High-altitude wind field hourly forecast circulation assimilation correction method for guaranteeing safety of air route of aircraft

Abstract

The invention discloses a high-altitude wind field hourly forecasting circulation assimilation correction method for guaranteeing the safety of an air route of an aircraft, belongs to the field of aerospace meteorology, and solves the problem of how to improve the vertical layering fineness and resolution in a WRF mode forecasting result. Comprising the following steps: carrying out observation data preprocessing and quality control, and setting a background field error covariance and an observation error covariance; constructing a detection data assimilation scheme; judging weather conditions at an initial moment, and inputting observation data bound at a corresponding moment for assimilation to obtain an observation field; assimilating the observation field and the background field to generate an analysis field, and starting mode numerical forecasting by using the analysis field to obtain a high-altitude wind field hourly forecasting result; comparing with observation data, correcting and evaluating a high-altitude wind field hourly forecast result, and adjusting a detection data assimilation scheme to carry out cyclic assimilation correction until a forecast result meeting a preset standard of an aircraft route and an optimal detection data assimilation scheme are obtained; the accuracy of high-altitude wind forecasting is improved, and the safety of the air route of the aircraft is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of aerospace meteorology, and relates to a method for hourly forecasting cyclic assimilation correction of high-altitude wind fields to ensure the safety of aircraft routes. Specifically, it is a method for hourly forecasting cyclic assimilation correction of high-altitude winds with a 250m layer-by-layer stratification below 20km. Background Technique

[0002] Numerical weather forecasting plays a very important role in the safety of aircraft routes. Precise route meteorological information is crucial for the safety and accurate trajectory of aircraft. Especially when a launch vehicle passes through the atmosphere, the wind load brought by high-altitude strong winds will affect the flight attitude of the rocket and have a certain impact on the structural reliability of the rocket.

[0003] The current mainstream regional model WRF (The Weather Research and Forecasting Model) can give relatively accurate aviation weather forecasts. However, with the continuous improvement of flight accuracy, the influence of meteorological factors on the flight trajectory route becomes more and more significant. China's new generation of launch vehicles has put forward higher requirements for high-altitude wind meteorological support, and requires the whole-layer wind field based on altitude distribution over a specific area at a specific moment, with high vertical resolution requirements.

[0004] There are problems in the high-altitude wind forecast results of the WRF model, such as insufficiently fine vertical stratification and insufficient resolution, which cannot accurately describe the detailed information of the high-altitude wind distribution with height, and the vertical stratification does not meet the requirements for aircraft route support. Summary of the Invention

[0005] In order to solve the problems of insufficiently fine vertical stratification and insufficient resolution in the high-altitude wind forecast results of the WRF model over a specific area at a specific moment; and the large differences in the assimilation effects of different assimilation methods and different observational data under different weather conditions, the present invention provides a method for hourly forecasting cyclic assimilation correction of high-altitude wind fields to ensure the safety of aircraft routes. By preprocessing and quality controlling the observational data, estimating the background field error, judging the weather conditions of the initial moment of the WRF model through the surface observational data in the observational data, and using three types of multi-source heterogeneous data, namely microwave radiometer data, wind profiler radar data, and Beidou navigation high-altitude wind detection data in the observational data for three-dimensional variational assimilation, a numerical forecasting system with the WRF model as the core is built to achieve hourly forecasting of high-altitude winds with a 250m layer-by-layer stratification below 20km over a specific area at a specific moment using different combinations of multi-source heterogeneous data under different weather conditions. This method improves the accuracy of high-altitude wind forecasting and ensures the safety of aircraft routes; this technical solution is applicable to cyclic assimilation correction in the process of hourly forecasting of high-resolution high-altitude winds.

[0006] The object of the present invention is specifically realized through the following technical solutions:

[0007] The present invention discloses a method for hourly forecasting cyclic assimilation correction of upper-air wind fields to ensure the safety of the flight path of an aircraft. The method includes:

[0008] Step 1: Collect the observation data for the required time period in the mission area. The observation data is measured multi-source heterogeneous data and ground observation data. After quality control of the historical extreme values and background field deviations of the observation data, it is converted into a format that can be processed by the obsproc module in the WRF model. At the same time, download the numerical forecast model product data for the same time period to obtain the background field;

[0009] Step 2: Construct a background field error covariance estimation model by using the background field error information pre-generated by the WRF model by default and the global background error covariance matrix CV3 provided by NCEP and included in the WRF model itself;

[0010] Step 3: Bind the multi-source heterogeneous data with the highest hourly forecast result accuracy of the historical upper-air wind fields under different weather conditions to construct a sounding data assimilation scheme. Estimate the observation error covariance through the obsproc module and set the observation error covariance parameters. Estimate the background field error covariance through the background field error covariance estimation model and set the background field error covariance parameters. Judge the weather conditions at the initial time of the WRF model based on the measured ground observation data corresponding to the mission area. Based on the sounding data assimilation scheme, input the multi-source heterogeneous data bound at the time corresponding to the weather conditions at the initial time into WRFDA, and use the three-dimensional variational assimilation method 3DVAR for assimilation to obtain the observation field required by the WRF model;

[0011] Step 4: Use the three-dimensional variational assimilation method 3DVAR to assimilate the observation field and the background field to generate an analysis field, and use the analysis field to start the numerical forecast of the WRF model to obtain the hourly forecast result of the upper-air wind field;

[0012] Step 5: Compare the hourly forecast result of the upper-air wind field with the observation data to obtain the accuracy rate of the hourly forecast result of the upper-air wind field. If the accuracy rate does not meet the preset standard of the flight path of the aircraft, then use the two-factor rolling MOS correction method to correct and evaluate the hourly forecast result of the upper-air wind field, adjust the combination method of the multi-source heterogeneous data, update the sounding data assimilation scheme, and return to Step 3 for cyclic assimilation correction; until the accuracy rate of the hourly forecast result of the upper-air wind field that meets the preset standard of the flight path of the aircraft is obtained. At this time, the corresponding sounding data assimilation scheme is the optimal sounding data assimilation scheme; if the accuracy rate of the hourly forecast result of the upper-air wind field that meets the preset standard of the flight path of the aircraft cannot be obtained, then select the sounding data assimilation scheme corresponding to the highest accuracy rate of the hourly forecast result of the upper-air wind field as the optimal sounding data assimilation scheme.

[0013] In Step 2, the constructed background field error covariance estimation model is:

[0014] B = (x background - x true )(x background - x true ) T ;

[0015] B 1 / 2 = U p U v U h ;

[0016] In the formula, B is the background field error covariance, which is a symmetric and positive semi - definite square matrix, x background is the background field, x true is the true field, and T is the transpose of the matrix; U p U v U h Adopt the background field error information pre - generated by the default of the WRF model, and use the global background error covariance matrix CV3 provided by NCEP and built - in in the WRF model to achieve the optimal fusion of the background field and the observed data. Among them, U p is the physical transformation or balance operator, U v is the vertical transformation, and U h is the horizontal transformation.

[0017] In step three, in the constructed sounding data assimilation scheme,

[0018] The different weather conditions are: four types of weather conditions: few clouds on sunny days, cloudy, light rain on cloudy days, and thunderstorm with strong wind;

[0019] The multi - source heterogeneous data are: Beidou radiosonde data, wind profiler radar data, and microwave radiometer data; among them, the combination methods of the multi - source heterogeneous data to be bound under different weather conditions include: no observed data, Beidou radiosonde data, Beidou radiosonde data combined with wind profiler radar data, Beidou radiosonde data combined with microwave radiometer data, wind profiler radar data combined with microwave radiometer data, and Beidou radiosonde data combined with wind profiler radar data and microwave radiometer data, a total of six combination methods.

[0020] In step three, the constructed sounding data assimilation scheme is:

[0021] When the weather condition is few clouds on sunny days, it is bound to the Beidou radiosonde data;

[0022] When the weather condition is cloudy, it is bound to the Beidou radiosonde data combined with the wind profiler radar data;

[0023] When the weather condition is light rain on cloudy days, it is bound to the Beidou radiosonde data combined with the microwave radiometer data;

[0024] When the weather condition is thunderstorm with strong wind, it is bound to the Beidou radiosonde data combined with the microwave radiometer data.

[0025] In step 3, the method for judging the weather conditions at the initial time of the WRF model from the measured surface observation data corresponding to the task area is as follows:

[0026] If the cloud cover in the measured surface observation data corresponding to the task area is ≤ 3 tenths, it is judged as few clouds on a sunny day;

[0027] If 3 tenths < cloud cover < 8 tenths, it is judged as cloudy;

[0028] If the cloud cover ≥ 8 tenths, or the rainfall in 1 hour ≤ 2.5 mm, it is judged as light rain on a cloudy day;

[0029] If the rainfall in 1 hour > 2.5 mm, or the average wind speed > 8 m / s, it is judged as thunderstorm with strong wind.

[0030] Among them, the objective function of the 3DVAR (Three-Dimensional Variational Assimilation Method) is:

[0031]

[0032] Among them, J(x) is the solution of the cost function, x represents the analysis variable on the WRF model grid points, x background is the background field, B represents the background field error covariance, y is the observation data, y o represents the observation operator, R represents the observation error covariance, and T represents the transpose of the matrix.

[0033] In step 5, the method for obtaining the hourly forecast result accuracy rate of the upper-air wind field by comparing the hourly forecast result of the upper-air wind field with the observation data is as follows:

[0034]

[0035] Among them,

[0036]

[0037]

[0038] In the formula, f w represents the hourly forecast result accuracy rate of the upper-air wind field obtained from the multi-source heterogeneous data bound by the weather condition w; among them, w is the weather condition; a is the east-west wind weight coefficient, b is the north-south wind weight coefficient; f mu represents the accuracy rate of the east-west wind in the hourly forecast result of the upper-air wind field under m, and m represents the detection data assimilation scheme; f mv represents the accuracy rate of the north-south wind in the hourly forecast result of the upper-air wind field under m; N represents the judgment result of whether the forecast result conforms to the wind preset standard of the aircraft route; i represents the i-th layer height data; n represents the total number of layers of the forecast wind product; u mi represents the east-west wind at the i-th layer height in the hourly forecast result of the upper-air wind field under m, uoi represents the east-west wind at the i-th upper-air level of the observed data; v mi represents the north-south wind at the i-th height level in the hourly forecast result of the upper-air wind field at m; v oi represents the north-south wind at the i-th upper-air level of the observed data.

[0039] In step five, the method for correcting and evaluating the hourly forecast result of the upper-air wind field using the two-factor rolling MOS correction method is as follows:

[0040] y' t+h = b1y t+h + b2(O t - Y t ) + b0

[0041] In the formula, y' t+h is the corrected value at time t + h, y t+h is the forecast value at time t + h, O t is the grid actual situation at time t, Y t is the forecast value at time t; b0, b1, and b2 are the first model coefficient, the second model coefficient, and the third model coefficient automatically established using the least squares method, t is the observed data time, h is the difference between the correction time and the observed data time, and h = 1, 2,..., 24.

[0042] In step five, when the accuracy rate of the hourly forecast result of the upper-air wind field that meets the preset standard of the aircraft route cannot be obtained, the method for selecting the optimal detection data assimilation scheme as the detection data assimilation scheme corresponding to the highest accuracy rate of the hourly forecast result of the upper-air wind field is as follows:

[0043]

[0044] If u mi - u oi ≤ u s , then is 1; if u mi - u oi > u s , then is 0;

[0045] If v mi - v oi ≤ v s , then is 1; if v mi - v oi > v s , then is 0;

[0046] In the formula, f' wDenote the maximum accuracy rate of the hourly forecast results of the upper-air wind field under weather condition w; f' mu Denote the average accuracy rate of the east-west wind in the upper air under the sounding data assimilation scheme m; f' mv Denote the average accuracy rate of the north-south wind in the upper air under the sounding data assimilation scheme m; u s Is the judgment standard for the east-west wind in the preset standard of the aircraft route; v s Is the judgment standard for the north-south wind in the preset standard of the aircraft route; f s Is the percentage standard in the preset standard of the aircraft route; K 总 Denote the total number of samples under weather condition w; k denotes the sample number under the sounding data assimilation scheme m.

[0047] In step five, the preset standard of the aircraft route is:

[0048] Under different weather conditions, the hourly forecast accuracy rate of the upper-air wind stratified every 250 m below 20 km in the overall assimilated data forecast is: for the east-west wind component and the north-south wind component, when the wind speed ≤ the judgment standard, the accuracy rate Accuracy ≥ the percentage standard.

[0049] The beneficial effects of the present invention are:

[0050] The technical solution disclosed by the present invention solves the problems of insufficient vertical stratification and low resolution in the upper-air wind forecast results of the WRF model over a specific area at a specific moment. By using the characteristics of the measured Beidou sounding data, wind profiler radar data, and microwave radiometer data in the multi-source heterogeneous data, which have stable and reliable quality, high vertical detection accuracy, can provide a complete description of the three-dimensional structure of the atmosphere, are one of the most basic data for numerical model forecasting, and are specimens for testing the forecasting effect of numerical models and the credibility of other observational data, the present invention inputs them into the numerical forecasting model for assimilation to improve the forecasting accuracy of the atmospheric temperature, humidity, and wind profiles and enhance the accuracy of weather forecasting.

[0051] The present invention also solves the problem that there are significant differences in the effects of different assimilation methods and the assimilation of different observational data under different weather conditions; in view of the forecasting guarantee requirements for high-resolution upper-air winds in aerospace meteorological support, using various multi-source heterogeneous data combination methods of measured Beidou sounding data, wind profiler radar data, and microwave radiometer data in the observational data, through preprocessing and quality control of the observational data, background field error estimation, and three-dimensional variational assimilation, a numerical forecasting system with the WRF model as the core is built. Combining local meteorological data, that is, binding the observational data with the highest hourly forecasting result accuracy of the historical upper-air wind field under different weather conditions, designing an assimilation scheme for sounding data under different weather conditions, and judging the weather conditions at the initial time of the WRF model through the measured surface observational data in the observational data. Input the observational data bound at the corresponding time of the initial time weather conditions into WRFDA for assimilation, and assimilate the observational field and the background field to generate an analysis field. Use the analysis field to start the numerical forecasting of the WRF model to obtain the hourly forecasting result of the upper-air wind field. If the accuracy of the forecasting result does not meet the preset standard of the aircraft route, the double-factor rolling MOS correction method is used to correct and evaluate the hourly forecasting result of the upper-air wind field, and adjust the combination method of multi-source heterogeneous data in the sounding data assimilation scheme, update the sounding data assimilation scheme for cyclic assimilation correction. Through the preset standard of the aircraft route that distinguishes east-west wind and north-south wind, the hourly forecasting of the upper-air wind is realized for each 250 m layer below 20 km above a specific area at a specific time, improving the accuracy of the whole-layer forecasting of the upper-air wind at a specific time and a specific area in the mission. The vertical resolution requirement is high, and it is applicable to the hourly forecasting of high-resolution upper-air winds. After experiments, the present invention can meet the standard of the east-west wind component and the north-south wind component wind speed ≤ 5 m / s (95%), accurately describe the detailed information of the upper-air wind distribution with height, and the vertical stratification meets the requirements of aircraft route guarantee, ensuring the safety of the aircraft route. Brief Description of the Drawings

[0052] The present invention will be further described in detail below with reference to the drawings and embodiments.

[0053] Figure 1 It is a schematic diagram of the design process of the assimilation scheme for sounding data under different weather conditions.

[0054] Figure 2 It is a schematic diagram of the cyclic assimilation process of the numerical model hourly forecasting.

[0055] Figure 3 It is the forecasting accuracy of the wind field from March to April in 2024; among them, each point is the accuracy of the comparison between the forecasting value and the observational value on that day. The red points are the accuracy of the eastward wind, and the blue points are the accuracy of the northward wind. The average accuracy of the eastward wind and the northward wind in 60 days is 95.48% and 95.14% respectively.

[0056] Figure 4is the wind field forecast accuracy rate on March 4, 2024; among them, the red line is the forecast result of the WRF model, the green line is the observed value, the accuracy rate of U wind is 94.05%, and the accuracy rate of V wind is 96.43%.

[0057] Figure 5 is the average accuracy rate of east-west winds under different assimilation schemes under different weather conditions from March to April 2024.

[0058] Figure 6 is the average accuracy rate of north-south winds under different assimilation schemes under different weather conditions from March to April 2024. Specific implementation manner

[0059] A high-altitude wind field hourly forecast cycle assimilation correction method for ensuring the safety of the aircraft route provided by the embodiment of the present invention mainly includes: preprocessing and quality control of observation data, estimation of background field error covariance, design of assimilation schemes under different weather conditions, design of the forecast process and result correction; specifically, the method includes:

[0060] Step 1, preprocessing and quality control of observation data: Collect observation data for the required time period in the mission area, and the observation data is measured multi-source heterogeneous data and ground observation data; after quality control of the historical extreme value and background field deviation of the observation data, convert it into a format that can be processed by the obsproc module in the WRF model; at the same time, download numerical forecast model product data for the same time period to obtain the background field;

[0061] Among them, the multi-source heterogeneous data is: measured Beidou sounding data, wind profiler radar data and microwave radiometer data; the numerical forecast model product data is GFS data, ECMWF data and GRAPES data downloaded from global model data providers.

[0062] In order to enable the observation data to effectively enter the WRFDA assimilation system, it is necessary to preprocess the observation data, perform quality control on the historical extreme value and background field deviation of the observation data, and convert it into the ASCII format that can be processed by obsproc in the WRF model;

[0063] Quality control of the historical extreme value and background field deviation of the observation data is used to identify and eliminate outliers in the observation data, and improve the quality and reliability of the observation data; in step 1, the method for quality control of the historical extreme value and background field deviation of the observation data includes:

[0064] a. Mark the observation data exceeding the historical extreme value range as suspicious values;

[0065] b. Calculate the difference between the observation data and the background field value to obtain the deviation; mark the observation data with the deviation exceeding the threshold as outliers;

[0066] c. Reject outliers or correct according to the background field value.

[0067] It mainly includes:

[0068] 1. Historical extreme value check: If the observed value of the observed data exceeds the historical extreme value range, it is marked as a suspicious value.

[0069] 2. Background field deviation check: Compare the observed value with the background field value, that is, observed value - background field value = deviation; if the deviation exceeds the threshold, it is marked as an outlier.

[0070] 3. Process outliers: After confirming the outliers, you can choose to remove the data point or correct it according to the background field value.

[0071] Through the above steps, outliers in the observed data can be effectively identified and processed, improving the quality and reliability of the data.

[0072] Observed data usually contains errors and needs to be quality-controlled during the assimilation process. Estimate the observation error covariance through the obsproc module and set the observation error covariance parameters; the obsproc module is mainly implemented in FORTRAN compilation and includes the following steps:

[0073] (1) Remove the observed values outside the specified time and space domain;

[0074] (2) Reorder and merge duplicate data reports;

[0075] (3) Calculate the air pressure or altitude according to the hydrostatic relationship and the observation information;

[0076] (4) Check the vertical consistency and superadiabatic conditions of multi-level observations;

[0077] (5) Estimate the observation error according to the pre-given error file;

[0078] (6) Write the upper-air sounding data file in the ASCII format recognized by the assimilation system;

[0079] The multi-source heterogeneous data selected in the present invention is stable, reliable, and has high vertical detection accuracy. It can provide a complete description of the three-dimensional atmospheric structure, which is one of the most basic data for numerical model forecasting and also a specimen for testing the forecasting effect of numerical models and the credibility of other observational data. Assimilating upper-air meteorological sounding data in numerical forecasting models can improve the forecasting accuracy of atmospheric temperature, humidity, and wind profiles and enhance the accuracy of weather forecasting. However, there are significant differences in the assimilation effects of different assimilation methods and different multi-source heterogeneous data under different weather conditions. Evaluating the sounding data can establish data assimilation schemes for different weather conditions. Taking the observational data of the local meteorological observatory on March 3, 2024 as an example; the numerical forecasting model product data is GFS data, ECMWF data, and GRAPES data downloaded from global model data providers. The technical solution of the present invention will be described taking the GFS data as an example.

[0080] Step 2, background field error covariance estimation: Use the background field error information pre-generated by the default of the WRF model and the global background error covariance matrix CV3 provided by NCEP and included in the WRF model to construct a background field error estimation model. Use a simplified model to estimate the background field error covariance and set the background field error covariance parameters.

[0081] There are correlation relationships among the variables in the WRF model due to various dynamic constraints, and these relationships also exist among their background errors. Coupled with the autocorrelation among the same variables, the matrix dimension reaches 107×107 and the structure is extremely complex. Inverting such a super-large matrix is extremely difficult, and it is solved by preconditioning through controlled variable transformation in actual operations.

[0082] The constructed background field error estimation model is:

[0083] B=(x background -x true )(x background -x true ) T ;

[0084] The method for estimating the background field error covariance using a simplified model is:

[0085] B 1 / 2 =U p U v U h ;

[0086] In the formula, B is the background field error covariance, which is a symmetric positive semi-definite square matrix, x background is the background field, x true is the true field, T is the transpose of the matrix; U p U v U hUsing the background field error information pre-generated by the default of the WRF model and the global background error covariance matrix CV3 provided by NCEP included in the WRF model, the optimal fusion of the background field and the observational data is achieved;

[0087] Among them, U p is a physical transformation or balance operator. The control variables are uncorrelated with each other. Through physical transformation, the control variables are restored to physical variables. The physical transformation actually reflects the cross relationships between physical variables, mainly the balance relationship;

[0088] U v is a vertical transformation. The vertical autocorrelation relationship of each control variable is simulated by isotropic empirical orthogonal functions (EOFs) or by applying a recursive iterative filter;

[0089] U h is a horizontal transformation that defines the horizontal autocorrelation relationship of the control variables and is simulated by continuously applying a horizontal recursive filter (approximate to a diffusion operator because it is affordable). Alternative methods include a diffusion operator, a spectral operator, and a wavelet operator (diagonal).

[0090] For the study area, using the background field error information pre-generated by the default of the WRF model and the global background error covariance matrix CV3 provided by NCEP included in the model, CV3 uses vertical recursive filtering to simulate the vertical covariance. Based on the preprocessed upper-air wind sounding data and microwave radiometer observational data on March 3, 2024, the initial background field is adjusted to approach the observed values. This process is regulated by weight parameters to effectively control the influence range and degree of the observational data in the spatial and variable dimensions, achieving the optimal fusion of the background field and the observational data and improving the accuracy of the model initial field.

[0091] Step 3: Bind multi-source heterogeneous data with the highest hourly forecast result accuracy of different weather conditions and historical upper-air wind fields to construct a sounding data assimilation scheme as the initial sounding data assimilation scheme; estimate the observational error covariance through the obsproc module and set the observational error covariance parameters; estimate the background field error covariance through the background field error covariance estimation model and set the background field error covariance parameters; judge the weather conditions at the initial time of the WRF model based on the measured surface observational data corresponding to the task area. Based on the sounding data assimilation scheme, input the multi-source heterogeneous data bound at the corresponding time of the weather conditions at the initial time into WRFDA and use the three-dimensional variational assimilation method 3DVAR for assimilation to obtain the observational field required by the WRF model;

[0092] Among them, the weather conditions at the initial time of the WRF model are judged through local actual ground observation data, and four types of weather, namely few clouds on sunny days, cloudy, light rain on cloudy days, and thunderstorms with strong winds, are distinguished. The combination methods of multi-source heterogeneous data include six combinations: no observation data, Beidou radiosonde data, Beidou radiosonde data combined with wind profiler radar data, Beidou radiosonde data combined with microwave radiometer data, wind profiler radar data combined with microwave radiometer data, and Beidou radiosonde data combined with wind profiler radar data and microwave radiometer data.

[0093] The method for judging the weather conditions at the initial time of the WRF model through the ground observation data corresponding to the mission area is as follows:

[0094] When the cloud amount in the ground observation data of the observation data corresponding to the mission area ≤ 3 tenths, it is judged as few clouds on sunny days;

[0095] When 3 tenths < cloud amount < 8 tenths, it is judged as cloudy;

[0096] When the cloud amount ≥ 8 tenths, or the rainfall in 1 hour ≤ 2.5 mm, it is judged as light rain on cloudy days;

[0097] When the rainfall in 1 hour > 2.5 mm, or the average wind speed > 8 m / s, it is judged as thunderstorms with strong winds.

[0098] The definitions of the four types of weather conditions are shown in Table 1.

[0099] Table 1

[0100]

[0101] In step three, the method for constructing the sounding data assimilation scheme is as follows:

[0102] Statistical historical forecast data, and bind the observation data with the highest accuracy of the hourly forecast results of the historical upper-air wind field for different weather conditions; for example: according to the actual observation data from March to April 2024, through assimilating different observation data for repeated experiments, the experimental results are as Figure 5 and Figure 6 shown. By statistically analyzing all the forecast cases from March to April 2024, the observation data with the highest accuracy of the obtained forecast results is used as the final binding scheme for this weather condition, and the constructed sounding data assimilation scheme includes:

[0103] When the weather condition is few clouds on sunny days, it is bound to the Beidou radiosonde data in the observation data;

[0104] When the weather condition is cloudy, it is bound to the Beidou radiosonde data combined with the wind profiler radar data in the observation data;

[0105] When the weather condition is light rain on cloudy days, it is bound to the Beidou radiosonde data combined with the microwave radiometer data in the observation data;

[0106] When the weather condition is thunderstorm with strong wind, it is combined with the Beidou sounding data in the observation data and the microwave radiometer data is bound.

[0107] According to the local actual ground observation data, on March 3, 2024, it was cloudy with light rain. According to the Figure 1 procedure, the Beidou sounding data and the microwave radiometer data are taken as the assimilation scheme. The corresponding time is 08:00 on March 3, 2024. Data processing is carried out, and then data assimilation is carried out to obtain the observation field.

[0108] Step 4, forecast process design: Using the three-dimensional variational assimilation method 3DVAR, the observation field is assimilated with the background field to generate the analysis field, and the numerical forecast of the WRF model is started using the analysis field to obtain the hourly forecast results of the upper-air wind field;

[0109] 3DVAR mainly solves the minimum value of the objective function. It mainly obtains the analysis field (the solution of the minimum value) to make the best fitting effect between the analysis field and both the background field and the observation field. And at this time, the minimum solution of the objective function is the analysis field; The objective function of the three-dimensional variational assimilation method 3DVAR is:

[0110]

[0111] where J(x) is the solution of the cost function, x represents the analysis variable on the WRF model grid points, x background is the background field, B represents the background field error covariance, y is the observation data, y o represents the observation operator, R represents the observation error covariance, and T represents the transpose of the matrix.

[0112] Preferably, the background field and the observation field are first input into the system. Among them, the background field can be the result of global or regional model forecasts. In this example, the background field is obtained from the WRF model. The observation field is the assimilation data put into the assimilation system. It should be noted that the WRFDA system only recognizes the BUFR (Binary Universal Form for the Representation of meteorological data) format. Before putting the data into WRFDA, the data format needs to be converted. In addition, statistical control (quality control) is also required. The background error covariance and the observation error covariance need to be set. The setting of both is indispensable and will have a great impact on the system simulation effect. During the WRFDA assimilation process, parameterization settings are also required, such as the setting of the assimilation time window. Different parameterization settings will result in different assimilation results. In the process of minimizing the cost function, there are mainly two outer loops, and the number of iterations can be set in the relevant modules of WRFDA according to the data type and total amount. After each loop ends, the forecast result will be updated until the optimal solution, that is, the analysis field, is obtained.

[0113] Step Five, result correction: As Figure 2 shown, compare the hourly forecast result of the upper-air wind field with the observation data to obtain the accuracy rate of the hourly forecast result of the upper-air wind field. If the accuracy rate does not meet the preset standard of the aircraft route, the double-factor rolling MOS correction method is used to correct and evaluate the hourly forecast result of the upper-air wind field, adjust the combination method of multi-source heterogeneous data, update the detection data assimilation scheme, and return to Step Three for cyclic assimilation correction until the accuracy rate of the hourly forecast result of the upper-air wind field that meets the preset standard of the aircraft route is obtained. At this time, the corresponding detection data assimilation scheme is the optimal detection data assimilation scheme. If the accuracy rate of the hourly forecast result of the upper-air wind field that meets the preset standard of the aircraft route cannot be obtained, then select the detection data assimilation scheme corresponding to the highest accuracy rate of the hourly forecast result of the upper-air wind field as the optimal detection data assimilation scheme.

[0114] In Step Five, the method for comparing the hourly forecast result of the upper-air wind field with the observation data to obtain the accuracy rate of the hourly forecast result of the upper-air wind field is:

[0115] Among them,

[0116]

[0117] In the formula, f w represents the accuracy rate of the hourly forecast result of the upper-air wind field obtained from the multi-source heterogeneous data bound by the weather condition w. Among them, w is the weather condition, which distinguishes four types of weather: clear sky with few clouds, cloudy, light rain on cloudy days, and thunderstorm with strong wind; a is the east-west wind weight coefficient, b is the north-south wind weight coefficient; fmu Denotes the accuracy rate of the east-west wind in the hourly forecast results of the upper-air wind field under the sounding data assimilation scheme m, where m represents the sounding data assimilation scheme, and f mv Denotes the accuracy rate of the north-south wind in the hourly forecast results of the upper-air wind field under the sounding data assimilation scheme m; N denotes the judgment result of whether the forecast result of the i-th layer conforms to the wind preset standard of the aircraft route in the hourly forecast results of the upper-air wind field under m; i denotes the height data of the i-th layer; n denotes the total number of layers of the forecast wind products; u mi Denotes the east-west wind at the height of the i-th layer in the hourly forecast results of the upper-air wind field under m, u oi Denotes the east-west wind at the i-th layer of the upper air in the observation data; v mi Denotes the north-south wind at the height of the i-th layer in the hourly forecast results of the upper-air wind field under m; v oi Denotes the north-south wind at the i-th layer of the upper air in the observation data.

[0118] The combination methods of multi-source heterogeneous data to be bound under different weather conditions in the sounding data assimilation scheme m include: six combination methods: no observation data, Beidou radiosonde data, Beidou radiosonde data combined with wind profiler radar data, Beidou radiosonde data combined with microwave radiometer data, wind profiler radar data combined with microwave radiometer data, and Beidou radiosonde data combined with wind profiler radar data and microwave radiometer data.

[0119] In step five, when the accuracy rate of the hourly forecast result of the upper-air wind field that meets the aircraft route preset standard cannot be obtained, the method of selecting the sounding data assimilation scheme corresponding to the highest accuracy rate of the hourly forecast result of the upper-air wind field as the optimal sounding data assimilation scheme is:

[0120]

[0121] If u mi -u oi ≤u s , then is 1; if u mi -u oi >u s , then is 0;

[0122] If v mi -v oi ≤v s , then is 1; if v mi -v oi >v s , then is 0;

[0123] In the formula, f' wIt represents the maximum accuracy rate of the hourly forecast result of the upper-air wind field under weather condition w; f' mu It represents the average accuracy rate of the east-west wind in the upper air under m; f' mv It represents the average accuracy rate of the north-south wind in the upper air under m; N represents whether the forecast result of the i-th layer in the hourly forecast result of the upper-air wind field of the k-th sample under m meets the judgment standard of the wind corresponding to the preset standard of the aircraft route; u s It is the judgment standard of the east-west wind in the preset standard of the aircraft route; v s It is the judgment standard of the north-south wind in the preset standard of the aircraft route; f s It is the percentage standard in the preset standard of the aircraft route; K 总 It represents the total number of samples under weather condition w; k represents the k-th sample data under m, u s and v s The default value is 5 m / s; f s The default value is 95%.

[0124] For example, when the weather condition is thunderstorm with strong wind, making an hourly forecast using the detection data assimilation scheme m for the thunderstorm with strong wind weather condition, and the process of comparing the hourly forecast result of the upper-air wind field with the observed data is a sample.

[0125] This formula realizes the effect confirmation of the detection data assimilation scheme for the updated multi-source heterogeneous data under different weather conditions. During the operation of the system, it realizes the selection of the optimal detection data assimilation scheme under different weather conditions.

[0126] In step five, the preset standard of the aircraft route is as follows:

[0127] Under different weather conditions, the hourly forecast accuracy rate of the upper-air wind with a 250-m stratification every 20 km for the overall assimilated data forecast is: for the east-west wind component and the north-south wind component, the wind speed ≤ the judgment standard, and the accuracy rate Accuracy ≥ the percentage standard.

[0128] Preferably: for the east-west wind component and the north-south wind component, the wind speed ≤ 5 m / s, and the accuracy rate Accuracy ≥ 95%.

[0129] In the embodiment of the present invention, the preset standard of the aircraft route is divided into the east-west wind and the north-south wind, and the accuracy rate of the hourly forecast result of the upper-air wind field is judged respectively, which solves the problems of insufficient vertical stratification and insufficient resolution in the upper-air wind forecast result of the WRF model, improves the accuracy rate of the upper-air wind forecast, and ensures the safety of the aircraft route.

[0130] In step five, the method for correcting and evaluating the hourly forecast result of the upper-air wind field by using the two-factor rolling MOS correction method is as follows:

[0131] y' t+h = b1yt+h +b2(O t -Y t )+b0

[0132] where y' t+h is the correction value at time t + h, y t+h is the forecast value at time t + h, O t is the grid point actual situation at time t, Y t is the forecast value at time t; b0, b1, and b2 are the first model coefficient, the second model coefficient, and the third model coefficient automatically established using the least squares method, t is the observation data time, h is the difference between the correction time and the observation data time, and h = 1, 2,..., 24.

[0133] To enable those skilled in the art to further understand the technical solution of the present invention, two examples are provided for illustration.

[0134] Example 1: Taking the east-west wind component and the north-south wind component wind speed ≤ 5 m / s (95%) as the standard, the forecast results are tested; as Figure 3 shown, it is the wind field forecast accuracy rate from March to April 2024. Among them, each point is the accuracy rate of the comparison between the forecast value and the observed value on that day. The red points are the accuracy rates of the eastward wind, and the blue points are the accuracy rates of the northward wind. The 60-day average accuracy rates of the eastward wind and the northward wind are 95.48% and 95.14% respectively; from Figure 3 it can be obtained that under different weather conditions, the accuracy of the overall assimilated data in forecasting the high-altitude wind every 250 m below 20 km reaches 95.48% for the U wind component and 95.14% for the V wind component, meeting the preset standards for the flight path of the aircraft.

[0135] Example 2: The background field is obtained from the WRF model, and the observation field is the assimilated data result of the above-mentioned observation data at 8:00 on March 3, 2024. After forecasting by the WRF model, the 24-hour forecast result is compared with the actual observation data result at 8:00 on March 4, 2024. The accuracy rate of the U wind component is 94.05%, and the accuracy rate of the V wind component is 96.43%. The forecast result and the actual observation data are shown in Figure 4 , Figure 4 which is the wind field forecast accuracy rate on March 4, 2024; among them, the red line is the forecast result of the WRF model, the green line is the observed value, the U wind accuracy rate is 94.05%, and the V wind accuracy rate is 96.43%. The stratified high-altitude wind forecast every 250 m below 20 km is realized, meeting the preset standards for the flight path of the aircraft.

[0136] The beneficial effects of the embodiments of the present invention are:

[0137] The technical solution disclosed in the embodiments of the present invention solves the problems of insufficient vertical stratification and low resolution in the high-altitude wind forecast results of the WRF model over a specific area at a specific moment. The present invention utilizes the characteristics of the measured Beidou sounding data, wind profiler radar data, and microwave radiometer data in multi-source heterogeneous data, which have stable and reliable quality, high vertical detection accuracy, can provide a complete description of the three-dimensional structure of the atmosphere, are one of the most basic data for numerical model forecasting, and are also specimens for testing the forecasting effect of numerical models and the credibility of other observational data. These data are input into the numerical forecasting model for assimilation to improve the forecasting accuracy of the atmospheric temperature, humidity, and wind profiles and enhance the accuracy of weather forecasting.

[0138] The present invention also solves the problem that there are significant differences in the assimilation effects of different assimilation methods and different observational data under different weather conditions. In response to the forecasting guarantee requirements for high-resolution high-altitude winds in aerospace meteorological support, using various multi-source heterogeneous data combination methods of the measured Beidou sounding data, wind profiler radar data, and microwave radiometer data in the observational data, through preprocessing and quality control of the observational data, background field error estimation, and three-dimensional variational assimilation, a numerical forecasting system with the WRF model as the core is built. Combining local meteorological data, that is, binding the observational data with the highest hourly forecasting result accuracy of the historical high-altitude wind field under different weather conditions, designing the assimilation scheme of the sounding data under different weather conditions, and judging the weather conditions at the initial moment of the WRF model through the measured surface observational data in the observational data. The observational data bound to the corresponding moment of the initial moment weather conditions are input into WRFDA for assimilation, and the observational field and the background field are assimilated to generate an analysis field. The numerical forecasting of the WRF model is started using the analysis field to obtain the hourly forecasting results of the high-altitude wind field. If the accuracy of the forecasting results does not meet the preset standard of the aircraft route, the double-factor rolling MOS correction method is used to correct and evaluate the hourly forecasting results of the high-altitude wind field, and the combination method of the multi-source heterogeneous data in the sounding data assimilation scheme is adjusted, and the sounding data assimilation scheme is updated for cyclic assimilation correction. Through the preset standard of the aircraft route that distinguishes between east-west and north-south winds, the hourly forecasting of the high-altitude wind with a 250m stratification below 20km over a specific area at a specific moment is realized, improving the accuracy of the whole-layer forecasting of the high-altitude wind at a specific moment and in a specific area in the mission, with high vertical resolution requirements, suitable for the hourly forecasting of high-resolution high-altitude winds. After experiments, the present invention can meet the standard of the east-west wind component and the north-south wind component wind speed ≤ 5m / s (95%), accurately describe the detailed information of the high-altitude wind distribution with height, and the vertical stratification meets the requirements of aircraft route guarantee, ensuring the safety of the aircraft route.

[0139] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claimed rights.

Claims

1. A method for hourly forecasting cyclic assimilation correction of upper air wind fields to ensure the safety of aircraft flight routes, characterized in that, The method includes: Step 1: Collect the observation data for the required time period in the task area. The observation data are measured multi-source heterogeneous data and ground observation data. After quality control of the historical extreme values and background field deviations of the observation data, convert them into the format that can be processed by the obsproc module in the WRF model. At the same time, download the numerical prediction model product data for the same time period to obtain the background field. Step 2: Construct a background field error covariance estimation model using the background field error information pre-generated by default in the WRF model and the global background error covariance matrix CV3 provided by NCEP and included in the WRF model. Step 3: Bind the multi-source heterogeneous data with the highest hourly prediction result accuracy of the historical upper-air wind field under different weather conditions to construct a sounding data assimilation scheme. Estimate the observation error covariance through the obsproc module and set the observation error covariance parameters. Estimate the background field error covariance through the background field error covariance estimation model and set the background field error covariance parameters. Determine the weather conditions at the initial time of the WRF model based on the measured ground observation data corresponding to the task area. Based on the sounding data assimilation scheme, input the multi-source heterogeneous data bound at the corresponding time of the weather conditions at the initial time into WRFDA, and use the three-dimensional variational assimilation method 3DVAR for assimilation to obtain the observation field required by the WRF model. Step 4: Use the three-dimensional variational assimilation method 3DVAR to assimilate the observation field and the background field to generate an analysis field, and use the analysis field to start the numerical prediction of the WRF model to obtain the hourly prediction result of the upper-air wind field. Step 5: Compare the hourly prediction result of the upper-air wind field with the observation data to obtain the accuracy rate of the hourly prediction result of the upper-air wind field. If the accuracy rate does not meet the preset standard of the aircraft route, use the two-factor rolling MOS correction method to correct and evaluate the hourly prediction result of the upper-air wind field, adjust the combination method of the multi-source heterogeneous data, update the sounding data assimilation scheme, and return to Step 3 for cyclic assimilation correction until the accuracy rate of the hourly prediction result of the upper-air wind field that meets the preset standard of the aircraft route is obtained. At this time, the corresponding sounding data assimilation scheme is the optimal sounding data assimilation scheme. If the accuracy rate of the hourly prediction result of the upper-air wind field that meets the preset standard of the aircraft route cannot be obtained, select the sounding data assimilation scheme corresponding to the highest accuracy rate of the hourly prediction result of the upper-air wind field as the optimal sounding data assimilation scheme.

2. The method according to claim 1, wherein In Step 2, the constructed background field error covariance estimation model is: B = (x background - x ture )(x background - x true ) T ; B 1 / 2 = U p U v U h ; where B is the background field error covariance, a symmetric and positive semi - definite square matrix, x background is the background field, x true is the true field, T is the transpose of the matrix; U p U v U h Adopt the background field error information pre - generated by default in the WRF model, and use the global background error covariance matrix CV3 provided by NCEP and included in the WRF model to achieve the optimal fusion of the background field and the observed data. Among them, U p is the physical transformation or balance operator, U v is the vertical transformation, U h is the horizontal transformation.

3. The method according to claim 1, wherein In Step 3, in the constructed sounding data assimilation scheme, The different weather conditions are: four types of weather conditions: few clouds on sunny days, cloudy, light rain on cloudy days, and thunderstorm gales. The multi-source heterogeneous data are: Beidou radiosonde data, wind profiler radar data, and microwave radiometer data. Among them, the combination methods of the multi-source heterogeneous data to be bound under different weather conditions include: no observation data, Beidou radiosonde data, Beidou radiosonde data combined with wind profiler radar data, Beidou radiosonde data combined with microwave radiometer data, wind profiler radar data combined with microwave radiometer data, and Beidou radiosonde data combined with wind profiler radar data and microwave radiometer data, a total of six combination methods.

4. The method according to claim 1 or 3, characterized in that, In Step 3, the constructed sounding data assimilation scheme is: When the weather condition is few clouds on sunny days, it is bound to the Beidou sounding data; When the weather condition is cloudy, it is bound to the Beidou sounding data combined with the wind profiler radar data; When the weather condition is light rain on overcast days, it is bound to the Beidou sounding data combined with the microwave radiometer data; When the weather condition is thunderstorm with strong wind, it is bound to the Beidou sounding data combined with the microwave radiometer data.

5. The method according to claim 1 or 3, characterized in that In step three, the method for judging the weather condition at the initial time of the WRF model by the measured surface observation data corresponding to the task area is as follows: If the cloud amount in the measured surface observation data corresponding to the task area ≤ 3 tenths, it is judged as few clouds on sunny days; 3 tenths < cloud amount < 8 tenths, it is judged as cloudy; Cloud amount ≥ 8 tenths, or 1-hour rainfall ≤ 2.5 mm, it is judged as light rain on overcast days; 1-hour rainfall > 2.5 mm, or average wind speed > 8 m / s, it is judged as thunderstorm with strong wind.

6. The method according to claim 1, characterized in that The objective function of the three-dimensional variational assimilation method 3DVAR is: Among them, J(x) is the solution of the cost function, x represents the analysis variable on the WRF model grid points, and x background is the background field, B represents the background field error covariance, y is the observed data, and y o represents the observation operator, R represents the observation error covariance, and T represents the transpose of the matrix.

7. The method according to claim 1, characterized in that In step five, the method for obtaining the hourly forecast accuracy rate of the upper-air wind field by comparing the hourly forecast results of the upper-air wind field with the observation data is as follows: Among them, In the formula, f w represents the hourly forecast result accuracy of the upper-air wind field obtained from the multi-source heterogeneous data bound by the weather condition w; where w is the weather condition; a is the east-west wind weight coefficient, and b is the north-south wind weight coefficient; f mu represents the accuracy of the east-west wind in the hourly forecast result of the upper-air wind field under m, where m represents the detection data assimilation scheme; f mv represents the accuracy of the north-south wind in the hourly forecast result of the upper-air wind field under m; N represents the judgment result of whether the forecast result conforms to the wind corresponding to the preset standard of the aircraft route; i represents the i-th layer height data; n represents the total number of layers of the forecast wind product; u mi represents the east-west wind at the i-th layer height in the hourly forecast result of the upper-air wind field under m, u oi represents the east-west wind at the i-th upper-air layer of the observed data; v mi represents the north-south wind at the i-th layer height in the hourly forecast result of the upper-air wind field under m; v oi represents the north-south wind at the i-th upper-air layer of the observed data.

8. The method according to claim 1, characterized in that, In step five, the method for correcting and evaluating the hourly forecast results of the upper-air wind field by using the two-factor rolling MOS correction method is as follows: y' t+h = b1y t+h + b2(O t - Y t ) + b0 where y' t+h is the correction value at time t + h, y t+h is the forecast value at time t + h, O t is the grid point actual situation at time t, Y t is the forecast value at time t; b0, b1, and b2 are the first model coefficient, the second model coefficient, and the third model coefficient automatically established by the least squares method, t is the observation data time, h is the difference between the correction time and the observation data time, and h = 1, 2,..., 24.

9. The method according to claim 7, wherein In step five, if the hourly forecast accuracy rate of the upper-air wind field that meets the preset standard of the flight path of the aircraft cannot be obtained, then the method for selecting the optimal sounding data assimilation scheme as the optimal sounding data assimilation scheme corresponding to the highest hourly forecast accuracy rate of the upper-air wind field is: If u mi -u oi ≤u s , then is 1; if u mi -u oi >u s , then is 0; If v mi -v oi ≤v s , then is 1; if v mi -v oi >v s , then is 0; where f' w represents the maximum accuracy rate of the hourly forecast results of the upper-air wind field under weather condition w; f' mu represents the average accuracy rate of the upper-air east-west winds under the sounding data assimilation scheme m; f' mv represents the average accuracy rate of the upper-air north-south winds under the sounding data assimilation scheme m; u s is the judgment criterion for east-west winds in the preset standard of the flight path of the aircraft; v s is the judgment criterion for north-south winds in the preset standard of the flight path of the aircraft; f s is the percentage standard in the preset standard of the flight path of the aircraft; K 总 represents the total number of samples under weather condition w; k represents the sample number under the sounding data assimilation scheme m.

10. The method according to claim 1 or 8, characterized in that, In step five, the preset standard of the flight path of the aircraft is: Under different weather conditions, the hourly forecast accuracy rate of the upper-air wind forecasted by the overall assimilation data in layers of 250 m each below 20 km is: for the east-west wind component and the north-south wind component, the wind speed ≤ the judgment standard, and the accuracy rate Accuracy ≥ the percentage standard.

Images