning tasks to minimize the impact of high-visibility samples on the model.
[0074] The data preprocessing unit is configured to perform the following process: according to the distribution of visibility and business needs, visibility is divided into multiple levels according to the size of visibility - for example, visibility can be divided into 5 levels from 0 to 4, and then different weight coefficients are configured for data samples of different visibility levels, wherein a larger weight coefficient is configured for low visibility samples than for high visibility samples to increase the proportion of low visibility samples in the loss function, so that the model training is inclined to low visibility events; and, multi-year observation data and model forecast data are divided into relatively independent training sets, validation sets and test sets, wherein the data of the most recent year is used as the test set, and the data of the previous year is sampled according to the preset ratio by using the segmented interval method to divide the training set and the validation set, and the independence between the data sets is guaranteed when the data distribution is consistent. The preset ratio can be set by the system default or personalized by the user.
[0075] As an example of a typical method, the weight coefficient configuration information of each visibility level can be shown in the following table.
[0076]
[0077] It can be seen that low-visibility samples are configured with a larger weight coefficient than high-visibility samples, thereby increasing the proportion of low-visibility samples in the loss function, making the model lean towards low-visibility events during training, thereby learning more low-visibility weather features. During the training process, the evaluation and selection criteria of the model are also mainly aimed at low-visibility events. As an example, combined with the above table, the evaluation and selection can mainly focus on the first two types of events (visibility ≤1km and visibility between 1-3km).
[0078] When configuring the dataset information of the training set, validation set, and test set, in order to better independently test the generalization ability of the model, the observation data and pattern data of multiple years (such as the observation data and pattern data from 2019 to 2023) can be divided into relatively independent training sets, validation sets, and test sets. As an example of a typical method, for example, from January to November 2023 is used as the test set, and the samples from 2019 to 2022 are divided into the training set and validation set by using the segmented interval method with a 3:1 sampling ratio, so as to make the datasets as independent as possible while ensuring the consistency of data distribution.
[0079] The data cleaning unit is used to alleviate the problem of data imbalance. It performs data cleaning on the training set and the validation set. The main idea is to remove high-visibility samples that meet certain conditions, thereby reducing the proportion of high-visibility samples that the business does not care about in the training samples. The cleaning process can be divided into two steps. Specifically, the data cleaning unit is configured as follows: first, according to the designed daily start time, only the data samples in which low visibility events occurred at any station on the day are retained; then, according to the observed data and the model forecast data, the high-visibility samples with correct model forecasts are removed, wherein the high-visibility samples in the training set and the validation set are removed according to a preset ratio.
[0080] As an example but not a limitation, if 8 pm to 7 pm the next day is counted as one day, first only samples of days with visibility < 3 km at any station on that day are retained. Then, samples with visibility greater than 15 km (the preset visibility threshold) in both observation data and central grid forecast data of the two models are eliminated, that is, samples with high visibility greater than 15 km that are correctly predicted by the model are eliminated, of which 30,000 and 10,000 samples that meet the conditions are eliminated from the training set and validation set respectively.
[0081] It should be noted that the specific values of the daily start time, visibility threshold, and preset ratio configured in the above implementation manner can be adjusted by those skilled in the art according to actual needs and should not be regarded as limitations to the present invention.
[0082] The cloud height and cloud amount forecast optimization module is configured as follows: based on the three-dimensional grid forecast field products of meteorological elements generated by numerical forecast tools and the Metar messages of airport observations, using a preset machine learning model, developing optimized cloud amount classification forecast model algorithms and cloud base height forecast model algorithms, and providing cloud amount classification forecast products and cloud height regression products with a preset time resolution.
[0083] In this embodiment, in the cloud height and cloud amount forecast optimization module, the machine learning model preferably adopts an XGBoost model and a 4-layer fully connected neural network model.
[0084] The input features of the model can include model forecast data, observation data, and external spatiotemporal background feature information.
[0085] The observation data mainly include airport observation cloud high cloud information provided by Metar data.
[0086] The model forecast is preferably a three-dimensional grid forecast field product of meteorological elements obtained by analyzing and processing the East China regional business model CMA-SH9. The ground forecast variables include 2m temperature, 2m relative humidity, precipitation, 10m meridional wind (V) and zonal wind (U), and cloud height and cloud amount diagnosed by the algorithm; the multi-layer forecast includes relative humidity, meridional wind (V) and zonal wind (U), and temperature. According to the station height, it is processed as an interval of 300 meters above the ground, and the top layer height is set to 3000 meters.
[0087] The above data are spliced according to time, with observations as learning targets and aviation meteorological data as learning factors.
[0088] Since cloud height and cloud amount have strong spatial locality and temporal variation characteristics such as daily variation and seasonal variation, in order to improve the forecasting skills of the model, external spatiotemporal background feature information is added to the input features of the model, including external temporal feature information and spatial feature information. By configuring external temporal feature information and spatial feature information in the input features, the temporal variation of cloud height and cloud amount, as well as the different characteristics of cloud height and cloud amount changes at different sites can be identified. Among them, the temporal feature information is the forecast step of the model, the day is the day of the year, and the hour is the hour of the day. Spatial feature information can include latitude, longitude and altitude information.
[0089] At this time, the cloud height and cloud amount forecast optimization module is configured to perform the following steps: configure the input feature information of the model; based on the input features, train the XGBoost model, configure the data set information of the training set, validation set and test set during model training, tune the model parameters through evaluation and analysis of the test set, and run the optimized forecast model (or optimal forecast model) finally tuned in real time, input pattern forecast data every day, and output hourly cloud amount classification forecast products; and, based on the input features, train the 4-layer fully connected neural network model, configure the data set information of the training set, validation set and test set during model training, tune the model parameters through evaluation and analysis of the test set, and run the optimized forecast model (or optimal forecast model) finally tuned in real time, input pattern forecast data every day, and output hourly cloud height regression products.
[0090] During model training, the Metar message contains multiple layers of records about cloud-like cloud cover. The bottom layer of record information is selected, and multiple visibility levels are coded according to the cloud cover ratio, for example, 0-4 coding. The cloud cover is capped at 3,000 meters (top layer height). The observations and forecasts above 3,000 meters are set to 3,000 and normalized to 3,000. The training set can be configured as the 2019-2022 dataset, and the validation set is configured as the 2023 dataset.
[0091] As an example of a typical method, the configuration information of 0-4 encoding can be shown in the following table.
[0092]
[0093] In this embodiment, when performing route planning analysis, the analysis can be performed based on the focus of aviation flight. The focus of aviation flight mainly includes the weather conditions of the take-off and landing airports and the flight conditions during the flight. Among them, the take-off and landing airports mainly focus on the visibility conditions, cloud base height and weather type of the airport area. Through the visibility optimization algorithm and model output results, it is possible to intuitively display the changes in weather conditions of the take-off and landing airports over time. The changes in humidity, ice accumulation, turbulence, etc. at different altitudes during the flight are the focus of attention, which directly affects the stability and safety of aviation flight. Therefore, a route planning analysis solution that simultaneously refers to time and space is provided based on model forecast data.
[0094] Specifically, a route planning spatial analysis module and a route planning temporal analysis module may be configured to respectively perform route planning spatial analysis and route planning temporal analysis to obtain the weather conditions of any waypoint on the planned route at any time. The meteorological element interpolation technology includes meteorological element spatial interpolation and meteorological element temporal interpolation.
[0095] The route planning spatial analysis module is configured to: obtain route information of each waypoint during the route flight, and interpolate the grid point data of the forecast data to the required route position; wherein, the inverse distance weighted method is used to interpolate the humidity, ice accumulation index and turbulence index to the waypoints, and the four points closest to the required position are selected as sample points of the inverse distance weighted method; in the route planning analysis, the grid data of the pattern is first interpolated to the waypoints selected by the route, and then the time analysis is performed.
[0096] The route planning time analysis module is configured as follows: according to the continuity of flight conditions, a time interpolation method is used to interpolate meteorological elements with longer time intervals to obtain data at required times; wherein, for humidity, icing index and turbulence index during route flight, a quadratic spline interpolation method is used to obtain data at required times.
[0097] Through the above route planning space and time analysis, the weather conditions at any point and any time on the custom route can be obtained. Custom route analysis can be performed based on this technology, specifically, through a custom route analysis module.
[0098] The custom route analysis module is configured to perform the following steps: (1) input the latitude and longitude information of each waypoint on the route into the system by coordinate input or map point selection to complete the route preset; (2) after selecting the preset route, input the starting flight time of the route. Subsequently, the system will calculate the time of passing each waypoint according to the domestic civil aviation average reference speed of 450 knots and the distance between the waypoints by default. As needed, the user can also individually set and modify the time of passing each waypoint; (3) through the route planning spatial analysis module, the humidity, turbulence index, and ice accumulation index of the aviation weather forecast data are interpolated to each waypoint, and the route planning time analysis module is used to obtain the aviation flight conditions when the route passes each waypoint; (4) the vertical information of the weather conditions of each waypoint in the route is displayed in the form of a profile diagram to provide a reference for route planning.
[0099] By way of example and not limitation, Figure 4 This example shows how to preset a route by inputting the longitude and latitude information of each waypoint on the route into the system through coordinate input. After the route is preset, the system calculates the time to pass each waypoint based on the average reference speed of 450 knots for domestic civil aviation and the distance between waypoints ( Figure 4 (not shown). Subsequently, the humidity, turbulence index, and ice accumulation index of the aviation weather forecast data are interpolated to each waypoint through the route planning spatial analysis module, and the aviation flight conditions (humidity, ice accumulation, turbulence, etc.) when the route passes through each waypoint are obtained using the route planning time analysis module. Then, the vertical information of the weather conditions of each waypoint in the route is displayed in the form of a profile diagram as needed, such as Figure 5 An example of a route turbulence profile is provided to provide a reference for route planning.
[0100] In this embodiment, the weather forecast and regional display of airports, waypoints and control areas can also be performed according to needs. For example, the temperature forecast information of the airport can be displayed, see Figure 6 shown.
[0101] Furthermore, the waypoints of the control area (flight control area) can be displayed and weather warnings can be issued.
[0102] Another embodiment of the present invention further provides an aviation flight weather analysis system.
[0103] The system includes a weather forecast big data device for optimizing aviation services and a route planning and analysis device.
[0104] The weather forecast big data device for aviation service optimization is used to construct weather forecast big data for aviation service optimization based on meteorological elements of aviation meteorology concern. The meteorological elements include meteorological elements related to weather phenomena such as visibility obstruction. When constructing the weather forecast big data for aviation service optimization, forecast optimization based on machine learning is performed on meteorological elements related to weather phenomena such as visibility obstruction.
[0105] The route planning and analysis device is used to obtain the planned route information, which includes the waypoint information on the route and the flight time information of the route; and, through the meteorological element interpolation technology and the meteorological forecast big data optimized by the aforementioned aviation service, analyze the aviation meteorological element information of each waypoint and each forecast time on the planned route to obtain the meteorological element forecast data of any waypoint in the entire route process, and visualize the aviation meteorological element forecast data according to the needs of the aviation flight.
[0106] In this embodiment, the weather forecast big data device for aviation service optimization can construct the aforementioned weather forecast big data for aviation service optimization based on the model forecast data provided by the numerical forecast tool. At this time, after obtaining the model forecast data provided by the preset numerical forecast tool, a three-dimensional grid forecast field product of meteorological elements of concern to aviation meteorology is generated based on the model forecast data. The three-dimensional grid forecast field product sets the vertical resolution according to the aviation flight requirements, sets a smaller altitude interval at the low layer below the preset altitude threshold, and sets a larger altitude interval at the high layer above the preset altitude threshold. The three-dimensional grid forecast field product is configured to: provide altitude layer data, and provide customized meteorological element forecast data of concern to aviation meteorology, and provide aviation index forecast data of ice accumulation index and turbulence index, and provide full-plane visualization display information of airports, waypoints and control areas.
[0107] Preferably, the weather forecast big data device for aviation service optimization may include a visibility forecast optimization module and a cloud height and cloud amount forecast optimization module.
[0108] The route planning analysis device may include a route planning space analysis module, a route planning time analysis module and a custom route analysis module.
[0109] For other technical features, please refer to the description of the previous embodiment and will not be repeated here.
[0110] In the above description, the disclosure of the present invention is not intended to limit itself to these aspects. Rather, within the scope of the target protection of the present disclosure, the components can be selectively and operatively combined in any number. In addition, terms such as "include", "encompass" and "have" should be interpreted as inclusive or open by default, rather than exclusive or closed, unless they are explicitly defined to the contrary. All technical, scientific or other terms are consistent with the meanings understood by those skilled in the art, unless they are defined to the contrary. Common terms found in dictionaries should not be interpreted too idealistically or too impractically in the context of relevant technical documents, unless the present disclosure explicitly defines them as such. Any changes and modifications made by a person of ordinary skill in the field of the present invention based on the above disclosure belong to the scope of protection of the claims.
Claims
1. An aviation flight meteorological analysis method, characterized in that: Includes steps: Based on the meteorological elements of concern to aviation meteorology, meteorological forecast big data for aviation service optimization is constructed; wherein the meteorological elements include meteorological elements related to weather phenomena such as visibility obstruction, and when constructing the meteorological forecast big data for aviation service optimization, forecast optimization based on machine learning is performed on meteorological elements related to weather phenomena such as visibility obstruction; Acquire planned route information, the route information including waypoint information on the route and flight time information of the route; Through the interpolation technology of meteorological elements and the meteorological forecast big data optimized by the aforementioned aviation services, the aviation meteorological element information of each waypoint and each forecast time on the planned route is analyzed to obtain the meteorological element forecast data of any waypoint in the entire route process, and the aviation meteorological element forecast data is visualized according to the needs of aviation flight.
2. The method according to claim 1, characterized in that Based on the model forecast data provided by the numerical forecast tool, the meteorological forecast big data for the aforementioned aviation service optimization is constructed. At this time, after obtaining the model forecast data provided by the preset numerical forecast tool, a three-dimensional grid forecast field product of meteorological elements of concern to aviation meteorology is generated based on the model forecast data. The three-dimensional grid forecast field product sets the vertical resolution according to the aviation flight requirements, sets a smaller height interval at the low layer below the preset height threshold, and sets a larger height interval at the high layer above the preset height threshold; The three-dimensional grid forecast field product is configured to provide altitude layer data, customized meteorological element forecast data of aviation meteorological concern, aviation index forecast data of icing index and turbulence index, and full-plane visualization display information of airports, waypoints and control areas.
3. The method according to claim 2, characterized in that The meteorological elements of concern to aviation meteorology also include one or more of temperature, precipitation, snowfall, relative humidity, wind direction and speed, 0°C altitude and low-altitude wind shear.
4. The method according to any one of claims 2 to 3, characterized in that: The meteorological elements related to the visibility obstruction weather phenomenon include visibility and cloud height and cloud amount, and the forecast optimization corresponding to the visibility and cloud height and cloud amount are respectively configured with a visibility forecast optimization module and a cloud height and cloud amount forecast optimization module; The visibility forecast optimization module is configured to: develop a visibility site optimization forecast model algorithm based on the model forecast data provided by the numerical forecast tool, using a preset machine learning model, supplemented by observation data from airport meteorological observation points, and external spatiotemporal background feature information, and provide visibility forecast products with a preset time resolution; The cloud height and cloud amount forecast optimization module is configured as follows: based on the three-dimensional grid forecast field products of meteorological elements generated by numerical forecast tools and the Metar messages of airport observations, using a preset machine learning model, developing optimized cloud amount classification forecast model algorithms and cloud base height forecast model algorithms, and providing cloud amount classification forecast products and cloud height regression products with a preset time resolution.
5. The method according to claim 4, characterized in that In the visibility forecast optimization module, the machine learning model adopts the LightGBM model. At this time, the visibility forecast optimization module is configured to perform the following steps: The input feature information of the configuration model includes model forecast data, site meteorological observation data and pollutant concentration data of the airport meteorological observation point, and external time feature information and space feature information. The time change of visibility and different features of visibility change at different sites are identified by configuring the external time feature information and space feature information in the input feature. The time feature information includes the forecast step, month and hour information of the model, and the space feature information includes the station number and the latitude and longitude information of each site. Train the LightGBM model based on the input features and analyze the importance of each feature; configure the dataset information of the training set, validation set, and test set during model training; The model parameters are tuned through evaluation and analysis of the test set to optimize the forecast model; the final optimized forecast model is run in real time, real-time observation data and model forecast data are input, and hourly site visibility forecast products are output.
6. The method according to claim 5, characterized in that The visibility forecast optimization module is provided with a data preprocessing unit and a data cleaning unit; the data preprocessing unit is used to focus on low visibility weather during the model training process; the data cleaning unit is used to alleviate the problem of data imbalance, and remove high visibility samples that meet the screening conditions in the training set and the validation set to reduce the proportion of high visibility samples in the sample; The data preprocessing unit is configured to: divide visibility into multiple levels according to visibility distribution and business needs, configure different weight coefficients for data samples of different visibility levels, wherein a larger weight coefficient is configured for low visibility samples than for high visibility samples to increase the proportion of low visibility samples in the loss function, so that the model training is inclined to low visibility events; and divide multi-year observation data and model forecast data into relatively independent training sets, validation sets and test sets, wherein the data of the most recent year is used as the test set, and the data of the previous year is sampled according to a preset ratio using a segmented interval method to divide the training set and the validation set; The data cleaning unit is configured as follows: first, according to the designed daily start time, only data samples in which low visibility events occurred at any station on that day are retained; then, according to the observation data and model forecast data, high visibility samples with correct model forecasts are eliminated, wherein high visibility samples in the training set and the validation set are eliminated according to a preset ratio.
7. The method according to claim 4, characterized in that In the cloud height and cloud amount forecast optimization module, the machine learning model adopts the XGBoost model and the 4-layer fully connected neural network model. At this time, the cloud height and cloud amount forecast optimization module is configured to perform the following steps: The input feature information of the configuration model includes the three-dimensional grid forecast field product of meteorological elements, the airport observation cloud height and cloud amount information provided by the Metar message, and the external time feature information and spatial feature information. The time change of cloud height and cloud amount, as well as the different characteristics of cloud height and cloud amount changes at different stations are identified by configuring the external time feature information and spatial feature information in the input feature; wherein the time feature information is the forecast step of the model, the day is the day of the year, and the hour is the hour of the day; the spatial feature information includes latitude, longitude and altitude information; Based on the input features, the XGBoost model is trained. During model training, the dataset information of the training set, validation set, and test set is configured. The model parameters are tuned through evaluation and analysis of the test set. The optimized forecast model finally tuned is run in real time, and the pattern forecast data is input daily to output hourly cloud classification forecast products. Also, based on the input features, a 4-layer fully connected neural network model is trained. During model training, the dataset information of the training set, validation set, and test set is configured. During model training, the model parameters are tuned through evaluation and analysis of the test set. The optimized forecast model finally tuned is run in real time, and the pattern forecast data is input daily to output hourly cloud height regression products.
8. The method according to claim 4, characterized in that In the visibility forecast optimization module, the model forecast data used include the model forecast data of the East China regional business model CMA-SH9 and the weather-atmosphere environment coupling model; In the cloud height and cloud amount forecast optimization module, the model forecast data used is the three-dimensional grid forecast field product of meteorological elements obtained by analyzing and processing the East China regional business model CMA-SH9; the Metar message records cloud shape, cloud height and cloud amount data, and there are multiple layers of records about cloud shape, cloud height and cloud amount. The bottom layer of record information is selected, and multi-level encoding is performed according to the cloud amount ratio; and, according to the set top layer height, the cloud amount observation and prediction values of the height layer greater than the top layer height are set to correspond to the top layer height, and normalized according to the top layer height.
9. The method according to claim 1, characterized in that: It includes a route planning space analysis module and a route planning time analysis module, which are used to perform route planning space analysis and route planning time analysis respectively to obtain the weather conditions of any waypoint on the planned route at any time; the meteorological element interpolation technology includes meteorological element space interpolation and meteorological element time interpolation; The route planning spatial analysis module is configured to: obtain route information of each waypoint during the route flight, and interpolate the grid point data of the forecast data to the required route position; wherein, the humidity, ice accumulation index and turbulence index at the waypoint are interpolated by using the inverse distance weighted method, and the four points closest to the required position are selected as sample points of the inverse distance weighted method; in the route planning analysis, the grid data of the pattern is first interpolated to the waypoint selected by the route, and then the time analysis is performed; The route planning time analysis module is configured as follows: according to the continuity of flight conditions, a time interpolation method is used to interpolate meteorological elements with longer time intervals to obtain data at required times; wherein, for humidity, icing index and turbulence index during route flight, a quadratic spline interpolation method is used to obtain data at required times.
10. The method according to claim 9, characterized in that The system also includes a custom route analysis module, which is configured to perform the following steps: Enter the longitude and latitude information of each waypoint on the route into the system by coordinate input or map selection to complete the route preset; After selecting a preset route, enter the route start flight time, and then obtain the time of each waypoint passing through the route based on the route start flight time; the time of each waypoint passing through the route is calculated by default according to the average reference speed of domestic civil aviation and the distance between waypoints, or the time of each waypoint passing through the route can be modified separately according to the user's settings; The humidity, turbulence index and ice accumulation index of aviation weather forecast data are interpolated to each waypoint through the route planning spatial analysis module, and the aviation flight conditions when the route passes through each waypoint are obtained using the route planning time analysis module; The vertical information of weather conditions at each waypoint in the route is displayed in the form of a profile.
11. An aviation flight weather analysis system, characterized in that The system comprises: A weather forecast big data device for aviation service optimization is used to construct weather forecast big data for aviation service optimization based on meteorological elements of aviation meteorology concern; wherein the meteorological elements include meteorological elements related to weather phenomena such as visibility obstruction, and when constructing the weather forecast big data for aviation service optimization, forecast optimization based on machine learning is performed for meteorological elements related to weather phenomena such as visibility obstruction; A route planning and analysis device is used to obtain planned route information, the route information including waypoint information on the route and flight time information of the route; and, through meteorological element interpolation technology and the aforementioned aviation service optimized weather forecast big data, analyze the aviation meteorological element information of each waypoint on the planned route and each forecast time to obtain the meteorological element forecast data of any waypoint in the entire route process, and visualize the aviation meteorological element forecast data according to the needs of aviation flight.
Citation Information
Patent Citations
Single-point meteorological element determining method based on Beidou positioning
CN107367773A
Meteorological numerical model element post-forecast correction method and device based on machine learning
CN114545528A
Complex underlying surface meteorological data grid analysis method and system based on big data and machine learning
CN116451801A
Helicopter route planning method based on weather forecast information
CN116659512A
Aircraft weather prediction system and demonstration method
CN117420617A
Cited By
Aircraft navigation-oriented meteorological display alarm method, system, equipment and medium
CN120847915A
Low-altitude turbulence forecasting and route planning method based on machine learning
CN121920226A
