An Arrival Sorting Method Based on Machine Learning and Airport Capacity

Abstract

The present invention discloses an arrival sorting method based on machine learning and airport capacity, including: performing flight time machine learning and flight track machine learning based on historical flight data in advance to establish an approach airspace flight time resource pool and a track resource pool; collecting arrival flight data, predicting the estimated landing time of all arrival flights based on the arrival flight data and the approach airspace flight time resource pool; sorting the arrival flight capacity according to the estimated landing time; traversing all arrival flights, and determining the airspace where each arrival flight is located; if the arrival flight is in the zone adjustment airspace, performing zone adjustment sorting on the arrival flight based on the capacity sorting result; if the arrival flight is in the approach airspace, performing approach sorting on the arrival flight based on the capacity sorting result, and making route recommendations according to the track resource pool. The present invention realizes arrival sorting based on machine learning and airport capacity and gives optimal time or route recommendations.

Description

Technical Field

[0001] The present invention relates to the technical field of airport arrival sorting, and in particular to an arrival sorting method based on machine learning and airport capacity. Background Art

[0002] With the rapid development of the economy, the civil aviation industry has also achieved rapid and long-term development. In the traditional flight arrival control process, the command is more based on the controller's personal experience. With the increasing number of incoming flights, its arrival efficiency and safety are increasingly challenged. As an auxiliary control command for flight arrival and landing, the importance of the arrival flight sequencing management system is increasing. The existing arrival sequencing schemes focus more on optimizing arrival efficiency, reducing delays and increasing throughput through various sequencing schemes and strategies. Arrival sequencing generally involves two core aspects: four-dimensional trajectory prediction and sequencing, in which the four-dimensional trajectory involves three-dimensional trajectories of time and space; sequencing involves the landing order of incoming flights and resolving landing time slot conflicts.

[0003] A Chinese invention patent discloses a method and system for automatically adjusting flight approach procedures based on the AMAN system (application number: CN202010745169.5), which provides a method for resolving conflicts in landing flights in the terminal area. The core idea is to achieve the purpose of adjusting the landing time by truncating and combining the arrival and approach procedure route segments, and splicing multiple flight routes based on routes, thereby avoiding the problem of landing time slot conflicts. However, it has the following disadvantages: although it solves the time slot conflict problem of landing flights in the approach area, its route has limitations and is not enough to describe the actual flight trajectory of flights in the approach airspace; the time slot conflict solution is given based on flight time data with insufficient accuracy, and its sorting suggestions are not credible.

[0004] The reasons why the Chinese invention patent (application number: CN202010745169.5) has the above-mentioned shortcomings are as follows:

[0005] (1) A new route is generated by truncating and combining the arrival and approach procedure routes to achieve the purpose of adjusting the flight landing time. However, in actual control, the approach airspace is commanded by radar guidance and is not subject to the constraints of the route. Therefore, the new route formed by combining the arrival and approach procedure routes is not a real flight route and is not sufficient to describe the real flight trajectory of the approach airspace.

[0006] (2) Traditional solutions focus on the technology itself, while ignoring the collaborative process between various control departments, especially the two major control departments of approach control and regional control. The main beneficiary of arrival sequencing is approach control, which can provide approach control with sequencing suggestions for incoming flights, effectively reducing the control burden. Regional control needs to pay attention to incoming, outgoing and overflying flights at the same time. If the arrival sequencing solution brings more additional control burden to regional control, it will lead to resistance from other control departments including regional control. Therefore, the arrival sequencing solution should not only focus on the advancement of the technology itself, but also take into account the demands of various control departments at the same time to reduce the difficulty of promoting new technologies. Summary of the invention

[0007] In view of the defects in the prior art, the present invention provides an arrival sorting method based on machine learning and airport capacity, comprising:

[0008] Perform flight time machine learning and flight track machine learning based on historical flight data in advance to establish approach airspace flight time resource pool and track resource pool;

[0009] Collecting inbound flight data, and predicting the estimated landing time of all inbound flights based on the inbound flight data and the approach airspace flight time resource pool;

[0010] sorting the inbound flight capacity according to the estimated landing time;

[0011] Traverse all incoming flights and determine the airspace where each incoming flight is located;

[0012] If the inbound flight is in the zone-adjusted airspace, the inbound flight is sorted based on the capacity sorting result;

[0013] If the inbound flight is in the approach airspace, the inbound flight is sorted for approach based on the capacity sorting result, and a route suggestion is made according to the track resource pool.

[0014] Furthermore, the flight time machine learning is performed in advance based on historical flight data, specifically:

[0015] Acquire historical flight data, record the entry time of historical flights into the corridor entrance point and the flight landing time based on the historical flight data, and acquire the approach flight time and the flight flow of the current approach airspace according to the entry time and landing time;

[0016] Based on the historical flight data, the corresponding corridor entrance points, arrival procedures, landing runways and flight traffic are combined to form a time key value, and the approach flight time is grouped based on the time key value and saved in a data pool;

[0017] Taking out all the approach flight time data in the data pool, and performing data cleaning on the approach flight time data in the group;

[0018] An average value is calculated for the cleaned approach flight time data, and an approach airspace flight time resource pool is established based on the average value calculation result.

[0019] Furthermore, the flight track machine learning is performed in advance based on historical flight data, specifically:

[0020] Acquire historical flight data, and sample the flight tracks of historical flights in the approach airspace based on the historical flight data, wherein the sampling starts from when the historical flight enters the approach airspace and ends when the flight lands;

[0021] Setting attribute tags for the sampled tracks based on the historical flight data, and saving the tracks with the tags set to a pre-processed track pool, wherein the attribute tags include corridor entrance points, arrival procedures, landing runways, and aircraft models;

[0022] Taking out the track data from the pre-processed track pool, combining the corresponding corridor entrance point, the arrival procedure and the landing runway to form a track key value based on the track data, and grouping the track data according to the track key value;

[0023] The track data in the group is cleaned, and a track resource pool is established based on the cleaned track data.

[0024] Furthermore, the collecting of the inbound flight data and the prediction of the estimated landing time of all the inbound flights based on the inbound flight data and the approach airspace flight time resource pool are specifically as follows:

[0025] Collect inbound flight data;

[0026] Traversing all incoming flights based on the incoming flight data, and determining the airspace where each incoming flight is located;

[0027] If the inbound flight is located in the regional airspace, a regional CTA calculation is performed for the inbound flight in the regional airspace based on the approach airspace flight time resource pool;

[0028] If the inbound flight is located in the approach area, performing approach CTA calculation for the inbound flight in the approach area based on the approach airspace flight time resource pool;

[0029] Until all incoming flights complete the CTA calculation, the estimated landing time of all incoming flights can be obtained.

[0030] Furthermore, if the inbound flight is located in the zone-controlled airspace, a zone-controlled CTA calculation is performed on the inbound flight in the zone-controlled airspace based on the approach airspace flight time resource pool, specifically:

[0031] If the inbound flight is located in the zoned airspace, the time for the inbound flight to arrive at the corridor entrance point is obtained according to the inbound flight data;

[0032] According to the inbound flight data, based on corridor entrance points, inbound procedures, landing runways and flight traffic, the corresponding approach flight time is obtained from the approach airspace flight time resource pool;

[0033] The remaining flight time in the CTA phase is calculated according to the approach flight time and the time to fly to the corridor entrance point, and the estimated landing time of the incoming flight is obtained according to the remaining flight time in the CTA phase.

[0034] Furthermore, if the inbound flight is located in the approach area, the approach CTA calculation is performed on the inbound flight in the approach area based on the approach airspace flight time resource pool, specifically:

[0035] If the inbound flight is located in the approach area, obtaining the remaining flight time and the full flight time of the inbound flight according to the inbound flight data;

[0036] According to the inbound flight data, based on corridor entrance points, inbound procedures, landing runways and flight traffic, the corresponding approach flight time is obtained from the approach airspace flight time resource pool;

[0037] The remaining flight time in the CTA phase is calculated according to the approach flight time, the flight time along the remaining route, and the flight time of the entire route, and the estimated landing time of the incoming flight is obtained according to the remaining flight time in the CTA phase.

[0038] Furthermore, the inbound flight capacity is sorted according to the estimated landing time, specifically:

[0039] Obtaining capacity configuration, where the capacity configuration includes a unit time and a maximum number of inbound flights allowed;

[0040] Based on the estimated landing time, calculating whether the estimated number of inbound flights in any unit time exceeds the maximum allowed number of inbound flights;

[0041] If it does not exceed the limit, the inbound flight will arrive normally;

[0042] If it exceeds, the capacity sorting delay time of the inbound flight is calculated according to the unit time and the maximum allowed number of inbound flights, and capacity sorting is performed according to the estimated landing time and the capacity sorting delay time.

[0043] Furthermore, if the inbound flight is in the zone-adjusted airspace, the inbound flight is sorted based on the capacity sorting result, specifically:

[0044] If the inbound flights are in the zoned airspace, the inbound flights are grouped according to the corridor entrance points of the routes they pass through, and each group of inbound flights is pre-sorted according to the time sequence of arrival at the corridor entrance points;

[0045] Performing zone adjustment conflict judgment on the pre-sorted inbound flights in turn, if any zone adjustment conflict exists for the inbound flights, calculating the zone adjustment airspace delay time according to the capacity sorting delay time;

[0046] Sorting the inbound flights based on the delay time of the airspace to eliminate airspace conflicts;

[0047] The zone adjustment conflicts include wake turbulence interval conflicts, designated interval conflicts and waypoint interval conflicts.

[0048] Furthermore, if the inbound flight is in the approach airspace, the inbound flight is sorted for approach based on the capacity sorting result, specifically:

[0049] If the inbound flight is in the approach airspace, an approach conflict judgment is performed on the inbound flight; if the inbound flight has any approach conflict, the approach airspace delay time is calculated according to the capacity sorting delay time;

[0050] Generating an approach sorting landing time according to the approach airspace delay time, and performing approach sorting on the inbound flights based on the approach sorting landing time to eliminate approach conflicts;

[0051] The approach conflicts include runway closing time conflicts, runway spacing conflicts, slant distance spacing conflicts, wake turbulence spacing conflicts and designated spacing conflicts.

[0052] Furthermore, the route suggestion is performed according to the track resource pool, specifically:

[0053] Calculate the remaining approach flight time of the inbound flight after the approach sequence according to the approach sequence landing time;

[0054] Obtain the current position of the incoming flight, obtain the track passing through the current position from the track resource pool, and store it in a queue of tracks to be processed;

[0055] Based on the corridor entrance point, arrival procedure, landing runway and aircraft type corresponding to the incoming flight, matching is performed with the pending track queue, and the optimal track close to the remaining flight time of the approach is obtained according to the matching result;

[0056] The optimal flight path is output as a route suggestion to the flight control terminal.

[0057] The beneficial effects of the present invention are as follows: based on machine learning of historical flight time and track routes of flights, an approach airspace flight time resource pool and a track resource pool are established, thereby improving the prediction accuracy of subsequent four-dimensional trajectories of the approach airspace; based on the approach airspace flight time resource pool, an estimated landing time of incoming flights is predicted, and in combination with the characteristics of approach and zone adjustment control, an arrival sorting is performed with the airport's incoming flight capacity as the core, taking into account the control demands of approach and zone adjustment; when performing approach sorting, the optimal route suggestion can also be given based on the track resource pool, and the traditional time suggestion is replaced by the route suggestion, and the landing conflict problem is solved by route adjustment, thereby improving control efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0059] Figure 1 A flowchart of an arrival sorting method based on machine learning and airport capacity provided by an embodiment of the present invention;

[0060] Figure 2 A flight time machine learning flow chart of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0061] Figure 3 A flight track machine learning flow chart of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0062] Figure 4 A flowchart of the estimated landing time calculation of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0063] Figure 5 A capacity sorting diagram of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0064] Figure 6 A schematic diagram of a regional sorting process of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0065] Figure 7 A schematic diagram of an approach sorting process of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0066] Figure 8 A schematic diagram of a route suggestion process of an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention;

[0067] Fig. 9 A schematic diagram of flight current position track matching for an arrival sorting method based on machine learning and airport capacity provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0069] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.

[0070] like Figure 1 As shown, a method for sorting arrivals based on machine learning and airport capacity includes:

[0071] S1: Perform flight time machine learning and flight track machine learning based on historical flight data in advance to establish the approach airspace flight time resource pool and track resource pool;

[0072] Specifically, the airport air traffic control system will collect flight data in real time in the background and store the data accordingly. Before sorting the arrivals, it is necessary to obtain historical flight data from the real-time collected flight data in advance, and perform machine learning of the four-dimensional trajectory of the approach airspace based on the historical flight data, including machine learning from two aspects: flight time and three-dimensional spatial trajectory.

[0073] like Figure 2 As shown, flight time machine learning includes two stages: time collection and time processing. In the time collection stage, historical flight data is obtained, and the corridor entrance point in the route is used as the boundary. The corridor entrance point is within the approach airspace. Based on the historical flight data, the flight time of the historical flight in the approach airspace is recorded, that is, the entry time of the historical flight into the corridor entrance point and the flight landing time are recorded, and the approach flight time is calculated based on the entry time and landing time. At the same time, the flight flow in the current approach airspace is calculated with the entry time as the time node.

[0074] Based on the historical flight data, the corresponding corridor entrance points, arrival procedures, landing runways and flight traffic are combined to form a unique time key value, and the approach flight time is grouped based on the time key value and saved in the data pool.

[0075] In the time processing stage, all the approach flight time data in the data pool are taken out, and all the groups and the approach flight time in all the groups are traversed to calculate the median of each group of approach flight time, and the approach flight time with the largest deviation from the median is removed according to a certain ratio to clean the data of each group of approach flight time data. The average value of the remaining approach flight time data after cleaning is calculated, and the average value calculation result is used as the approach flight time under the group, and an approach airspace flight time resource pool is established according to each group of approach flight time until all groups are processed. Since each group of flight time data corresponds to a unique time key value, the approach airspace flight time resource pool contains the approach flight time corresponding to the corridor entrance point, arrival procedure, landing runway and flight flow.

[0076] like Figure 3 As shown in the figure, flight track machine learning includes two stages: track collection and track processing. In the track collection stage, historical flight data is obtained, and the corridor entrance point in the route is used as the boundary. The corridor entrance point is within the approach airspace. Based on the historical flight data, the track of the historical flight in the approach airspace is sampled, and the real track of the historical flight from entering the corridor entrance point to the landing of the flight is collected.

[0077] Attribute tags are set for the sampled tracks based on historical flight data. The attribute tags include corridor entrance points, arrival procedures, landing runways, and aircraft models. After the attribute tags are set, the tracks are stored in a pre-processed track pool.

[0078] Furthermore, in the track processing stage, the track data is taken from the pre-processed track pool, and the corresponding corridor entrance point, arrival procedure and landing runway are combined to form a unique track key value based on the track data, and the track data is grouped according to the track key value. All groups and track data within the groups are traversed, the median of each group of approach track data is calculated, and the track with the largest deviation from the median is removed according to a certain ratio to clean each group of approach track data, and a track resource pool is established based on the remaining track data after data cleaning for subsequent arrival sorting until all track processing is completed.

[0079] The flight time machine learning and flight track machine learning are performed in advance based on the historical flight data, providing data support for the subsequent arrival sorting of air traffic control personnel, and improving the prediction accuracy of the four-dimensional trajectory. It should be noted that the learning cycle and learning content of the machine learning described in this embodiment can be specifically set by the air traffic control personnel. For example, the air traffic control personnel can set the system to start machine learning at 4 o'clock every morning based on the historical flight data of the previous day.

[0080] S2: Collecting inbound flight data, and predicting the estimated landing time of all inbound flights based on the approaching flight data and the approaching airspace flight time resource pool;

[0081] Specifically, based on the machine learning conducted in advance, the estimated landing time of all inbound flights is predicted based on the machine learning results. Depending on the airspace the flight is located in, the estimated landing time calculation method used is also different, including zone CTA calculation and approach CTA calculation.

[0082] like Figure 4 As shown, the data of incoming flights is collected, and all incoming flights are traversed to determine the airspace where each incoming flight is located. If the incoming flight is currently located in the zone adjustment airspace, it will be divided into a zone adjustment flight, and the zone adjustment CTA calculation method will be used to calculate the estimated landing time of the zone adjustment flight; if the incoming flight is currently located in the approach airspace, it will be divided into an approach flight, and the approach CTA calculation method will be used to calculate the estimated landing time of the approach flight, until all incoming flights complete the CTA calculation, and the estimated landing time of all incoming flights is obtained. The process of zone adjustment CTA calculation and approach CTA calculation is as follows:

[0083] (1) If the incoming flight is currently in the zoned airspace, the flight time to the corridor entrance point T is calculated based on the incoming flight data. g , the calculation formula is:

[0084]

[0085] Where, t now is the current system time, d1 is the distance from the current position of the aircraft to the first unflown waypoint, v now is the current flight speed of the aircraft, d i-1 is the distance between the i-1th and i-th unflown waypoints, v i is the passing speed of the i-th point, and N is the serial number of the corridor entrance point.

[0086] According to the arrival flight data, the corridor entrance point, arrival procedure, landing runway and flight flow information are obtained, and the corridor entrance point, arrival procedure, landing runway and flight flow are used to form a unique time key value. According to the time key value, the corresponding approach flight time T is obtained from the approach airspace flight time resource pool established in step S1. m . According to the flight time to the corridor entrance point T g and approach flight time T m , calculate the remaining flight time T in the CTA phase, the calculation formula is:

[0087] T=T g +T m

[0088] The current system time is obtained based on the incoming flight data, and the estimated landing time CTA of all incoming flights in the zone control airspace can be calculated by adding the current system time to the remaining flight time T of the corresponding CTA stage.

[0089] (2) If the inbound flight is currently in the approach airspace, the remaining flight time T along the route of the inbound flight is calculated based on the inbound flight data. r and the total flight time T a The full route flight time is the full route flight time of the incoming flight from the corridor entrance point along the route to the runway touchdown point. According to the incoming flight data, the corridor entrance point, the arrival procedure, the landing runway and the flight flow information are obtained, and the corridor entrance point, the arrival procedure, the landing runway and the flight flow constitute a unique time key value. According to the time key value, the corresponding approach flight time T is obtained from the approach airspace flight time resource pool established in step S1. m According to the approach flight time T m , the remaining flight time T r And the total flight time T a , calculate the remaining flight time T in the CTA phase, the calculation formula is:

[0090]

[0091] The current system time is obtained based on the incoming flight data, and the estimated landing time CTA of all incoming flights in the approach airspace can be calculated by adding the current system time to the remaining flight time T of the corresponding CTA stage.

[0092] S3: sorting the inbound flight capacity according to the estimated landing time;

[0093] Specifically, the capacity configuration is obtained, and the capacity configuration includes the unit time P and the maximum number of inbound flights allowed Q. The capacity configuration described in this embodiment can be jointly negotiated and formulated by the airport approach control and regional control staff, balancing the control demands of all parties. The specific configuration can be adjusted in real time according to the control needs.

[0094] Based on the expected landing time, calculate whether the expected number of inbound flights exceeds the maximum allowed number of inbound flights Q within any unit time P minutes. If not, the inbound flights will arrive normally without flight delays. If it exceeds the maximum allowed number of inbound flights Q, calculate the capacity sorting delay time ΔT of the inbound flights based on the unit time P and the maximum allowed number of inbound flights Q. c , assuming that the mth flight exceeds the capacity, its capacity sorting delay time ΔT c The calculation formula is:

[0095]

[0096] Finally, the delay time ΔT is sorted according to the estimated landing time CTA and capacity cTo facilitate understanding, this embodiment takes P as 60 minutes and Q as 30 flights as an example to illustrate the capacity sorting. Figure 5 As shown in the figure, the first incoming flight is expected to land at T0. Within the time range of T0+60 minutes, 30 flights are expected to land, which does not exceed the capacity and does not require delay. The second incoming flight is expected to land at T1. Within the time range of T1+60 minutes, 32 flights are expected to land, which exceeds the capacity limit of 30. Therefore, the capacity sorting delay time of the 31st flight within T1+60 is: That is, the 31st flight is delayed by 2 minutes. Similarly, the 32nd flight is delayed by 4 minutes, and the third and subsequent flights are delayed in the same way to calculate the capacity sorting delay time.

[0097] By sorting the incoming flights by capacity, when the number of incoming flights does not exceed the capacity, there will not be too much interference with the regional control work, and no additional workload will be added due to the arrival sorting; when the capacity is exceeded, the regional control will appropriately control the number of flights handed over to the approach control, effectively reducing the burden of the approach control.

[0098] S4: Traverse all incoming flights and determine the airspace where each incoming flight is located;

[0099] Specifically, after all incoming flights are ranked in capacity, they are ranked in zone adjustment or approach order according to the airspace where they are located. All incoming flights are traversed to determine the airspace where each incoming flight is located, with the corridor entrance point in the route as the boundary. The airspace outside the corridor entrance point is the zone adjustment airspace, and the airspace inside the corridor entrance point is the approach airspace. The reference point for the sorting of the zone adjustment airspace is the corridor entrance point.

[0100] S41: If the inbound flight is in the zone adjustment airspace, perform zone adjustment sorting on the inbound flight based on the capacity sorting result;

[0101] Specifically, Figure 6 As shown in the figure, if the inbound flights are in the zoned airspace, they are grouped according to the corridor points of the routes they pass through, and the estimated arrival time t of the inbound flights at the corridor points is calculated. gate , the calculation formula is:

[0102]

[0103] Where, t now is the current system time, d1 is the distance from the current position of the aircraft to the first unflown waypoint, v now is the current flight speed of the aircraft, d i-1 is the distance between the i-1th and i-th unflown waypoints, v i is the passing speed of the i-th point.

[0104] After calculating the estimated arrival time of the inbound flights at the corridor entrance, pre-sort each group of inbound flights according to the time sequence of arrival at the corridor entrance. Perform zone adjustment sorting on the pre-sorted inbound flights in turn. If the inbound flight is the first landing flight, the zone adjustment sorting is completed directly. If the inbound flight is not the first landing flight, the zone adjustment conflict judgment is performed on the pre-sorted inbound flights in turn to determine whether there is a wake interval conflict, a designated interval conflict, and a waypoint interval conflict.

[0105] It should be noted that the order of judging wake turbulence interval conflict, designated interval conflict and waypoint interval can be set by the controller and is not limited here.

[0106] If there is any zone scheduling conflict for the incoming flights, the delay time ΔT will be sorted according to capacity c Calculate the airspace delay time ΔT1 of the area control, and the calculation formula is:

[0107] ΔT1=ΔT c +ΔT a1

[0108] In the formula, ΔT a1 It indicates the time difference between the arrival of similar inbound flights in the regional airspace at the corridor entrance point.

[0109] After obtaining the zone-adjusted airspace delay time ΔT1, the inbound flights are sorted based on the zone-adjusted airspace delay time ΔT1 to eliminate zone-adjusted conflicts until all inbound flights are sorted.

[0110] S42: If the inbound flight is in the approach airspace, approach sorting the inbound flight based on the capacity sorting result, and making route suggestions according to the track resource pool;

[0111] Specifically, Figure 7 As shown, if the incoming flight is within the approach airspace, all incoming flights in the approach area are traversed, and the approach sorting of the incoming flights is performed in sequence. If the incoming flight is the first landing flight, the approach sorting is completed directly. If the incoming flight is not the first landing flight, the approach conflict judgment is performed on the incoming flights in sequence to determine whether the flight has a runway closure time conflict, a runway interval conflict, a slant distance interval conflict, a wake turbulence interval conflict, and a designated interval conflict. The runway closure time conflict judgment refers to checking whether the estimated landing time CTA of the incoming flight is during the temporary closure of the runway.

[0112] It should be noted that the judgment order of runway closure time conflict, runway interval conflict, slant distance interval conflict, wake turbulence interval conflict and designated interval conflict can be set by the controller and is not limited here.

[0113] If there is any approach conflict for the incoming flight, the approach airspace delay time ΔT2 is calculated based on the capacity sorting delay time. The calculation formula is:

[0114] ΔT2=ΔT c +ΔT a2

[0115] In the formula, ΔT a2 Indicates the time difference between the arrival points of similar inbound flights in the approach area.

[0116] After the approach airspace delay time ΔT2 is obtained, the approach sorting landing time STA is generated according to the approach airspace delay time ΔT2 and the estimated landing time CTA. The calculation formula is:

[0117] STA=CTA+ΔT2

[0118] The inbound flights are sequenced based on the approach sequencing landing time STA to eliminate approach conflicts.

[0119] Furthermore, after resolving the problem of inbound flights exceeding capacity and flight landing conflicts and obtaining a new approach sequence landing time STA, it also includes route suggestions based on the track resource pool, such as Figure 8 shown.

[0120] Subtract the current system time from the approach sorting landing time STA to get the remaining approach flight time after the approach sorting of the inbound flight. The calculation formula is:

[0121] T ′ =STA-T now

[0122] Where, T now is the current system time. Get the current position of the incoming flight, obtain the track passing through the current position from the track resource pool established in step S1, and store the obtained track in the queue of pending tracks. For ease of understanding, Fig. 9 As shown in the figure, with the current position of the incoming flight as the center of the sphere and S kilometers as the radius (S is a configurable system parameter), a track matching area is formed. If a track in the track resource pool has a track point passing through this area, the track will be included in the queue of pending tracks, otherwise it will be discarded. That is, tracks L1, L2, and L3 pass through the matching area of ​​this flight and are put into the queue of pending tracks, while L4 and L5 do not pass through and are discarded.

[0123] Obtain the corridor entrance point, arrival procedure, landing runway and aircraft model corresponding to the incoming flight, and match them with the pending track queue. According to the matching results, obtain the optimal track close to the remaining flight time of the approach to make route suggestions. The specific matching process is as follows: Figure 8 As shown, including:

[0124] (a) Match the tracks with the same exit point, arrival procedure, landing runway and aircraft type from the queue of pending tracks, and determine whether there is a track with a similar remaining flight time to the current position. If so, adopt the track and exit track matching;

[0125] (b) If not, match the track with the same corridor entrance point, arrival procedure and landing runway from the queue of pending tracks, and determine whether there is a track with a similar remaining approach flight time to the current position. If so, adopt the track and exit track matching;

[0126] (c) If not, match the track with the same landing runway and aircraft type from the queue of pending tracks, and determine whether there is a track with a similar remaining approach time to the current position. If so, adopt the track and exit track matching;

[0127] (d) If not, match a track with the same landing runway from the queue of pending tracks, and determine whether there is a track with a similar remaining approach time to the current position. If so, adopt the track and exit track matching;

[0128] (e) If not, use the full en-route track.

[0129] After obtaining the corresponding matching track according to the matching results, the time of each passing point on the matching track is corrected according to the flight time, and the corrected matching track is output to the flight control terminal as the optimal track to provide reference suggestions for the control personnel.

[0130] In actual approach control, the flight route is adjusted more often than the speed adjustment to resolve landing flight conflicts (speed adjustment is reflected in time suggestions), so the traditional method of calculating time suggestions does not conform to the actual control scenario. The approach airspace route suggestions of this embodiment replace the delay time suggestions, which can better meet the actual needs. In addition, the route suggestions output by this embodiment are based on machine learning of historical real tracks, and their flight time is real and reliable. The more flight tracks, the more accurate it is, and at the same time, it avoids the complex modeling process of the flight track of the flight in the approach airspace.

[0131] The present invention conducts machine learning based on the historical flight time and track route of the flight, establishes an approach airspace flight time resource pool and a track resource pool, thereby improving the prediction accuracy of the subsequent approach airspace four-dimensional trajectory; predicts the estimated landing time of the incoming flight based on the approach airspace flight time resource pool, and combines the characteristics of approach and zone adjustment control to perform arrival sorting with the airport's incoming flight capacity as the core, taking into account the control demands of approach and zone adjustment; when performing approach sorting, it can also give the best route suggestion based on the track resource pool, replace the traditional time suggestion with the route suggestion, solve the landing conflict problem through route adjustment, and improve control efficiency.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for sorting arrivals based on machine learning and airport capacity, characterized in that: include: Perform flight time machine learning and flight track machine learning based on historical flight data in advance to establish approach airspace flight time resource pool and track resource pool; Collecting inbound flight data, and predicting the estimated landing time of all inbound flights based on the inbound flight data and the approach airspace flight time resource pool; sorting the inbound flight capacity according to the estimated landing time; Traverse all incoming flights and determine the airspace where each incoming flight is located; If the inbound flight is in the zone-adjusted airspace, the inbound flight is sorted based on the capacity sorting result; If the inbound flight is in the approach airspace, the inbound flight is sorted based on the capacity sorting result, and a route suggestion is made according to the track resource pool; The inbound flights are sorted based on the capacity sorting result, specifically: If the inbound flights are in the zoned airspace, the inbound flights are grouped according to the corridor entrance points of the routes they pass through, and each group of inbound flights is pre-sorted according to the time sequence of arrival at the corridor entrance points; The pre-sorted inbound flights are sequentially judged for zone adjustment conflicts. If any zone adjustment conflict exists in the inbound flights, the zone adjustment airspace delay time is calculated according to the capacity sorting delay time. Sorting the inbound flights based on the delay time of the airspace to eliminate airspace conflicts; the airspace conflicts include wake turbulence interval conflicts, designated interval conflicts, and waypoint interval conflicts; The approach sorting of the inbound flights is performed based on the capacity sorting result, specifically: If the inbound flight is in the approach airspace, an approach conflict judgment is performed on the inbound flight; if the inbound flight has any approach conflict, the approach airspace delay time is calculated according to the capacity sorting delay time; An approach sorting and landing time is generated according to the approach airspace delay time, and the inbound flights are sorted on the approach based on the approach sorting and landing time to eliminate approach conflicts; the approach conflicts include runway closure time conflicts, runway spacing conflicts, slant distance spacing conflicts, wake turbulence spacing conflicts and designated spacing conflicts.

2. The method for sorting arrivals based on machine learning and airport capacity according to claim 1, characterized in that: The flight time machine learning is performed in advance based on historical flight data, specifically: Acquire historical flight data, record the entry time of historical flights into the corridor entrance point and the flight landing time based on the historical flight data, and acquire the approach flight time and the flight flow of the current approach airspace according to the entry time and landing time; Based on the historical flight data, the corresponding corridor entrance points, arrival procedures, landing runways and flight traffic are combined to form a time key value, and the approach flight time is grouped based on the time key value and saved in a data pool; Taking out all the approach flight time data in the data pool, and performing data cleaning on the approach flight time data in the group; An average value is calculated for the cleaned approach flight time data, and an approach airspace flight time resource pool is established based on the average value calculation result.

3. The method for sorting arrivals based on machine learning and airport capacity according to claim 1, characterized in that: The flight track machine learning is performed in advance based on historical flight data, specifically: Acquire historical flight data, and sample the flight tracks of historical flights in the approach airspace based on the historical flight data, wherein the sampling starts from when the historical flight enters the approach airspace and ends when the flight lands; Setting attribute tags for the sampled tracks based on the historical flight data, and saving the tracks with the tags set to a pre-processed track pool, wherein the attribute tags include corridor entrance points, arrival procedures, landing runways, and aircraft models; Taking out the track data from the pre-processed track pool, combining the corresponding corridor entrance point, the arrival procedure and the landing runway to form a track key value based on the track data, and grouping the track data according to the track key value; The track data in the group is cleaned, and a track resource pool is established based on the cleaned track data.

4. The method for sorting arrivals based on machine learning and airport capacity according to claim 2, characterized in that: The collecting of the inbound flight data and the prediction of the estimated landing time of all the inbound flights based on the inbound flight data and the approach airspace flight time resource pool are specifically as follows: Collect inbound flight data; Traversing all incoming flights based on the incoming flight data, and determining the airspace where each incoming flight is located; If the inbound flight is located in the regional airspace, a regional CTA calculation is performed for the inbound flight in the regional airspace based on the approach airspace flight time resource pool; If the inbound flight is located in the approach area, performing approach CTA calculation for the inbound flight in the approach area based on the approach airspace flight time resource pool; Until all incoming flights complete the CTA calculation, the estimated landing time of all incoming flights can be obtained.

5. The method for sorting arrivals based on machine learning and airport capacity according to claim 4, characterized in that: If the inbound flight is located in the zone-controlled airspace, the zone-controlled CTA calculation is performed on the inbound flight in the zone-controlled airspace based on the approach airspace flight time resource pool, specifically: If the inbound flight is located in the zoned airspace, the time for the inbound flight to arrive at the corridor entrance point is obtained according to the inbound flight data; According to the inbound flight data, based on corridor entrance points, inbound procedures, landing runways and flight traffic, the corresponding approach flight time is obtained from the approach airspace flight time resource pool; The remaining flight time in the CTA phase is calculated according to the approach flight time and the time to fly to the corridor entrance point, and the estimated landing time of the incoming flight is obtained according to the remaining flight time in the CTA phase.

6. The method for sorting arrivals based on machine learning and airport capacity according to claim 4, characterized in that: If the inbound flight is located in the approach area, the approach CTA calculation is performed on the inbound flight in the approach area based on the approach airspace flight time resource pool, specifically: If the inbound flight is located in the approach area, obtaining the remaining flight time and the full flight time of the inbound flight according to the inbound flight data; According to the inbound flight data, based on corridor entrance points, inbound procedures, landing runways and flight traffic, the corresponding approach flight time is obtained from the approach airspace flight time resource pool; The remaining flight time in the CTA phase is calculated according to the approach flight time, the flight time along the remaining route, and the flight time of the entire route, and the estimated landing time of the incoming flight is obtained according to the remaining flight time in the CTA phase.

7. The method for sorting arrivals based on machine learning and airport capacity according to claim 5 or 6, characterized in that: The inbound flight capacity is sorted according to the estimated landing time, specifically: Obtaining capacity configuration, where the capacity configuration includes a unit time and a maximum number of inbound flights allowed; Based on the estimated landing time, calculating whether the estimated number of inbound flights in any unit time exceeds the maximum allowed number of inbound flights; If it does not exceed the limit, the inbound flight will arrive normally; If it exceeds, the capacity sorting delay time of the inbound flight is calculated according to the unit time and the maximum allowed number of inbound flights, and capacity sorting is performed according to the estimated landing time and the capacity sorting delay time.

8. The method for sorting arrivals based on machine learning and airport capacity according to claim 7, characterized in that: The route suggestion is made according to the track resource pool, specifically: Calculate the remaining approach flight time of the inbound flight after the approach sequence according to the approach sequence landing time; Obtain the current position of the incoming flight, obtain the track passing through the current position from the track resource pool, and store it in a queue of tracks to be processed; Based on the corridor entrance point, arrival procedure, landing runway and aircraft type corresponding to the incoming flight, matching is performed with the pending track queue, and the optimal track close to the remaining flight time of the approach is obtained according to the matching result; The optimal flight path is output as a route suggestion to the flight control terminal.

Images