Decision support system, method and storage medium for abnormal flight recovery

Abstract

Disclosed are a decision support system, method, and storage medium for abnormal flight recovery. The system includes a storage unit storing a dispatcher's decision information on abnormal flight recovery and historical abnormal flight recovery data; a scoring unit constructing a scoring standard based on the decision information and the historical abnormal flight recovery data, and using the scoring standard to score the abnormal flight based on the abnormal flight information; and an adjustment plan generating unit determining an abnormal flight with a score greater than a first predetermined threshold as a flight requiring recovery, determining multiple candidate adjustment flights for the flight requiring recovery, generating multiple adjustment plans based on the multiple candidate adjustment flights, and selecting the adjustment plan with the lowest score from the generated multiple adjustment plans for use in adjusting the flight requiring recovery, wherein the score of the adjustment plan is the sum of the scores of subsequent flights of the candidate adjustment flight.

Description

Technical Field

[0001] The present disclosure relates generally to the field of aviation, and more particularly to a decision support system, method, and storage medium for irregular flight recovery, and more particularly to a method, system, and computer storage medium for evaluating the importance of irregular flights and generating an adjustment plan by simulating a dispatcher's decision-making process. Background Art

[0002] With the rapid development of civil aviation in recent years, factors such as severe weather, aircraft failures, and air traffic control have sometimes disrupted normal flight schedules. When these disrupted flights occur, they not only reduce service quality and cause inconvenience to passengers, but also increase airline operating costs and affect subsequent flights.

[0003] In recent years, a great deal of research has been conducted to develop optimal flight scheduling, however only a small amount of research has been devoted to addressing the impact of irregular flights and developing feasible decision support systems to assist in short-term aircraft rescheduling.

[0004] The problem of recovering irregular flights is a real-time large-scale integer programming (Barnhart, C., Boland, N. L., Clarke, L. W., Johnson, E. L., Nemhauser, G. L., & Shenoi, R. G. (1998). Flight string models for aircraft fleeting and routing. Transportation Sci. 32(3), 208 - 220; Abdelghany, K. F., Abdelghany, A. F., & Ekollu, G. (2008). An integrated decision support tool for airlines schedule recovery during irregular operations. European Journal of Operational Research, 185(2), 825 - 848; Clausen, J., Larsen, A., Larsen, J., & Rezanova, N. J. (2010). Disruption management in the airline industry—Concepts, models and methods. Computers & Operations Research, 37(5), 809 - 821; Sinclair, K., Cordeau, J. F., & Laporte, G. (2014). Improvements to a large neighborhood search heuristic for an integrated aircraft and passenger recovery problem. European Journal of Operational Research, 233(1), 234 - 245; Maher, S. J. (2015). Solving the integrated airline recovery problem using column-and-row generation. Transportation Science, 50(1), 216 - 239; Ar1kan, U., Gürel, S., & Aktürk, M. S. (2017).Flight network-based approach for integrated airline recovery with cruise speed control. Transportation Sci. 51 (4), 1259-1287), which has complex variables and constraints, and the operating mechanisms of different airlines are different. Known are mixed integer programming (MIP) and its variations, such as mixed integer nonlinear programming (MINP), cone mixed integer programming (CMIP), and second-order cone mixed integer programming (CQMIP). In addition, decomposition algorithms such as Benders decomposition (BD), Dantzig-Wolfe decomposition (DWD), and Lagrangian decomposition (LD) are often used to simplify mixed integer programming problems. Another approach is to formulate the recovery problem as a set covering model, which can be solved by using a column generation (CG) algorithm, a branch and bound method (BB), or a branch and cut method (BC). Finally, instead of solving the optimal solution, many studies have attempted to find approximate optimal solutions by applying methods such as integer programming, nonlinear programming, heuristic learning, and network flow (Clausen, J., Larsen, A., Larsen, J., & Rezanova, NJ (2010). Disruption management in the airline industry—Concepts, models and methods. Computers & Operations Research, 37(5), 809-821).

[0005] However, existing research has the following defects: First, most studies are used to solve the global (approximate) optimal solution. Therefore, the running time of the algorithm is very sensitive to the scale of the aviation network, and it is difficult to obtain a solution quickly for large aviation networks; second, even if the (approximate) optimal solution is obtained, it will be difficult to implement when many flights need to be adjusted. For example, sometimes the current schedule is not optimal and needs further adjustment; third, the efficiency of the algorithm is seriously dependent on the specific form of the objective function. Most studies consider one or two of the four optimization objectives (i.e., routes, crew, passengers, and rescheduling). Therefore, a slight change in the objective function (e.g., adding a factor) will cause the algorithm to fail to work. Summary of the Invention

[0006] Technical problems to be solved

[0007] The embodiments of the present disclosure provide a decision support system, method, and storage medium for abnormal flight recovery, which at least solve the problems of low recovery efficiency and large computational complexity in abnormal flight recovery in related technologies.

[0008] Means of solving technical problems

[0009] One aspect of the present disclosure provides a decision support system for abnormal flight recovery, comprising: a storage unit storing a dispatcher's decision information on the recovery of an abnormal flight and historical recovery data of the abnormal flight; a scoring unit constructing a scoring standard based on the decision information and the historical recovery data of the abnormal flight, and using the scoring standard to score the abnormal flight according to the information of the abnormal flight; and an adjustment plan generating unit determining an abnormal flight having a score greater than a first predetermined threshold as a flight that needs to be recovered, determining multiple candidate adjustment flights for the flight that needs to be recovered, generating multiple adjustment plans based on the multiple candidate adjustment flights, and selecting an adjustment plan with the lowest score from the generated multiple adjustment plans for adjusting the flight that needs to be recovered, wherein the score of the adjustment plan refers to the sum of the scores of subsequent flights of the candidate adjustment flight, and the time interval between the arrival time of the previous flight of each candidate adjustment flight and the departure time of the flight that needs to be recovered is within a predetermined time.

[0010] The concept of the present disclosure is to simulate the decision-making process of the dispatcher based on interviews with the dispatcher, abnormal flight scheduling information obtained by questionnaire surveys of the dispatcher, and abnormal flight historical recovery data obtained from historical system records.

[0011] The concept of the present disclosure is not to try to find an optimal solution, but to provide feasible suggestions for restoring irregular flights. There are two reasons for not trying to find an optimal solution: first, the optimal solution needs to be determined through an objective function. However, the objectives of different airlines vary greatly, so it is impossible to provide a universal optimal solution; second, the current flight schedule is not optimal, but in the absence of irregular flights, the system does not need to make changes to the current flight schedule, and changes to the flight schedule will lead to unnecessary increases in operating costs. Moreover, even if some flights are unprofitable due to many reasons, these flights cannot be canceled.

[0012] The decision support system of this disclosure is built on two initial assumptions: first, that the current flight schedule is feasible and does not require change; second, if an irregular flight needs to be restored, the change is preferably made to a flight with the same origin, destination, arrival time, and departure time. The first assumption means that the decision support algorithm is not activated when there are no irregular flights. The second assumption means that the algorithm begins with a local search for an adjustment plan.

[0013] Preferably, the storage unit stores a database structure constructed based on the decision information and abnormal flight historical recovery data, the database structure including flight data, aircraft data, passenger data, and delay time. The scoring unit includes: a model construction unit, which, based on the analysis of the decision information and the abnormal flight historical recovery data, uses the hierarchical analysis method to construct a hierarchical model for the flight data, aircraft data, passenger data, and delay time. The hierarchical model includes a target layer, a criterion layer, and an alternative solution layer. The target layer is for determining flights that need to be restored. The criterion layer includes at least the following parent layers: flight data, aircraft data, passenger data, and delay time. Each parent layer includes child elements. The alternative solution layer is all abnormal flights. The child elements of the flight data include at least flight type and flight density. The child elements of the aircraft data include aircraft type. The child elements of the passenger data include VIP information. The child elements of the delay time include a first delay time greater than a first predetermined duration, a second delay time less than a second predetermined duration, and a third delay time greater than the second predetermined duration and less than the first predetermined duration.

[0014] Preferably, the scoring unit also includes: a first scoring unit, which uses a hierarchical model to calculate the relative weight of each child element of each parent level in the corresponding parent level based on the priorities of pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of pairs of different abnormal flights in the abnormal flight history recovery data.

[0015] Preferably, the scoring unit also includes: a second scoring unit, which uses a machine learning algorithm to calculate the absolute weight of each parent level as a scoring criterion for the relative weight of each child element in the corresponding parent level based on the priorities of the pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of the pairs of different abnormal flights in the abnormal flight history recovery data.

[0016] Preferably, the second scoring unit is also used to calculate the absolute weight of each child element using the relative weight of each child element and the absolute weight of each parent level according to the hierarchical model, and the second scoring unit calculates the score of the abnormal flight using the absolute weight of each parent level according to the information of the abnormal flight or calculates the score of the abnormal flight using the absolute weight of each child element according to the information of the abnormal flight.

[0017] Preferably, assuming that the number of parent levels is m, and parent level i includes ni child elements, the relative weight y of child element j in parent level i is i (j) Satisfaction And assuming that the priority of K for abnormal flights is extracted from the decision information and the abnormal flight history recovery data, the weight vector of abnormal flight k is z k =(y1 k , y2 k ,...,ym k ),in, is the weight vector of parent level i, so the score of abnormal flight k is Among them, w i is the absolute weight of the parent level, assuming z k1 than z k2 has a higher priority, then z k1 >z k2 , and set Δy i k =y i k1 -y i k2 , but Take λ k ≤0 so that Δz k ≥λ k , and λ k The weight of In satisfaction and Δz k ≥λ k Under the conditions of Maximize to obtain the absolute weight w of the parent level i .

[0018] Preferably, the adjustment plan generating unit includes: a restoration flight determining unit, which receives the score of each abnormal flight from the scoring unit, determines whether the score of each abnormal flight is greater than a first predetermined threshold, and determines the abnormal flight with a score greater than the first predetermined threshold as a flight that needs to be restored; a candidate flight generating unit, which generates a plurality of candidate adjustment flights for the flight that needs to be restored based on the scores of the flight that needs to be restored; and an adjustment plan selecting unit, which generates a plurality of adjustment plans based on the plurality of candidate adjustment flights and selects an adjustment plan with the lowest score from the generated plurality of adjustment plans for adjusting the flight that needs to be restored.

[0019] Preferably, the candidate flight generating unit generates respective candidate adjustment flights for the flights that need to be restored in order of priority from high to low according to the scores of the abnormal flights, and the adjustment plan selecting unit generates multiple adjustment plans for the flights that need to be restored in order of priority from high to low according to the scores of the flights that need to be restored and selects the adjustment plan with the lowest score from the generated multiple adjustment plans for adjusting the flights that need to be restored.

[0020] Preferably, the adjustment plan selection unit is further used to determine whether the delay time of the flight to be restored is greater than a second predetermined threshold, and when it is determined that the delay time of the flight to be restored is greater than or equal to the second predetermined threshold, use the candidate adjustment flight to replace the flight to be restored, and when it is determined that the delay time of the flight to be restored is less than the second predetermined threshold, exchange the flight to be restored with the candidate adjustment flight.

[0021] Preferably, the candidate adjusted flight has the same destination as the flight that needs to be restored.

[0022] Another aspect of the present disclosure provides a decision-making method for abnormal flight recovery, comprising: storing a dispatcher's decision information on the recovery of an abnormal flight and historical recovery data of the abnormal flight; constructing a scoring standard based on the decision information and the historical recovery data of the abnormal flight, and using the scoring standard to score the abnormal flight according to the information of the abnormal flight, determining an abnormal flight with a score greater than a first predetermined threshold as a flight that needs to be recovered, and determining a plurality of candidate adjustment flights for the flight that needs to be recovered; and generating a plurality of adjustment plans based on the plurality of candidate adjustment flights, and selecting an adjustment plan with the lowest score from the generated plurality of adjustment plans for adjusting the flight that needs to be recovered, the score of the adjustment plan being the sum of the scores of subsequent flights of the candidate adjustment flight, wherein the time interval between the arrival time of the preceding flight of each candidate adjustment flight at the airport and the departure time of the flight that needs to be recovered is within a predetermined time.

[0023] Preferably, the method further includes: constructing a database structure based on the decision information and the abnormal flight history recovery data, the database structure including flight data, aircraft data, passenger data and delay time; based on the analysis of the decision information and the abnormal flight history recovery data, constructing a hierarchical model for the flight data, aircraft data, passenger data and delay time using the hierarchical analysis method, the hierarchical model including a target layer, a criterion layer and an alternative solution layer, the target layer is for determining the flights that need to be restored, the criterion layer includes at least the following parent layers: flight data, aircraft data, passenger data and delay time, each parent layer includes child elements, the alternative solution layer is for all abnormal flights, the child elements of the flight data include at least flight type and flight density, the child elements of the aircraft data include aircraft type, the child elements of the passenger data include VIP information, and the child elements of the delay time include a first delay time greater than a first predetermined duration, a second delay time less than a second predetermined duration, and a third delay time greater than the second predetermined duration and less than the first predetermined duration.

[0024] Preferably, constructing a scoring standard based on the decision information and the abnormal flight history recovery data includes: calculating the relative weight of each child element of each parent level in the corresponding parent level using a hierarchical model based on the analysis of the decision information.

[0025] Preferably, constructing a scoring criterion based on the decision information and the abnormal flight history recovery data further includes: based on the priorities of pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of pairs of different abnormal flights in the abnormal flight history recovery data, using a machine learning algorithm to calculate the absolute weight of each parent level as a scoring criterion for the relative weight of each child element in the corresponding parent level.

[0026] Preferably, constructing a scoring standard based on decision information and abnormal flight historical recovery data also includes: calculating the absolute weight of each child element using the relative weight of each child element and the absolute weight of each parent level according to a hierarchical model, and calculating the score of the abnormal flight using the absolute weight of each parent level according to the information of the abnormal flight, or calculating the score of the abnormal flight using the absolute weight of each child element according to the information of the abnormal flight.

[0027] Preferably, calculating the absolute weight of each parent level includes: assuming that the number of parent levels is m, and parent level i includes ni child elements, then the relative weight y of child element j in parent level i is i (j) Satisfaction And assuming that the priority of K for abnormal flights is extracted from the decision information and the abnormal flight history recovery data, the weight vector of abnormal flight k is z k =(y1 k , y2 k ,...,y m k ),in, is the weight vector of parent level i, so the score of abnormal flight k is Among them, w i is the absolute weight of the parent level, assuming z k1 than z k2 has a higher priority, then z k1 >z k2 , and set Δy i k =y i k1 -y i k2 , but Take λ k ≤0 so that Δz k ≥λ k , and λ k The weight of In satisfaction and Δz k ≥λ k Under the conditions of Maximize to obtain the absolute weight w of the parent leveli .

[0028] Preferably, the decision-making method further includes: determining the number of flights that need to be restored, and when the number of flights that need to be restored is greater than 1, generating respective candidate adjustment flights for the flights that need to be restored in descending order of priority according to the scores of the flights that need to be restored, and determining adjustment plans for the flights that need to be restored in descending order of priority according to the scores of the flights that need to be restored.

[0029] Preferably, the decision-making method also includes: before generating multiple adjustment plans, determining whether the delay time that needs to be restored is greater than a second predetermined threshold; if so, replacing the flight that needs to be restored with a candidate adjustment flight; if not, exchanging the candidate adjustment flight with the flight that needs to be restored.

[0030] Preferably, the first predetermined threshold is set to 0.15, the predetermined time is 3 hours, and the second predetermined threshold is 4 hours.

[0031] Another aspect of the present disclosure provides a computer storage medium having a computer program stored thereon, wherein the computer program implements the above method when executed by a processor.

[0032] Technical effects

[0033] Compared to previous studies, the method disclosed in this paper has the following three advantages: First, it does not rely on a specific form of the objective function. In fact, this paper establishes a flight scoring system based on existing information to evaluate the impact of irregular flights and measures the bias of the adjustment plan. Second, the method disclosed in this paper can generate several feasible adjustments for the dispatcher to choose from. Moreover, if a feasible adjustment plan does not exist, the algorithm can generate a plan with lower compensation. Third, the algorithm's running time depends on the network size because the system focuses on local solutions rather than finding the global optimum. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0035] Figure 1 is a schematic diagram of a decision support system for abnormal flight recovery according to an embodiment of the present disclosure;

[0036] Figure 2 is a schematic diagram of a database structure according to an embodiment of the present disclosure;

[0037] Figure 3 is a schematic diagram of the structure of a hierarchical model according to an embodiment of the present disclosure;

[0038] Figure 4 is a flow chart of a decision-making method for abnormal flight recovery according to an embodiment of the present disclosure;

[0039] Figure 5 is a flow chart of another decision method for abnormal flight recovery according to an embodiment of the present disclosure;

[0040] Figure 6A and Figure 6B Schematic diagram of an example of generating candidate adjusted flights according to an embodiment of the present disclosure DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the present disclosure, the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present disclosure.

[0042] Figure 1 is a schematic diagram showing a decision support system for abnormal flight recovery according to an embodiment of the present disclosure, such as Figure 1 As shown, the decision support system 100 includes a storage unit 10, which stores the dispatcher's decision information on the recovery of abnormal flights and historical recovery data of abnormal flights; a scoring unit 20, which constructs a scoring standard based on the decision information and the historical recovery data of abnormal flights, and uses the scoring standard to score the abnormal flights according to the information of the abnormal flights; and an adjustment plan generating unit 30, which determines the abnormal flights with scores greater than a first predetermined threshold as flights that need to be recovered, and determines multiple candidate adjustment flights for the flights that need to be recovered and generates multiple adjustment plans based on the multiple candidate adjustment flights, and selects the adjustment plan with the lowest score from the generated multiple adjustment plans for adjusting the flight that needs to be recovered, the score of the adjustment plan refers to the sum of the scores of subsequent flights of the candidate adjustment flight, and the time interval between the arrival time of the previous flight of each candidate adjustment flight and the departure time of the flight that needs to be recovered is within a predetermined time.

[0043] Specifically, multiple candidate adjustment flights are located at the same airport as the flight to be resumed, and the subsequent flights of each candidate adjustment flight are flights with the same aircraft tail number as the candidate adjustment flight, that is, flights using the same aircraft. Similarly, the predecessor flight of each candidate adjustment flight is the predecessor flight with the same aircraft tail number as the candidate adjustment flight, that is, each candidate adjustment flight and the predecessor flight use the same aircraft. Furthermore, the arrival time of the predecessor flight of each candidate adjustment flight at the airport can be within, for example, three hours of the departure time of the flight to be resumed.

[0044] Here, the dispatcher's decision information on abnormal flight recovery is obtained based on interviews with the dispatcher, questionnaires, etc., and abnormal flight recovery history data is obtained from historical system records. In addition, through interviews with the dispatcher, the first predetermined threshold is set to 0.15 based on the dispatcher's inspection and strategy.

[0045] In addition, the decision information and abnormal flight history recovery data stored in the storage unit 10 can be constructed into a database structure, such as Figure 2 As shown, the database structure can include flight data, aircraft data, passenger data, and event data such as delay time. Flight data includes the flight number, route, flight schedule before and after adjustment, and whether the flight has any subsequent missions (if so, the subsequent missions need to be adjusted accordingly). Aircraft data can include the aircraft's tail number, aircraft type, and any special attributes that may affect flight recovery (wide-body or narrow-body). Passenger data can include VIP information and the number of passengers in different cabin classes (this data is not used in this article due to data collection limitations). Event data is the most important data and can be further divided into three dimensions: aircraft events, route events, and delay time. Aircraft events can include aircraft malfunctions, previous aircraft delays, and aircraft maintenance. In these cases, the aircraft cannot fly, and the dispatcher may need to replace it with another aircraft. Route events refer to situations where the route is unavailable, but the aircraft is available. These situations may include severe weather, traffic control, military control, etc. In these cases, the aircraft on the route can be flown on other routes. Note that delay time is used to measure the severity of the event.

[0046] like Figure 1As further shown, the scoring unit 20 includes: a model building unit 201, which uses a hierarchical analysis method to build a hierarchical model for flight data, aircraft data, passenger data and delay time based on the analysis of decision information and historical recovery data of abnormal flights. The hierarchical model includes a target layer, a criterion layer and an alternative solution layer. The target layer is to determine the flights that need to be restored. The criterion layer includes at least the following parent layers: flight data, aircraft data, passenger data and delay time. Each parent layer includes child elements. The alternative solution layer is all abnormal flights. The child elements of flight data include at least flight type and flight density. The child elements of aircraft data include at least aircraft type. The child elements of passenger data include at least VIP information. The child elements of delay time include at least a first delay time greater than a first predetermined time, a second delay time less than a second predetermined time, and a third delay time greater than the second predetermined time and less than the first predetermined time.

[0047] Note that aircraft and route reasons in the event data are not included in the analysis. This is because for route reasons, dispatchers cannot recover abnormal flights. For aircraft reasons, although abnormal operations due to specific reasons play an important role in flight recovery, dispatchers are not concerned about them. Therefore, the analysis in this article does not include aircraft and route reasons in the event data. The following specific examples will further illustrate and explain them.

[0048] Specifically, Figure 3 A specific structural diagram of the hierarchical model according to an embodiment of the present disclosure is shown in FIG. Figure 3 As shown, the target layer is to identify irregular flights, and the criterion layer includes the following parent layers: flights (B1), aircraft (B2), aircraft events (B3), route events (B4), passengers (B5), and delay times (B6). In the parent layer flight B1, there are four sub-elements used to measure the attributes and density of flights. Among them, sub-element C1 refers to international flights, sub-element C2 refers to the average number of flights on this route per day being less than 1, sub-element C3 refers to the route having one or two flights per day, and sub-element C4 refers to the route having more than two flights per day. The sub-elements in parent layers B2-B5 are derived from data stored in storage unit 10. In parent layer B6, delay times can be divided into three levels: short delay times of less than one hour, long delay times between one and four hours, and very long delay times of more than four hours.

[0049] In each parent level, upper elements have higher priority than lower elements, such as Figure 3 The priorities indicated by arrows are from low to high.

[0050] Back to Figure 1The scoring unit 20 further includes: a first scoring unit 202, based on the priorities of the pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of the pairs of different abnormal flights in the abnormal flight history recovery data, using Figure 3 The hierarchical model shown calculates the relative weight of each child element of each parent level in the corresponding parent level; and the second scoring unit 203, based on the priorities of pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of pairs of different abnormal flights in the abnormal flight history recovery data, uses a machine learning algorithm to calculate the absolute weight of each parent level for the relative weight of each child element in the corresponding parent level, and the absolute weight can be used as a scoring criterion.

[0051] In addition, the second scoring unit 203 can also be used to calculate the absolute weight of each child element using the relative weight of each child element and the absolute weight of each parent level according to the hierarchical model, and calculate the score of the abnormal flight using the absolute weight of each parent level according to the information of the abnormal flight, or calculate the score of the abnormal flight using the absolute weight of each child element according to the information of the abnormal flight.

[0052] The working principles of the first scoring unit 202 and the second scoring unit 203 will be described in detail below. Figure 3 As shown, a hierarchical model can be established by the analytic hierarchy process (AHP). It is known that AHP is a very flexible and powerful tool because it obtains scores based on the pairwise relative evaluation of the criteria and solutions provided by the decision maker (for example, the dispatcher in this article) (Saaty, TL, (1986). Axiomatic foundation of the analytic hierarchy process. 32(7), 841-855; Saaty, TL, & Vargas, LG (2012). Models, methods, concepts & applications of the analytic hierarchy process (Vol. 175): Springer Science & Business Media.). The calculations made by AHP are always guided by the experience of the decision maker. Therefore, AHP can be considered as being able to convert the evaluations (quantitative and qualitative) made by the decision maker into multiple standard hierarchies. In addition, AHP is simple because it does not require the construction of a complex expert system that needs to embed the experience of the decision maker.

[0053] AHP is based on the following three basic principles: decomposition, comparative judgment, and hierarchical synthesis (Saaty, TL, (1986). Axiomatic foundation of the analytic hierarchy process. 32(7), 841-855; Saaty, TL, & Vargas, LG (2012). Models, methods, concepts & applications of the analytic hierarchy process (Vol. 175): Springer Science & Business Media). The decomposition principle is used to structure complex problems into a hierarchical structure of parent levels, child elements, sub-child elements, etc. The comparative judgment principle is used to construct pairwise comparisons of all combinations of child elements in the parent level relative to the parent level. These pairwise comparisons are used to obtain the local priorities of child elements in the parent level relative to the parent level. The principle of hierarchical synthesis is used to multiply the local priorities of child elements in the parent level by the global priority of the parent element to generate the global priority of all levels, and then add up the global priorities of each child element.

[0054] Based on interviews, questionnaires and decision information obtained from historical system records, the following Figure 3 The AHP hierarchical model shown. The first level has six parent levels, namely flights (B1), aircraft (B2), aircraft events (B3), route events (B4), passengers (B5) and delay time (B6). In the parent level flight B1, there are 4 child elements used to measure the attributes and density of flights, among which child element C1 refers to international flights, child element C2 refers to the average number of flights on this route per day is less than 1, child element C3 refers to the route with one or two flights per day, and child element C4 refers to the route with more than two flights per day. The child elements in B2-B5 are data from the database. In the parent level B6, delay time can be divided into three levels, namely short delay time less than one hour, long delay time between 1 hour and 4 hours, and very long delay time greater than 4 hours. In each parent level, the upper child element has a higher priority than the lower child element.

[0055] Although elements within a parent hierarchy have clear priorities, the relationships between different parent hierarchies are unclear because they cannot be directly compared. In fact, dispatchers can only compare the priorities of different abnormal flight recovery attempts. To address this issue, the present invention applies a machine learning algorithm to optimize the unknown priorities between parent hierarchies based on decision-making information obtained through questionnaires and historical abnormal flight recovery data in a database.

[0056] Specifically, the questionnaire consisted of 26 questions, which were broken down into 36 paired comparisons of irregular flight recoveries with varying flight types, aircraft, passengers, and delay durations (see questionnaire structure). In addition to the questionnaire, 10 swap and replacement scenarios were extracted from historical irregular flight recovery data, assuming that irregular flights had higher priority than adjusted flights. This resulted in a total of 46 paired comparisons. It should be noted that some questionnaire results were not consistent with the final recovery data; that is, no single scoring system could explain all the results.

[0057] Therefore, the present disclosure develops a machine learning algorithm to calculate the optimal weights of different parent levels, so that the scoring unit 20 can account for most of the priorities in historical data and questionnaires.

[0058] Table 1: Questionnaire results for paired comparisons of different irregular flight recovery scenarios. The questionnaire included 26 questions in total, and two examples (questions 1 and 17) are shown in Table 1. The questionnaire was collected from dispatchers at China Southern Airlines' Beijing office. The results are shown in the priority column.

[0059]

[0060]

[0061] This method combines the ideas of support vector machines and linear optimization. We can consider an AHP model with m parent levels, where the weight of parent level i is w i ,and Assume that parent level i has n i elements, the relative weight of element j is y i (j), then In this paper, the priority of child elements in the parent level is determined through questionnaire survey, i.e., i (j) is known, but the weights w of different parent levels i i is unknown. In addition, through the questionnaire survey on pairwise irregular flight recovery, the priorities of K pairs of irregular flight recovery with different flight types, aircraft types, passenger types and delay times are known. Specifically, the irregular flight k can be defined as an m-dimensional vector z k =(y1 k , y2 k ,...,y m k ), is the weight vector of parent level i and can be calculated using the standard analytic hierarchy process based on the mutual comparison of the priorities of K pairs of irregular flight recovery with different flight types, aircraft types, passenger types and delay times. Then the irregular flight zk The score is Assume that the abnormal flight z k1 Than abnormal flight z k2 has a higher priority, (i.e., irregular flight z k1 Than abnormal flight z k2 has a higher impact), the scores of the two abnormal flights should satisfy z k1 >z k2 For the convenience of calculation, we can consider Δy i k =y i k1 -y i k2 ,therefore, This creates an inequality problem: solving for w i Make And Δz k >0.

[0062] In reality, the dispatcher may make wrong decisions, so some priority orders are not consistent. Therefore, no single scoring system can explain all inequalities. To solve this problem, we introduce the latent variable λ k ≤0, so that z k1 -z k2 >λ k , to replace z k1 -z k2 >0, the idea comes from the support vector machine of machine learning. In addition, each inequality is given a weight δ k , which is used to measure the two abnormal flights z k1 and z k2 The difference, Specifically, the comparison of similar abnormal flights should have a lower weight because the result carries less information. Now, we can try to make Maximize, i.e., the scoring unit should explain as much of the priority order as possible, which then leads to a standard linear optimization problem: by making Maximize and satisfy and Δz k ≥λ k This problem can be solved by a simple method, such as the method described in Jafari N, Zegordi SH. (2010). The airline perturbation problem: Considering disrupted passengers. Transportation planning Tech. 33(2), 203–220.

[0063] This method is compared with the standard logistic regression method LRM through numerical simulation, in which the weight values ​​w are randomly generated for different parent levels. i , and based on the generated weights, the questionnaire about the irregular flight comparison is answered. When the error rate e = 0, i.e., all priorities are consistent, the LRM performs better than the above-mentioned method of the present disclosure, where the estimated value is very close to the actual value. However, when there are random errors in the questionnaire, i.e., the error rate e > 0.1, the LRM will not converge to the actual value, while the method of the present disclosure will produce better results, i.e., a smaller standard deviation.

[0064] In the example of the present disclosure, there are 4 parent levels (B1, B2, B5, and B6) and 46 paired comparisons, where some priorities in the paired comparisons are not consistent. Applying this algorithm, the absolute weight values ​​for the parent levels flight (B1), aircraft (B2), passenger (B5), and delay time (B6) are 0.1217, 0.1020, 0.1688, and 0.6076, respectively. Figure 3 The absolute weight of each child element is shown, along with the absolute weight of the parent element. Note that the scores for parent elements B3 and B4 are not included in the analysis because, due to route reason B4, dispatchers are unable to recover irregular flights. Specifically, a route reason in this context refers to a situation where the aircraft is not faulty but the route cannot be used on schedule. For example, due to weather conditions or traffic control, flights cannot take off or land at an airport. Dispatchers can only adjust the aircraft, thus preventing recovery. In aircraft reason B3, abnormal operations due to specific reasons play a significant role in flight recovery, but these are not of concern to dispatchers. Specifically, aircraft reason B3 includes aircraft failure, aircraft maintenance, and previous delays (i.e., aircraft cannot perform their next mission on time). Dispatchers primarily adjust based on the impact of irregular flights, rather than the specific cause of the delay, and therefore do not consider these factors in the scoring mechanism. Undoubtedly, the three most important elements that influence the score are very long delays, VIP passengers, and long delays. In practice, the majority of adjustments are caused by long delays and VIP passengers. Other important elements include international flights and wide-body aircraft.

[0065] Overall, the scoring unit of the present disclosure was able to explain 88.9% (32 out of 36) of the priorities in the questionnaire, 77% of the replacements and 87% of the exchanges in the recovery data of the event type of the previous flight delay.

[0066] In addition, when solving the weight w of parent level i i Afterwards, the second scoring unit 203 scores the abnormal flight based on the weight vector z of the m-dimensional parent level. k =(y1 k , y2 k,...,y m k ) and the weight w corresponding to each weight vector i Calculate the score of irregular flights. In this paper, m is equal to 4.

[0067] Alternatively, since the importance of sub-elements within a hierarchy can be compared with each other, the weight vectors of these sub-elements can also be calculated using the standard hierarchical analysis method, so that the score of the irregular flight can be calculated based on the weight vectors of the sub-elements in each parent hierarchy of the irregular flight and the absolute weight corresponding to the weight vector of each sub-element.

[0068] return Figure 1 ,like Figure 1 As shown, the adjustment plan generating unit 30 includes: a recovery flight determining unit 301, which receives the score of each abnormal flight from the scoring unit 20, and determines whether the score of each abnormal flight is greater than a first predetermined threshold, and determines the abnormal flight with a score greater than the first predetermined threshold as a flight that needs to be recovered; a candidate flight generating unit 302, which generates a plurality of candidate adjustment flights for the flight that needs to be recovered based on the scores of the flight that needs to be recovered; and an adjustment plan selecting unit 303, which generates a plurality of adjustment plans based on the plurality of candidate adjustment flights and selects an adjustment plan with the lowest score from the generated plurality of adjustment plans for use in adjusting the flight that needs to be recovered.

[0069] In addition, when the restoration flight determination unit 301 determines that the number of flights that need to be restored is greater than 1, the candidate flight generation unit 302 generates respective candidate adjustment flights for the flights that need to be restored in order of priority from high to low according to the scores of the flights that need to be restored, and the adjustment plan selection unit 303 generates multiple adjustment plans for the flights that need to be restored based on multiple candidate adjustment flights in order of priority from high to low according to the scores of the flights that need to be restored, and selects the adjustment plan with the lowest score from the generated multiple adjustment plans for adjusting the flight that needs to be restored.

[0070] The adjustment plan selection unit 303 is further configured to determine whether the delay time of the flight to be resumed is greater than a second predetermined threshold, and when it is determined that the delay time of the flight to be resumed is greater than or equal to the second predetermined threshold, replace the flight to be resumed with a candidate adjustment flight, and when it is determined that the delay time of the flight to be resumed is less than the second predetermined threshold, swap the flight to be resumed with the candidate adjustment flight. The second predetermined threshold may be 4 hours.

[0071] After an adjustment such as a replacement or exchange, if the scores of some subsequent flights of a candidate adjustment flight exceed a threshold, especially for replacements, further changes may be required for these flights with scores exceeding the threshold. In this case, the adjustment plan selection unit 303 may be further configured to feed information about these flights with scores exceeding the threshold back to the candidate flight generation unit 302, so that the candidate flight generation unit 302 can generate multiple candidate adjustment flights for such flights, thereby enabling the adjustment plan selection unit 303 to further refine the adjustment plan.

[0072] Preferably, the candidate adjusted flight has the same destination as the flight that needs to be restored.

[0073] In addition, the adjustment plan generation unit 30 includes an adjustment plan output unit 304 for sorting and outputting all feasible adjustment plans according to their final scores. If no feasible adjustment plan is found, the adjustment plan with the lowest score is output. Furthermore, the adjustment plan output unit 304 can also be used to output the generated flight adjustment plan and the flight adjustment plan obtained from the dispatcher interview to a display for display to the user in a comparative manner. Furthermore, the adjustment plan output unit 304 can also output the scores and delay reduction levels of each adjusted flight adjustment plan in pairs for benefit analysis. The adjustment plan output unit 304 can also output the adjustment plan in the form of a web page.

[0074] In this article, a feasible adjustment plan means that the score of the flight that needs to be restored decreases after the adjustment and the sum of the scores of subsequent flights of the candidate adjustment flight does not increase after the adjustment, and the score of the adjustment plan means the sum of the scores of subsequent flights of the candidate adjustment flight.

[0075] Next, the process of generating an adjustment plan will be described. It consists of three steps. The first step is to determine the threshold at which the scoring system triggers the recovery process. If there are no abnormal flights or the impact of abnormal flights is small, for example, the flight is only delayed by a few minutes, the flight should have a score below the threshold, so that there is no need to change the flight itinerary. If the scores of some flights are higher than the threshold, it is necessary to start the recovery process and find a suitable adjustment plan. Through interviews with dispatchers, we set the threshold to 0.15, that is, Z0=0.15, that is, a long delay (>1 hour) or a short delay (0-1 hour) but VIP will trigger the recovery process.

[0076] Specifically, the delay is calculated according to the following formula:

[0077] Delay = Estimated arrival time (previous flight) + rotation time - planned departure time.

[0078] The second step is to find a feasible adjustment plan when starting the recovery process. Specifically, if the scores of more than two flights are higher than the threshold, the adjustment is made from the flight with the highest score. According to the recovery data, if the delay is relatively long, for example 1 to 4 hours, priority is given to exchange, and if the delay is very long (greater than 4 hours), priority is given to replacement. From the interview with the dispatcher, the change should be limited to flights located at the same airport, the previous flight arrives at the airport, and the arrival time of the previous flight should be close to the departure time of the flight that needs to be restored. In addition, the change is made as small as possible. Therefore, the search is limited to flights with an arrival time of 0 to 3 hours from the departure time of the flight that needs to be restored. However, all flights that can be used to adjust the flight network can be sorted according to their matching degree with the affected flights, where the matching degree depends on the aircraft type, crew and arrival time.

[0079] The third step is to evaluate different adjustment plans. For greater accuracy, the scores of the relevant flights before and after the adjustment can be checked. After the adjustment, the scores of some flights may become higher than the threshold, especially for replacements, which require further changes. In this case, the second step can be repeated to improve the adjustment plan. If the changes do not increase the total scores of all relevant flights, the plan is feasible. Note that a feasible adjustment plan may not exist or require more changes. Therefore, if no feasible adjustment plan is found in several repetitions (most changes are less than four repetitions), the search will be stopped. Finally, all feasible adjustment plans are sorted according to the scores of the final state and the results are output. If there is no feasible adjustment plan, the result with the lowest score will be output.

[0080] Next, it is explained how to reproduce the historical status from the operational data and compare the adjustment plan generated by the method of the present disclosure with the decision made by the dispatcher.

[0081] Data from three different sources is used to reconstruct the historical state: flight history, change logs (system-recorded recovery actions for irregular flights), and change records (manually recorded recovery actions for irregular flights). These databases contain different flight attributes and need to be combined to present a complete picture of the actions taken by the dispatcher. The historical state reconstruction logic is as follows: first, check the change log for the date, flight number, aircraft tail number, and adjustment reason. Then, use this information to find the adjustment time in the change record. Finally, extract the planned time and actual time from the flight history.

[0082] Table 2: Data from different sources

[0083]

[0084] In addition, actual recovery history data may be input into the second scoring unit 203 to verify whether the score of the recovered abnormal flight is greater than the first predetermined threshold, thereby verifying the effectiveness of the second scoring unit 203 .

[0085] The following will refer to Figure 4 A decision-making method for recovering an abnormal flight is described. The method begins with step S401, storing a dispatcher's decision information on recovering an abnormal flight and historical recovery data of the abnormal flight, and includes: step S402, constructing a scoring standard based on the decision information and the historical recovery data of the abnormal flight, and using the scoring standard to score the abnormal flight according to the information of the abnormal flight; step S403, determining abnormal flights with scores greater than a first predetermined threshold as flights that need to be recovered; step S404, determining the number of flights that need to be recovered; step S405, determining whether the number of flights that need to be recovered is greater than 1; when the judgment result is no, proceeding to step S406, generating multiple candidate adjustment flights for the flights that need to be recovered; when the judgment result in step S405 is yes, proceeding to step S407, sorting the flights that need to be recovered from high to low priority according to the scores of the flights that need to be recovered, and then proceeding to step S406. The method also includes executing step S408 after step S406, in which multiple adjustment plans are generated based on multiple candidate adjustment flights; and step S409, determining all feasible adjustment plans and sorting and outputting all feasible adjustment plans according to their final scores, and if there is no feasible adjustment plan, outputting the adjustment plan with the lowest score, and then proceeding to step S410, selecting the adjustment plan with the lowest score to adjust the flight that needs to be restored.

[0086] Furthermore, step S408 may include determining whether the delay time of the flight to be resumed is greater than a second predetermined threshold, and if so, replacing the flight to be resumed with a candidate adjustment flight, and if not, exchanging the candidate adjustment flight with the flight to be resumed. Note that after making adjustments such as replacements and exchanges, if the scores of some subsequent flights of the candidate adjustment flight become higher than the threshold, particularly for replacements, further changes may need to be made to the flights with scores higher than the threshold. In this case, step S408 may also include returning to step S406 to make further adjustments after making adjustments such as replacements and exchanges, thereby improving the adjustment plan. If the adjustments do not increase the total score of all relevant flights, the plan is feasible. Note that a feasible adjustment plan may not exist or require more adjustments. Therefore, if no feasible adjustment plan is found within several cycles (most changes are less than four cycles), the search will be stopped. If no feasible adjustment plan is found, the adjustment plan with the lowest score will be output.

[0087] The following will describe how the method of the present disclosure works with a specific example. Figure 6A and Figure 6B As shown, at 14:55, the system estimates that flight CZ3255 (tail number B6578) from Guangzhou to Beijing will be delayed by 57 minutes. As a result, flight CZ3684 from Beijing to Changbaishan using the same aircraft will be delayed by 62 minutes. Since the delay time is greater than 1 hour, the score of CZ3684 will be higher than 0.15, which triggers the recovery process. Next, all flights with arrival times between 16:20 and 19:20 are searched, and 4 flights are found, which are CZ0309, CZ3115, CZ3999 and CZ3902 (marked as plans 1-4, see Figure 6A ), these four flights can be swapped to adjust CZ3684. The third step evaluates the four adjustment plans. Table 2 shows their scores. Plan 1 (i.e., swapping CZ3684 and CZ3188) is the only feasible plan because it lowers CZ3684's score without increasing CZ3188's. In reality, the dispatcher would also take this recovery action. By swapping CZ3684 and CZ3188, CZ3684's delay remains constant.

[0088] Table 2: Scores obtained for different adjustment plans

[0089] The irregular flight is CZ3684, and Plan 1 (i.e., swapping CZ3684 and CZ3188) is the only feasible plan. B to N means changing the aircraft type from wide to narrow (which will increase the flight's score).

[0090]

[0091] Figure 5 FIG. 4 shows a decision-making method according to another embodiment of the present disclosure.

[0092] like Figure 5As shown, after determining a flight that needs to be restored and has a score greater than a first predetermined threshold, the method includes: Step S501: determining the event type of the flight that needs to be restored. If the event type is route-related, the process ends. If the event type is aircraft-related and the passengers of the flight that needs to be restored include VIP passengers, Step S502: determining whether the delay time is greater than a predetermined time, such as 4 hours. If it is determined to be greater than 4 hours, Step S503: searching for all flights that arrive at the airport where the flight needs to be restored and whose arrival time is close to the departure time of the flight that needs to be restored, and replacing the flight that needs to be restored with a flight that has a relatively high degree of match with the flight that needs to be restored and meets predetermined conditions, and outputting the result. If it is determined to be less than 4 hours, Step S504: searching for all flights that arrive at the airport where the flight needs to be restored and whose arrival time is close to the departure time of the flight that needs to be restored, and replacing the flight that needs to be restored with a flight that has a relatively high degree of match with the flight that needs to be restored and meets predetermined conditions, and outputting the result.

[0093] The predetermined condition may be that the score of the flight to be restored decreases after the adjustment and the sum of the scores of subsequent flights of the adjusted (exchanged or replaced) flight does not increase after the adjustment.

[0094] The idea of ​​the present disclosure is that, first, a database is established, in which flight information, aircraft information, passenger information and event information are stored. In addition, the database also stores the decision information of the dispatcher to restore abnormal flights, and the decision information can be obtained by interviewing the dispatcher and conducting a questionnaire on the dispatcher. The second step is to establish a scoring system, which can assign a real-time score to each flight, and the score represents the importance of the flight and the impact on system operations due to the current status / delay. Based on the scoring system, it can be evaluated whether the abnormal flight needs to be restored. The third step is to generate an adjustment plan based on the scoring system, in which the solution is first provided to the flight with a higher score, and the feasible adjustment plan needs to meet the condition of reducing the score of the related flight. The related flight may refer to the subsequent flight with the same aircraft tail number as the adjustment flight.

[0095] Compared to previous studies, this disclosure offers the following three advantages: First, it does not rely on a specific form of the objective function. In fact, this disclosure establishes a flight scoring system based on existing information to evaluate the impact of irregular flights and measures the bias of adjustment plans. Second, the disclosed method can generate several feasible adjustments for dispatchers to choose from. Furthermore, if a feasible adjustment plan does not exist, the algorithm can generate a plan with lower compensation. Third, the algorithm's runtime depends on the network size because the system focuses on local solutions rather than finding the global optimum.

[0096] The above is only a preferred embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as within the scope of protection of the present disclosure.

Claims

1. A decision support system for abnormal flight recovery, characterized in that: include: A storage unit storing the dispatcher's decision information on recovery of abnormal flights and historical recovery data of abnormal flights; a scoring unit, which constructs a scoring standard based on the decision information and the abnormal flight history recovery data, and uses the scoring standard to score the abnormal flight according to the abnormal flight information, and An adjustment plan generating unit is configured to determine the abnormal flight having a score greater than a first predetermined threshold as a flight that needs to be restored, determine multiple candidate adjustment flights for the flight that needs to be restored, generate multiple adjustment plans based on the multiple candidate adjustment flights, and select an adjustment plan with a lowest score from the generated multiple adjustment plans to be used for adjusting the flight that needs to be restored, wherein the score of the adjustment plan refers to the sum of the scores of subsequent flights of the candidate adjustment flight, and the time interval between the arrival time of the previous flight of each candidate adjustment flight and the departure time of the flight that needs to be restored is within a predetermined time. The storage unit stores a database structure constructed based on the decision information and the abnormal flight history recovery data, the database structure including flight data, aircraft data, passenger data, and delay time, and the scoring unit includes: The model building unit uses a hierarchical analysis method to build a hierarchical model for the flight data, the aircraft data, the passenger data, and the delay time based on an analysis of the decision information and the historical recovery data of the abnormal flights. The hierarchical model includes a target layer, a criterion layer, and an alternative solution layer. The target layer is for determining flights that need to be restored. The criterion layer includes at least the following parent layers: the flight data, the aircraft data, the passenger data, and the delay time. Each parent layer includes child elements. The alternative solution layer is for all abnormal flights. The sub-elements of the flight data include at least the flight type and flight density, the sub-elements of the aircraft data include at least the aircraft type, the sub-elements of the passenger data include at least VIP information, and the sub-elements of the delay time include at least a first delay time greater than a first predetermined time, a second delay time less than a second predetermined time, and a third delay time greater than the second predetermined time and less than the first predetermined time.

2. The decision support system according to claim 1, characterized in that: The scoring unit further includes: The first scoring unit uses the hierarchical model to calculate the relative weight of each child element of each parent level in the corresponding parent level based on the priorities of the pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of the pairs of different abnormal flights in the abnormal flight history recovery data.

3. The decision support system according to claim 2, characterized in that: The scoring unit further includes: The second scoring unit uses a machine learning algorithm to calculate the absolute weight of each parent level as the scoring criteria based on the relative weight of each child element in the corresponding parent level, based on the priorities of the pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of the pairs of different abnormal flights in the abnormal flight history recovery data.

4. The decision support system according to claim 3, characterized in that: The second scoring unit is also used to calculate the absolute weight of each child element according to the hierarchical model using the relative weight of each child element and the absolute weight of each parent level, and the second scoring unit calculates the score of the abnormal flight according to the information of the abnormal flight using the absolute weight of each parent level or calculates the score of the abnormal flight according to the information of the abnormal flight using the absolute weight of each child element.

5. The decision support system according to claim 3, characterized in that: Assuming that the number of parent levels is m, and parent level i includes ni child elements, the relative weight y of child element j in parent level i is i (j) Satisfaction And assuming that the priority of K for abnormal flights is extracted from the decision information and the abnormal flight history recovery data, the weight vector of abnormal flight k is z k =(y1 k ,y2 k ,…,y m k ),in, is the weight vector of parent level i, so the score of abnormal flight k is Among them, w i is the absolute weight of the parent level, assuming z k1 than z k2 has a higher priority, then z k1 >z k2 , and set Δy i k =y i k1 -y i k2 ,but Take λ k ≤0 so that Δz k ≥λ k , and λ k The weight is In satisfaction and Δz k ≥λ k Under the conditions of Maximize to find the absolute weight w of parent level i i .

6. The decision support system according to claim 1, characterized in that: The adjustment plan generating unit includes: a resumption flight determination unit, receiving the score of each abnormal flight from the scoring unit, determining whether the score of each abnormal flight is greater than the first predetermined threshold, and determining the abnormal flight having a score greater than the first predetermined threshold as the flight that needs to be resumed; a candidate flight generating unit, configured to generate the plurality of candidate adjustment flights for the flight that needs to be restored based on the score of the flight that needs to be restored; and An adjustment plan selection unit generates a plurality of adjustment plans based on the plurality of candidate adjustment flights and selects an adjustment plan with a lowest score from the generated plurality of adjustment plans for use in adjusting the flight that needs to be restored.

7. The decision support system according to claim 6, characterized in that: The candidate flight generating unit generates respective candidate adjustment flights for the flights that need to be restored in order of priority from high to low according to the scores of the flights that need to be restored, and the adjustment plan selecting unit generates multiple adjustment plans for the flights that need to be restored in order of priority from high to low according to the scores of the flights that need to be restored and selects an adjustment plan with a lowest score from the generated multiple adjustment plans for use in adjusting the flight that needs to be restored.

8. The decision support system according to claim 6, characterized in that: The adjustment plan selection unit is further configured to determine whether the delay time of the flight to be restored is greater than a second predetermined threshold, and when it is determined that the delay time of the flight to be restored is greater than or equal to the second predetermined threshold, use the candidate adjustment flight to replace the flight to be restored, and when it is determined that the delay time of the flight to be restored is less than the second predetermined threshold, exchange the flight to be restored with the candidate adjustment flight.

9. The decision support system according to claim 1, characterized in that: The candidate adjustment flight and the flight that needs to be restored have the same destination.

10. A decision-making method for abnormal flight recovery, characterized in that: include: Store the dispatcher's decision information on the recovery of abnormal flights and the historical recovery data of abnormal flights; Constructing a scoring standard based on the decision information and the abnormal flight history recovery data, and using the scoring standard to score the abnormal flight according to the abnormal flight information, determining the abnormal flight having a score greater than a first predetermined threshold as a flight that needs to be restored, and determining a plurality of candidate adjustment flights for the flight that needs to be restored; as well as Generating multiple adjustment plans based on the multiple candidate adjustment flights, and selecting an adjustment plan with a lowest score from the generated multiple adjustment plans for use in adjusting the flight that needs to be resumed, wherein the score of the adjustment plan refers to the sum of the scores of subsequent flights of the candidate adjustment flight, wherein the time interval between the arrival time of the preceding flight of each candidate adjustment flight at the airport and the departure time of the flight that needs to be resumed is within a predetermined time, wherein the decision-making method further includes: A database structure is constructed based on the decision information and the abnormal flight history recovery data, wherein the database structure includes flight data, aircraft data, passenger data and delay time. Based on the analysis of the decision information and the historical recovery data of the abnormal flights, a hierarchical model is constructed for the flight data, the aircraft data, the passenger data, and the delay time using the hierarchical analysis method. The hierarchical model includes a target layer, a criterion layer, and an alternative solution layer. The target layer is for determining the flights that need to be restored. The criterion layer includes at least the following parent layers: the flight data, the aircraft data, the passenger data, and the delay time. Each parent layer includes child elements. The alternative solution layer is for all abnormal flights. The sub-elements of the flight data include at least the flight type and flight density, the sub-elements of the aircraft data include the aircraft type, the sub-elements of the passenger data include VIP information, and the sub-elements of the delay time include a first delay time greater than a first predetermined time, a second delay time less than a second predetermined time, and a third delay time greater than the second predetermined time and less than the first predetermined time.

11. The decision-making method according to claim 10, characterized in that: The scoring criteria constructed based on the decision information and the abnormal flight history recovery data include: The relative weight of each child element of each parent level in the corresponding parent level is calculated based on the analysis of the decision information using the hierarchical model.

12. The decision-making method according to claim 11, characterized in that: Constructing a scoring standard based on the decision information and the abnormal flight history recovery data further includes: Based on the priorities of pairs of different abnormal flights determined by the dispatcher in the decision information and the priorities of pairs of different abnormal flights in the abnormal flight history recovery data, a machine learning algorithm is used to calculate the absolute weight of each parent level as the scoring criteria for the relative weight of each child element in the corresponding parent level.

13. The decision-making method according to claim 12, characterized in that: The scoring criteria constructed based on the decision information and the abnormal flight history recovery data also include: The absolute weight of each child element is calculated using the relative weight of each child element and the absolute weight of each parent level according to the hierarchical model, and Scoring the abnormal flight based on the information of the abnormal flight using the scoring criteria includes: The score of the abnormal flight is calculated based on the information of the abnormal flight using the absolute weight of each parent level, or the score of the abnormal flight is calculated based on the information of the abnormal flight using the absolute weight of each child element.

14. The decision-making method according to claim 12, characterized in that: Calculating the absolute weight of each parent level involves: Assuming that the number of parent levels is m, and parent level i includes ni child elements, the relative weight y of child element j in parent level i is i (j) Satisfaction And assuming that the priority of K for abnormal flights is extracted from the decision information and the abnormal flight history recovery data, the weight vector of abnormal flight k is z k =(y1 k ,y2 k ,…,y m k ),in, is the weight vector of parent level i, so the score of abnormal flight k is Among them, w i is the absolute weight of the parent level, assuming z k1 than z k2 has a higher priority, then z k1 >z k2 , and set Δy i k =y i k1 -y i k2 , but Take λ k ≤0 so that Δz k ≥λ k , and λ k The weight is In satisfaction and Δz k ≥λ k Under the conditions of Maximize to find the absolute weight w of parent level i i .

15. The decision-making method according to claim 10, before determining a plurality of candidate adjustment flights for the flight that needs to be restored, the decision-making method further comprises: The number of flights that need to be restored is determined, and when the number of flights that need to be restored is greater than 1, respective candidate adjustment flights are generated for the flights that need to be restored in descending order of priority based on the scores of the flights that need to be restored, and adjustment plans are determined for the flights that need to be restored in descending order of priority based on the scores of the flights that need to be restored.

16. The decision-making method according to claim 10, further comprising: Before generating multiple adjustment plans, determine whether the delay time of the flight that needs to be restored is greater than a second predetermined threshold. If so, use the candidate adjustment flight to replace the flight that needs to be restored. If not, exchange the candidate adjustment flight with the flight that needs to be restored.

17. The decision-making method according to claim 16, characterized in that: The candidate adjustment flight and the flight that needs to be restored have the same destination.

18. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 10 to 17 is implemented.

Images