QAR data and airport self-observation meteorological data-based approach and departure program prediction method

Abstract

The invention discloses an arrival and departure program prediction method based on QAR data and airport self-observation meteorological data. Data matching is carried out based on a historical QAR actual flight track and an intelligence track; date attributes and time information corresponding to flights and runway weather information corresponding to the time are obtained from the matched QAR data to construct a terminal area program prediction model, data learning training is carried out, the trained model is used for predicting available terminal area programs, the available terminal area programs are recommended to a controller, and the controller can refer to a prediction result to carry out commanding rapidly. Compared with the prior art, the QAR data and airport self-observation meteorological data-based approach and departure program prediction method has the advantages that historical data matching and machine learning are performed according to a large amount of historical QAR data in combination with information such as meteorology and intelligence, available and optimal terminal area programs are found out, and the optimal terminal area programs are recommended to control personnel; therefore, the controller can quickly control and command.

Description

Technical Field

[0001] The present invention relates to the field of aviation, and in particular to an approach and departure procedure prediction method based on QAR data and airport self-observation meteorological data. Background Art

[0002] Terminal area air traffic control is carried out by terminal area air traffic controllers to direct incoming and outgoing aircraft based on the airport's incoming and outgoing routes, flight schedules, weather conditions, navigation facilities and other conditions. One of the important aspects of this control is to direct each aircraft on which incoming and outgoing procedures to use.

[0003] When directing aircraft to operate at the airport, the air traffic controller in the terminal area must give the arrival and departure procedures used in the terminal area. Usually, both ends of the runway at the airport can take off and land, and there are multiple available arrival and departure procedure paths at each end of the runway. When the air traffic controller is giving instructions, he needs to give instructions to the pilot and inform him of the arrival and departure procedure ID mark that he should use. The pilot must follow the controller's control.

[0004] There are many factors that controllers need to consider when selecting a terminal procedure path:

[0005] First: the effective arrival and departure procedure ID in the flight information data;

[0006] Second: operational standards affected by meteorological conditions. For example, when taking off, the tailwind must not be greater than x knots, the crosswind must not be greater than y knots, and the cloud base height must not be less than z feet when landing.

[0007] Third: the availability of the arrival and departure procedures can be judged by the availability of the navigation station equipment;

[0008] Fourth: other factors, etc.

[0009] Therefore, the controller has to consider multiple factors, and the command efficiency is low. If accurate terminal area procedure predictions can be made and the prediction results can be pushed to the controller, it will greatly facilitate the controller to conduct quick commands.

[0010] There are two major difficulties in predicting terminal area procedures: on the one hand, most prediction models use limited data sets, and a large amount of historical data is required to achieve accurate predictions; on the other hand, the predictions of terminal area trajectories in existing technologies have not been able to be combined with actual operating weather, intelligence, navigation equipment and restriction information, and the accuracy of predictions based solely on historical trajectories is limited. Summary of the invention

[0011] The purpose of the present invention is to provide a method for predicting the terminal area procedure of a certain flight in accordance with the actual operation requirements and control characteristics of civil aviation, combined with a large amount of historical QAR data and historical airport meteorological data, which solves the above problems.

[0012] In order to achieve the above object, the technical solution adopted by the present invention is: a method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data, the method flow is as follows:

[0013] Step 1: Match the actual flight trajectory of the historical QAR with the intelligence trajectory to find the best matching flight procedure;

[0014] Step 2: Obtain flight information from the QAR data after the terminal area program matches, and obtain the date attribute d corresponding to flight A from the flight information;

[0015] Step 3: Obtain the take-off or landing time t corresponding to flight A from the flight information, i.e., A(t), and extract the runway weather information q corresponding to time t from the historical airport self-observation weather data;

[0016] Step 4: Construct a terminal area procedure prediction model based on the flight procedures in the matched QAR data, the date attributes and time information in the flight information, and the meteorological information in the airport self-observation meteorological data;

[0017] Step 5: Perform supervised learning on the terminal area program prediction model to implement data training of the model;

[0018] Step 6: Use the trained model to calculate future weather forecasts and takeoff and landing dates and times to predict available terminal area procedures;

[0019] Step 7: The training model recommends the predicted terminal area procedure to the controller, who can then make quick commands based on the prediction results.

[0020] Preferably, in step one, a hidden Markov model is used to match the trajectory model, and then a Viterbi algorithm is used to find the best matching flight program.

[0021] As a preferred method, in the invisible Markov model, r1 is the flight segment defined in the intelligence data, and z is i is the GPS track point,

[0022] Defining the probability of observation

[0023] Among them, r i It is the flight segment defined in the intelligence data; i is the GPS track point; ||z t -x t，i|| great circle is the great circle distance of the GPS track point at time t from the defined flight segment; z For GPS accuracy, set this value to 4.0;

[0024] Define state transition probability

[0025] Among them, d t =||z t -z t+1 || greatcircle -||x t，i -x t+1，j || route || is the great circle distance between the GPS track point at time t+1 and the GPS track point at time t minus the defined segment distance; e is a constant 2.71828; β is an estimated median formula as follows:

[0026]

[0027] Define the initial state probability π i =p(z1|r i ).

[0028] As a preference, the optimal path trajectory vector is defined as: The closest trajectory vector is found through the Viterbi algorithm. The specific formula is as follows:

[0029]

[0030] Preferably, in step 2, the flight working day attribute is defined, and the date attribute is mainly divided into two categories, one is holidays and the other is working days, namely holidays A(h) and working days A(w).

[0031] Preferably, in step three, the runway meteorological information includes wind speed Ws(t), wind direction Wd(t), and cloud base height Cu(t).

[0032] Preferably, in step 5, the supervised learning method in pathon is used to train the terminal area program prediction model.

[0033] The input sample for its learning is,

[0034] d = [h, w], h is assigned to 0, w is assigned to 1; d is the date, h represents holidays, and w represents working days;

[0035] t=[t1,t2,t3,……t24], t is the time when flight A takes off or lands, that is, the vector is divided according to each hour of the day, and the corresponding value is a number 1-24;

[0036] q = [Ws, Wd, Cu], Ws is wind speed, Wd is wind direction, Cu is cloud base height;

[0037] p = [p1, p2, ... pn], replace the flight program name in the intelligence with the number pn. This vector is the flight program number used when the corresponding date attribute, corresponding time, and corresponding weather conditions are in the QAR data;

[0038] The output sample of its learning is,

[0039] ans=f(d, t, q, p).

[0040] Compared with the prior art, the advantages of the present invention are: the present invention is based on the arrival and departure procedure prediction method of QAR data and airport self-observed meteorological data. According to a large amount of historical QAR data, combined with meteorological, intelligence and other information, historical data matching and machine learning are carried out to find the available and optimal terminal area procedures, and the optimal terminal area procedures are recommended to the controller for rapid control and command. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the method of the present invention;

[0042] Figure 2 A comparison chart of the departure procedure of the aeronautical chart and the corresponding GPS track in the intelligence data of the present invention;

[0043] Figure 3 The invisible Markov model structure diagram of the present invention;

[0044] Figure 4 This is a schematic diagram of the path matching of the present invention. DETAILED DESCRIPTION

[0045] QAR (Quick access recorder)-specifically refers to an airborne flight data recorder with a protective device, which can record continuously for up to 600 hours and can collect hundreds of flight data at the same time, covering most parameters such as the aircraft's flight control, flight trajectory, equipment failure, etc.

[0046] Airport self-observed meteorological data is measured by ground-based meteorological wired telemetry instruments installed in the airport. Its detection part is installed near the runway and can automatically measure meteorological elements such as wind direction, wind speed, temperature, air pressure, humidity, cloud base height, runway visual range, etc. on different runways.

[0047] Terminal Area Procedures - Civil Aviation regulations specify the horizontal and profile trajectories that aircraft must follow when taking off and landing at an airport, and identify the corresponding speed and flight mode requirements.

[0048] Terminal area guidance equipment refers to a type of navigation equipment deployed at the airport, including very high frequency navigation stations (VHF), instrument landing equipment (ILS), etc. This equipment can transmit guidance signals to guide aircraft to take off or land along a safe trajectory.

[0049] The present invention will be further described below. A method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data is described in detail in detail below. Figure 1 , the method flow is as follows,

[0050] Step 1: Match the actual flight trajectory of the historical QAR with the intelligence trajectory to find the best matching flight procedure;

[0051] Since QAR data is stored at intervals of 1 second, there are many factors in the actual flight process that may cause the actual flight trajectory to not perfectly match the intelligence trajectory, mainly including the different turning radius caused by aircraft performance, point-to-point direct flight and detour commanded by control, etc. Figure 2 As shown, the left picture is intelligence data, in which the red line is the standard departure procedure, and the program name is LUVEN-09D. The green trajectory line in the right picture is the actual flight trajectory of QAR, and the program corresponding to the red line is LUVEN-09D.

[0052] Although the actual trajectory is different from the trajectory specified by the intelligence, the real QAR data needs to be matched to the corresponding program before the next analysis can be carried out. Therefore, the first step is data matching to exclude unmatched QAR data.

[0053] The present invention uses the invisible Markov model to match the trajectory model, with T i The flight segment defined in the intelligence data is represented by z i is a GPS track point, such as Figure 3 As shown,

[0054] Defining the probability of observation

[0055] Among them, r i It is the flight segment defined in the intelligence data; i is the GPS track point; ||z t -x t，i || great circle is the great circle distance of the GPS track point at time t from the defined flight segment; z For GPS accuracy, set this value to 4.0;

[0056] Define state transition probability

[0057] Among them, d t =||z t -zt+1 || greatcircle -||x t，i -x t+1，j || route || is the great circle distance between the GPS track point at time t+1 and the GPS track point at time t minus the defined segment distance; e is a constant 2.71828; β is an estimated median formula as follows:

[0058]

[0059] Define the initial state probability π i =p(z1|r i ),like Figure 4 shown.

[0060] Then use the Viterbi algorithm to find the best matching flight program, as follows:

[0061] The optimal path trajectory vector is defined as: The closest trajectory vector is found through the Viterbi algorithm. The specific formula is as follows:

[0062]

[0063] Due to the huge amount of industry QAR data, there are about 5 million flights every year, which is a necessary condition for the subsequent supervised learning to be better. Through this step, the industry QAR operation big data of China Civil Aviation will be matched with the intelligence data to find accurate samples from the massive data for subsequent supervised learning.

[0064] Step 2: Obtain flight information from the QAR data after the terminal area program matches, and obtain the date attribute d corresponding to flight A from the flight information;

[0065] Since holidays and working days are important parameters that affect flight scheduling, the present invention defines flight working day attributes, and mainly divides date attributes into two categories, one is holidays and the other is working days, namely holidays A(h) and working days A(w);

[0066] Step 3: Obtain the take-off or landing time t corresponding to flight A from the flight information, i.e., A(t), and extract the runway weather information q corresponding to time t from the historical airport self-observation weather data; the runway weather information includes wind speed Ws(t), wind direction Wd(t), and cloud base height Cu(t);

[0067] Step 4: Construct a terminal area procedure prediction model based on the flight procedures in the matched QAR data, the date attributes and time information in the flight information, and the meteorological information in the airport self-observation meteorological data;

[0068] Step 5: Use the supervised learning method in pathon to supervise the terminal area program prediction model to implement data training of the model.

[0069] The input sample for its learning is,

[0070] d = [h, w], h is assigned to 0, w is assigned to 1; d is the date, h represents holidays, and w represents working days;

[0071] t=[t1,t2,t3,……t24], t is the time when flight A takes off or lands, that is, the vector is divided according to each hour of the day, and the corresponding value is a number 1-24;

[0072] q = [Ws, Wd, Cu], Ws is wind speed, Wd is wind direction, Cu is cloud base height;

[0073] p = [p1, p2, ... pn], replace the flight program name in the intelligence with the number pn. This vector is the flight program number used in the corresponding date attribute, corresponding time, and corresponding weather conditions in the QAR data;

[0074] The output sample of its learning is,

[0075] ans = f(d, t, q, p);

[0076] Step 6: Use the trained model to calculate future weather forecasts and takeoff and landing dates and times to predict available terminal area procedures;

[0077] Steps 2 to 6 of the present invention conduct model training and prediction based on two major factors affecting control services, time and meteorological conditions, which are more in line with the requirements and characteristics of actual control.

[0078] Step 7: The training model recommends the predicted terminal area procedure to the controller, who can refer to the prediction results to quickly give instructions. The purpose of this step is to provide the controller with the prediction results in advance, which can avoid some risks in advance and improve the controller's work efficiency.

[0079] The present invention uses the above-mentioned arrival and departure procedure prediction method based on QAR data and airport self-observed meteorological data, and performs historical data matching and machine learning based on a large amount of historical QAR data in combination with meteorological, intelligence and other information, so as to find out the available and optimal terminal area procedures, and recommend the optimal terminal area procedures to the controller for rapid control and command.

[0080] The above is a detailed introduction to the arrival and departure procedure prediction method based on QAR data and airport self-observation meteorological data provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope, and changes and improvements to the present invention will be possible without exceeding the concept and scope specified in the attached claims. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data, characterized by: The method flow is as follows: Step 1: Match the actual flight trajectory of the historical QAR with the intelligence trajectory to find the best matching flight procedure; Step 2: Obtain flight information from the QAR data after the terminal area program matches, and obtain the date attribute d corresponding to flight A from the flight information; Step 3: Obtain the take-off or landing time t corresponding to flight A from the flight information, i.e., A(t), and extract the runway weather information q corresponding to time t from the historical airport self-observation weather data; Step 4: Construct a terminal area procedure prediction model based on the flight procedures in the matched QAR data, the date attributes and time information in the flight information, and the meteorological information in the airport's self-observed meteorological data; Step 5: Perform supervised learning on the terminal area program prediction model to implement data training of the model; Step 6: Use the trained model to calculate future weather forecasts and takeoff and landing dates and times to predict available terminal area procedures; Step 7: The training model recommends the predicted terminal area procedure to the controller, who can then make quick commands based on the prediction results.

2. The method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data according to claim 1, characterized in that: In step 1, a hidden Markov model is used to match the trajectory model, and then the Viterbi algorithm is used to find the best matching flight program.

3. The method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data according to claim 2, characterized in that: In the hidden Markov model, r i The flight segment defined in the intelligence data is represented by z i is the GPS track point, Defining the probability of observation Among them, r i It is the flight segment defined in the intelligence data; i is the GPS track point; ||z t -x t，i || great circle is the great circle distance of the GPS track point at time t from the defined flight segment; z For GPS accuracy, set this value to 4.0; Define state transition probability Among them, d t =||z t -z t+1 || greatcircle -||x t，i -x t+1，j || route || is the great circle distance between the GPS track point at time t+1 and the GPS track point at time t minus the defined segment distance; e is a constant 2.71828; β is an estimated median formula as follows: Define the initial state probability π i =p(z1|r i ).

4. The method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data according to claim 3, characterized in that: The optimal path trajectory vector is defined as: The closest trajectory vector is found through the Viterbi algorithm. The specific formula is as follows:

5. The method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data according to claim 1, characterized in that: In step 2, the flight working day attribute is defined, and the date attribute is mainly divided into two categories, one is holidays and the other is working days, namely holidays A(h) and working days A(w).

6. The method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data according to claim 1, characterized in that: In step 3, the runway meteorological information includes wind speed Ws(t), wind direction Wd(t), and cloud base height Cu(t).

7. The method for predicting arrival and departure procedures based on QAR data and airport self-observation meteorological data according to claim 1, characterized in that: In step 5, supervised learning method in pathon is used to train the terminal area program prediction model. The input sample for its learning is, d = [h, w], h is assigned to 0, w is assigned to 1; d is the date, h represents holidays, and w represents working days; t = [t1, t2, t3, ... t24], t is the time when flight A takes off or lands, that is, the vector is divided according to each hour of the day, and the corresponding values ​​are numbers 1-24; q = [Ws, Wd, Cu], Ws is wind speed, Wd is wind direction, Cu is cloud base height; p = [p1, p2, ... pn], replace the flight program name in the intelligence with the number pn. This vector is the flight program number used when the corresponding date attribute, corresponding time, and corresponding weather conditions are in the QAR data; The output sample of its learning is, ans=f(d, t, q, p).

Images