Airport parking stand reallocation optimization method and system for abnormal operation scenarios

Abstract

This invention relates to the fields of airport operation management and intelligent transportation technology, specifically to an optimization method and system for airport parking stand reallocation under abnormal operational scenarios. The method includes: establishing a parking stand allocation model based on information about flights to be reallocated and airport parking stand resource information; constructing a macro-level apron allocation model and a micro-level parking stand allocation model based on the parking stand allocation model; constructing a reinforcement learning model based on the macro-level apron allocation model, and iterating the macro-level apron allocation model using the reinforcement learning model to obtain a target apron; inputting the target apron into the micro-level parking stand allocation model, optimizing the candidate parking stands within the target apron to obtain the target parking stand; obtaining feedback information corresponding to the target parking stand, and feeding this feedback information back to the reinforcement learning model to update the model and continue executing apron allocation for subsequent flights to be reallocated. This invention is applicable to the dynamic reallocation of airport parking stands under abnormal operational scenarios.

Description

Technical Field

[0001] This invention relates to the fields of airport operation management and intelligent transportation technology, specifically to an optimized method and system for airport parking space reallocation in response to abnormal operating scenarios. Background Technology

[0002] With the continuous expansion of air transport, airport parking stands, as a crucial resource for airport ground operations, directly impact the efficiency of aircraft turnaround support, apron operational order, and the overall operational capacity of the airport. Under normal airport conditions, parking stand allocation plans are typically pre-formulated based on flight schedules, aircraft attributes, parking stand resources, and airport operating rules. However, in abnormal operational scenarios such as flight delays, early flights, temporary parking stand closures, equipment failures, and extreme weather, the original parking stand allocation plan can easily become ineffective, necessitating a re-allocation of parking stands for affected flights to ensure the continuity of airport ground operations.

[0003] Currently, airport parking space reallocation methods mainly employ optimization techniques such as integer programming, mixed integer programming, and heuristic algorithms. These methods establish parking space allocation models and reallocate parking spaces that meet the constraints. While these methods demonstrate good solution capabilities for static planning or small-to-medium-scale problems, under abnormal operational scenarios, the constantly changing number of flights to be reallocated, real-time changes in parking space resources, and continuous shifts in operational constraints necessitate frequent model reconstruction and overall solution, resulting in high computational complexity and making it difficult to meet the real-time dynamic adjustment requirements of airports.

[0004] In recent years, reinforcement learning methods have been increasingly applied to airport resource scheduling. By learning resource allocation strategies through interaction between agents and the environment, dynamic decision-making capabilities have been improved. However, for the airport parking stand reallocation problem, directly using single-layer reinforcement learning to make decisions on all candidate parking stands can lead to low training efficiency, insufficient decision feasibility, and long solution time due to the large size of the state and action spaces, as well as the need to satisfy multiple constraints such as parking stand resource constraints, aircraft attribute constraints, and operational constraints. It is also difficult to balance the coordination between the overall airport resource allocation and the specific parking stand allocation. Summary of the Invention

[0005] This invention aims to solve the above-mentioned technical problems by providing an optimization method and system for airport parking stand reallocation under abnormal operating scenarios. It realizes a closed-loop collaborative decision-making mechanism between apron allocation and parking stand allocation in the airport parking stand reallocation process, and is applicable to the dynamic reallocation of airport parking stands under abnormal operating scenarios.

[0006] The objective of this invention can be achieved through the following technical solutions: An optimization method for airport parking stand reallocation under abnormal operational scenarios includes the following steps: S1. Obtain information on flights to be reassigned and airport parking space resources under abnormal operating scenarios, and establish a parking space allocation model based on the information on flights to be reassigned and airport parking space resources. S2. Based on the parking stand allocation model, construct a macro apron allocation model and a micro parking stand allocation model. The macro apron allocation model is used to determine the target apron corresponding to the flight to be reassigned, and the micro parking stand allocation model is used to determine the target parking stand based on the target apron. S3. Construct a reinforcement learning model based on the macro-level apron allocation model, and use the reinforcement learning model to iterate the strategy of the macro-level apron allocation model to obtain the target apron. S4. Input the target apron into the micro-parking space allocation model, optimize and solve the candidate parking spaces within the target apron, and obtain the target parking spaces; S5. Obtain feedback information corresponding to the target parking position and feed the feedback information back to the reinforcement learning model to update the reinforcement learning model so as to continue to execute the apron allocation for subsequent flights to be reassigned.

[0007] Furthermore, the parking space allocation model includes near-parking space configuration targets, apron load targets, taxiing distance targets, and conflict avoidance targets; The target for near-gate allocation is determined based on the number of flights awaiting reallocation assigned to near-gates. The target for near-gate allocation is as follows: in, apron The target value for the near-station configuration; Indicates the camera position The near and far camera position attributes; Indicates aircraft Have you been assigned to a work station? The aircraft collection within the planned time period is The target airport apron assembly point is The set of available aircraft stands at the target airport is ; Taxiing distance targets include taxiing distance targets within the apron and taxiing distance targets outside the apron. Taxiing distance targets within the apron are represented as follows: The target taxiing distance outside the apron is represented as: in, apron Target gliding distance within; This represents the target taxiing distance outside the apron; and They represent aircraft Taxis entry and exit distances within the apron; and They represent aircraft The taxiing distance and taxiing distance outside the apron; Indicates the camera position Does it belong to the apron? ; Apron load targets are determined based on the number of flights allocated to each apron and the number of parking spaces on each apron. Apron load targets are expressed as follows: in, Indicates the target value of the apron load; Indicates allocation to the apron The number of aircraft; apron The number of available workstations; The taxiing distance target is determined based on the taxiing in distance, taxiing out distance, taxiing in distance, and taxiing out distance corresponding to the flights to be reassigned. The taxiing distance target within the apron is expressed as: The target taxiing distance outside the apron is represented as: in, apron Target gliding distance within; This represents the target taxiing distance outside the apron; and They represent aircraft Taxis entry and exit distances within the apron; and They represent aircraft The taxiing distance and taxiing distance outside the apron; Indicates the camera position Does it belong to the apron? ; The conflict avoidance objective is determined based on the number of potential conflicts during the taxiing process of the flights to be reassigned on the apron. The conflict avoidance objective is expressed as: in, This indicates the conflict avoidance target value; Indicates aircraft On the tarmac Does it collide with the aircraft during the taxiing process? There is a potential conflict; Indicates aircraft On the tarmac Did it collide with the aircraft during the taxiing process? There is a potential conflict; potential conflicts are identified through a conflict determination time window. and minimum safe interval time Perform constraint discrimination.

[0008] Furthermore, a multi-objective integrated optimization function is constructed, which integrates the near-gate allocation objective, apron load objective, taxiing objective, and conflict avoidance objective to establish a multi-objective parking space allocation optimization function: in, Represents a multi-objective comprehensive optimization function; , , and These represent the weighting coefficients for the near-gate configuration target, apron load target, transit taxiing target, and conflict avoidance target, respectively, and each weighting coefficient is a non-negative number.

[0009] Furthermore, the parking space allocation model includes uniqueness constraints, minimum guarantee time constraints, aircraft attribute matching constraints, buffer time constraints, and potential conflict constraints. Uniqueness constraints are used to limit the allocation of each flight to a specific parking position for any given target flight. It can only be assigned to one parking position, and for any given parking position Under the same allocation criteria, a flight can be assigned to at most one target flight. The minimum guarantee time constraint is used to limit the duration of time that a flight awaiting reallocation remains parked at the parking stand, for any target flight. This ensures that the minimum support time requirement is met between the aircraft's push-out time and push-in time. Aircraft attribute matching constraints are used to define the matching relationship between parking positions and aircraft type, airline affiliation, and parking position support capabilities. For any given parking position... With any target flight The aircraft attribute matching relationship satisfies: in, Indicates the parking position With the target flight Corresponding attribute matching parameters between aircraft; when parking positions Meet the target flight When corresponding to the parking conditions of the aircraft ;otherwise, ; Buffer time constraints are used to limit the time interval between adjacent aircraft assigned to the same parking position; Potential conflict constraints are used to limit the taxiing time conflict relationships between flights awaiting reassignment and other flights on the same apron.

[0010] Furthermore, the macroscopic apron allocation model is a Markov decision process model; The state vector of the Markov decision process model includes: the state vectors of each apron during flight operations. Occupancy of various camera positions at arrival time With available quantity ,flight and before and after Aircraft type information ,flight landing runway information ,flight Before and after Arrival interval of aircraft And whether it is the same as the runway indicator vector The state vector of the Markov decision process model is represented as: Flight Choose a parking apron As a macroscopic apron allocation action; in the state vector Next action Then transition to the next state vector Establish a system targeting flights Execute action The corresponding instant reward function: in, , , , as well as These represent the target weight of apron load, the weight of taxiing path, the weight of apron conflict, the weight of lower-level target feedback, and the weight of delay, respectively. Indicates the target load on the apron. Indicates the target taxiing path outside the apron. Indicates the target of the apron conflict. This represents the lower-level target value fed back by the micro-machine position allocation model. Indicates the delay time caused by the taxiing conflict. This represents a penalty term introduced when there is no feasible solution in the lower-level model.

[0011] Furthermore, the macroscopic apron allocation model defines the mapping strategy between the state space and the action space. Using the maximization of long-term cumulative discount returns as the optimization objective, the optimal apron allocation strategy for the target flight is obtained by solving the problem. : in, This represents the discount factor, used to characterize the trade-off between current rewards and subsequent rewards; The macro-level apron allocation model outputs the apron assignment results corresponding to the target flight based on the optimal apron allocation strategy, and then transmits the apron assignment results to the micro-level gate allocation model.

[0012] Furthermore, the micro-level parking stand allocation model determines a set of candidate parking stands within the target apron. Based on the candidate parking stand set, proximity stand configuration targets, taxiing distance targets within the apron, uniqueness constraints, minimum guarantee time constraints, aircraft attribute matching constraints, buffer time constraints, and potential conflict constraints, it optimizes the matching relationship between the flight to be reassigned and the candidate parking stands in the candidate parking stand set. For any apron... Define the objective function of its microscopic aircraft parking allocation model within the apron. for: and These represent the target weight for near-gate configuration and the weight for taxiing paths within the apron, respectively. The parking position allocation result for the target flight within the target apron is obtained, and the objective function value corresponding to the parking position allocation result is fed back to the reward function of the macro apron allocation model.

[0013] Furthermore, reinforcement learning models include online Q-networks. With the target Q network , Using the macro-level apron allocation model as a reinforcement learning environment, an apron allocation agent is established. This agent selects the apron allocation action corresponding to the target flight based on the current state and receives an immediate reward based on the environmental transition result after the apron allocation action is executed. An online Q-network is used to learn the state-action value function, and a target Q-network is used to calculate the target Q-value. Construct a loss function based on the immediate reward, the next state, and the target Q value: in Indicates the parameters of the online Q network. Indicates the target Q-network parameters. Indicates the discount factor. Candidate actions to maximize the output function of the target Q-network; and updating the online Q-network parameters based on the loss function; The target apron corresponding to the target flight is output based on the updated online Q network, and the target apron is input into the micro-gate allocation model.

[0014] Furthermore, the micro-level parking space allocation model determines a set of candidate parking spaces based on the target apron, and combines the occupancy status, availability status, and flight support requirements of each parking space in the candidate parking space set. Under the conditions of uniqueness constraint, minimum support time constraint, aircraft attribute matching constraint, buffer time constraint, and potential conflict constraint, the matching relationship between the target flight and the candidate parking spaces is optimized and solved to obtain the target parking space. The system obtains feedback information corresponding to the target parking position and feeds the feedback information back to the reinforcement learning model to update the real-time reward and environmental state. The feedback information includes the lower-level optimization target value, the parking position occupancy status update result, and the candidate parking position set update result.

[0015] An airport parking stand reallocation optimization system for abnormal operational scenarios includes: The data acquisition module is used to acquire information on flights to be reassigned and airport parking space resources under abnormal operating scenarios. The parking stand allocation model construction module is used to build a parking stand allocation model based on the flight information to be reassigned and the airport parking stand resource information; The hierarchical model construction module is used to build a macro apron allocation model and a micro apron allocation model based on the parking stand allocation model. The macro apron allocation model is used to determine the target apron corresponding to the flight to be reassigned, and the micro apron allocation model is used to determine the target parking stand based on the target apron. The reinforcement learning module is used to build a reinforcement learning model based on the macro-level apron allocation model, and to iterate the macro-level apron allocation model through the reinforcement learning model to obtain the target apron. The aircraft parking position optimization solution module is used to input the target apron into the micro-aircraft parking position allocation model, optimize and solve the candidate parking positions within the target apron, and obtain the target parking position; The feedback update module is used to obtain feedback information corresponding to the target parking position and feed the feedback information back to the reinforcement learning model to update the reinforcement learning model so as to continue to execute the apron allocation for subsequent flights to be reallocated.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention provides an optimization method for airport parking stand reallocation under abnormal operating scenarios. It establishes a parking stand allocation model by acquiring flight information to be reallocated and airport parking stand resource information under abnormal operating scenarios. Based on the parking stand allocation model, it constructs a macro-level apron allocation model and a micro-level parking stand allocation model, performing hierarchical processing of apron allocation and parking stand allocation. A reinforcement learning model is used to iterate the macro-level apron allocation model to determine the target apron, and the micro-level parking stand allocation model is invoked based on the target apron to optimize the candidate parking stands. Simultaneously, feedback information corresponding to the target parking stand is fed back to the reinforcement learning model for model updates, forming a closed-loop collaborative decision-making mechanism between apron allocation and parking stand allocation in the airport parking stand reallocation process. This method is suitable for the dynamic reallocation of airport parking stands under abnormal operating scenarios. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall architecture of an airport parking space reallocation optimization method for abnormal operation scenarios provided by an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the collaborative decision-making process of a macroscopic apron allocation model and a microscopic aircraft stand allocation model, provided in an embodiment of the present invention. Figure 3 The training convergence curve of an airport parking space reallocation optimization method provided in this embodiment of the invention; Figure 4 A comparative diagram illustrating the optimization results of an airport parking space reallocation optimization method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the airport parking space reallocation result provided by an embodiment of the present invention; Figure 6 This is a schematic diagram of the results of a multi-objective optimization weight sensitivity analysis provided in an embodiment of the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0019] like Figures 1 to 6 The airport parking stand reallocation optimization method for abnormal operation scenarios, as shown, includes the following steps: S1. Obtain information on flights to be reassigned and airport parking space resources under abnormal operating scenarios, and establish a parking space allocation model based on the information on flights to be reassigned and airport parking space resources; determine the basic data and modeling objects required for parking space reassignment, and provide a data foundation for subsequent model construction.

[0020] S2. Based on the parking stand allocation model, construct a macro-level apron allocation model and a micro-level parking stand allocation model. The macro-level apron allocation model is used to determine the target apron corresponding to the flight to be reassigned, and the micro-level parking stand allocation model is used to determine the target parking stand based on the target apron. The parking stand reassignment problem is divided into two sub-problems: apron-level decision-making and parking stand-level solution, clarifying the determination logic of the target apron and the target parking stand.

[0021] S3. Construct a reinforcement learning model based on the macro apron allocation model, and use the reinforcement learning model to iterate the macro apron allocation model to obtain the target apron; learn the macro apron allocation strategy through the reinforcement learning model, and determine the target apron corresponding to the flight to be reassigned based on the strategy iteration results.

[0022] S4. Input the target apron into the micro-parking stand allocation model, optimize the candidate parking stands within the target apron to obtain the target parking stands; perform fine matching and solution on the candidate parking stands within the target apron area to determine the target parking stands corresponding to the flights to be reassigned.

[0023] S5. Obtain the feedback information corresponding to the target parking position and feed the feedback information back to the reinforcement learning model to update the reinforcement learning model so as to continue to perform apron allocation for subsequent flights to be reassigned; feed the solution result of the target parking position back to the reinforcement learning model to update the reinforcement learning model so that subsequent flights to be reassigned can continue to perform apron allocation based on the updated model.

[0024] Specifically, we define the set, parameters, and decision variables. The set is defined as follows: Let the set of aircraft within the planning time be... ,in For indexing a single aircraft within a planning period, The number of elements in the middle is denoted as Let the set of available aircraft stands at the target airport be... ,in This is the index of available aircraft stands at the target airport. The number of elements in the middle is denoted as Let the target airport apron assembly be... ,in An index of available aprons at the target airport. The number of elements in the middle is denoted as Set up an apron The set of positions in the middle is The number of its elements is denoted as .

[0025] The parameters are defined as follows: Let Indicates the camera position The camera position attribute, when the camera position When it is a near-station position, ,otherwise ;set up Indicates flight From the apron entrance to the aircraft stand The distance the aircraft can slide into from the tarmac. Indicates flight Self-positioning The distance from the apron exit to the apron exit; (This is the distance within the apron.) Indicates flight From the landing runway to the apron The distance of the taxiway outside the apron, Indicates flight self-propelled apron Taxiway distance from the runway to the outside of the apron; [Issued] Indicates flight Arrival time, Indicates flight Departure time Indicates flight The gliding speed, Indicates flight Required guarantee time; set Indicates the parking position Is it related to the flight? Attribute matching, during matching ,otherwise ;set up Indicates the parking position Does it belong to the apron? If it belongs to ,otherwise ;set up Indicates the parking position With parking position Do they belong to the same apron? If they both belong to the apron but ,otherwise ;set up This indicates the minimum safe interval between two aircraft at the same parking position. This indicates the size of the time window for determining potential conflicts. To represent a sufficiently large constant, This represents a sufficiently small constant.

[0026] The variables are defined as follows: Let These are 0-1 decision variables used to characterize aircraft. Has it been assigned to a parking position? If allocated ,otherwise ;set up , These are 0-1 variables, representing aircraft respectively. , Is it in an aircraft? Within the potential conflict judgment window when sliding in or out; set , A 0-1 variable, representing an aircraft On the tarmac When taxiing in or out, with the aircraft or Are there any potential conflicts?

[0027] Establish near-gate configuration goals. The near-gate configuration goals are established as follows: in, apron The target value for the near-station configuration; Indicates the camera position The near and far camera position attributes; Indicates aircraft Have you been assigned to a work station? The aircraft collection within the planned time period is The target airport apron assembly point is The set of available aircraft stands at the target airport is .

[0028] By maximizing the number of gates assigned to each apron, the utilization rate of gates is improved, passenger shuttle processes are reduced, and passenger transit efficiency is increased.

[0029] Establish apron load targets. The apron load targets are established as follows: in, Indicates the target value of the apron load; Indicates allocation to the apron The number of aircraft; apron The number of available workstations.

[0030] By minimizing the apron load target, the load-bearing pressure of each apron can be characterized and balanced, avoiding excessive concentration of aircraft that leads to excessive local support pressure and idleness of other aprons, thereby improving the overall utilization efficiency of apron resources.

[0031] Construct transit taxiing targets. Transit taxiing targets include taxiing distance targets within the apron and taxiing distance targets outside the apron.

[0032] The target taxiing distance within the apron is represented as: The target taxiing distance outside the apron is represented as: in, apron Target gliding distance within; This represents the target taxiing distance outside the apron; and They represent aircraft Taxis entry and exit distances within the apron; and They represent aircraft The taxiing distance and taxiing distance outside the apron; Indicates the camera position Does it belong to the apron? .

[0033] By minimizing the total taxiing distance inside and outside the apron, the ground operating path of aircraft can be shortened, taxiing time and operating energy consumption can be reduced, and ground turnaround efficiency can be improved.

[0034] Establish conflict avoidance goals. The conflict avoidance goals are established as follows: in, This indicates the conflict avoidance target value; Indicates aircraft On the tarmac Does it collide with the aircraft during the taxiing process? There is a potential conflict; Indicates aircraft On the tarmac Did it collide with the aircraft during the taxiing process? There is a potential conflict; potential conflicts are identified through a conflict determination time window. and minimum safe interval time Perform constraint discrimination.

[0035] By minimizing the number of potential conflicts, the risk of operational conflicts during taxiing in and out of the same apron can be reduced, coordination costs can be decreased, and apron operation safety and organizational efficiency can be improved.

[0036] Construct a multi-objective integrated optimization function. Integrate the objectives of near-gate allocation, apron load, taxiing, and conflict avoidance to establish a multi-objective parking space allocation optimization function.

[0037] In one implementation, a weighted summation method can be used to construct the comprehensive objective function, which is expressed as: in, Represents a multi-objective comprehensive optimization function; , , and These represent the weighting coefficients for the near-gate configuration target, apron load target, transit taxiing target, and conflict avoidance target, respectively, and each weighting coefficient is a non-negative number.

[0038] By jointly optimizing multiple objectives, we can achieve overall coordination among near-gate utilization, apron load balancing, taxiing distance control, and conflict avoidance, providing a basic optimization model for subsequent hierarchical decision-making.

[0039] Construct a unique constraint. The unique constraint is used to restrict each flight to be reassigned to only one parking position, and each parking position is not repeatedly assigned to multiple flights to be reassigned under the allocation conditions.

[0040] For any flight to be reassigned It can only be assigned to one parking position, as shown below: For any stop position Under the same allocation determination conditions, it can be allocated to at most one flight awaiting reallocation, as expressed as: in, This indicates the set of flights to be reassigned. This represents the set of parking positions. Indicates the parking position Flights awaiting reassignment The allocation relationship variable; when flights are to be reassigned Assigned to parking position hour, ;otherwise, .

[0041] Establish minimum support time constraints. Minimum support time constraints are used to ensure that transit aircraft have a minimum parking duration that meets ground support operation requirements from the time they enter the parking position until they leave the parking position.

[0042] For any flight to be reassigned The minimum support time requirement is met between the aircraft's pushback time and push-in time, as expressed as: in, Indicates flights awaiting reallocation The arrival time, Indicates flights awaiting reallocation The moment of departure, Indicates flights awaiting reallocation The minimum guarantee time corresponding to the aircraft, Indicates flights awaiting reallocation Corresponding to the aircraft's taxiing speed, Indicates flights awaiting reallocation The corresponding aircraft enters the parking position from the apron entrance. The gliding distance, Indicates flights awaiting reallocation The corresponding aircraft is from the approach side to the apron. The gliding distance, Indicates flights awaiting reallocation The corresponding aircraft is located at the parking position. Taxi distance from push-out to the apron exit Indicates flights awaiting reallocation The corresponding aircraft is located on the apron. The distance of the slide to the departure side.

[0043] Specifically, aircraft attribute matching constraints are constructed. These constraints ensure that the attributes of the parking positions and the attributes of the aircraft corresponding to the flights to be reassigned meet the compatibility requirements, thus preventing aircraft that do not meet the conditions for parking position usage from being assigned to the corresponding parking positions.

[0044] For any stop position With any flight to be reassigned The attribute matching relationship satisfies: in, Indicates the parking position Flights awaiting reassignment Corresponding attribute matching parameters between aircraft; when parking positions Meets the requirements for flight reassignment When corresponding to the parking conditions of the aircraft ;otherwise, In some implementations, the attribute matching parameters may include at least one or more of the following: aircraft type matching relationship, airline affiliation relationship, and gate support capability matching relationship.

[0045] Establish buffer time constraints. Buffer time constraints are used to ensure that there is a preset safe time interval between adjacent aircraft assigned to the same parking position, so as to avoid operational conflicts caused by continuous occupation of parking positions.

[0046] For any two flights to be reassigned and When both are assigned to the same parking position At that time, the occupancy relationship of its parking positions satisfies: in, This indicates the minimum buffer time between two adjacent aircraft at the same parking position. This indicates a pre-defined large constant. This indicates a preset small constant. This indicates the flight to be reassigned. Flights awaiting reassignment Auxiliary variables for determining the order of events.

[0047] Construct potential conflict constraints. Potential conflict constraints are used to reduce the probability of potential conflicts between flights awaiting reallocation and other aircraft on the same apron during taxiing or taxiing out after being assigned to candidate parking positions, thereby reducing operational interference caused by close taxiing times.

[0048] Potential conflict constraints regarding reassignment flights and taxiing aircraft: For reassignment flights Corresponding flight to another taxiing aircraft Their potential conflict relationships satisfy: in, This indicates the threshold for determining potential conflicts. This indicates the flight to be reassigned. Flights awaiting reassignment Auxiliary variables relating the time sequence during the sliding process Indicates flights awaiting reallocation Flights awaiting reassignment Is there a potential for conflict? Indicates flights awaiting reallocation Flights awaiting reassignment On the tarmac Is there a potential for conflict to slip into the interior? Indicates the parking position With parking position Are they located on the same apron? The associated parameters.

[0049] Potential conflict constraints regarding reassignment flights and taxiing aircraft: For reassignment flights Flight corresponding to the aircraft that has taxied out Their potential conflict relationships satisfy: in, This indicates the flight to be reassigned. Flights awaiting reassignment Auxiliary variables relating the time sequence of the sliding in and sliding out processes. Indicates flights awaiting reallocation Flights awaiting reassignment Are there any potential conflicts during the slide-in / slide-out process? Indicates flights awaiting reallocation Flights awaiting reassignment On the tarmac Is there a potential conflict between sliding in and sliding out?

[0050] Specifically, a macro-level apron allocation model is constructed. For flights awaiting reassignment due to the failure of the original parking stand allocation scheme under abnormal operational scenarios, a macro-level apron allocation model is constructed based on a Markov decision process to make assignment decisions for flights awaiting reassignment at the apron level. For any flight awaiting reassignment... Using state vectors To characterize its current operational state, the state vector must include at least: the state of each apron during flight operations. Occupancy of various camera positions at arrival time With available quantity ,flight and before and after Aircraft type information ,flight landing runway information ,flight Before and after Arrival interval of aircraft And whether it is the same as the runway indicator vector , is represented as: Flight The set of candidate aprons is defined as the action space. , will the flight Choose a parking apron As a macroscopic apron allocation action; the system is in a state Next action Then transition to the next state Its state transition relationship is expressed as .

[0051] By decomposing the objective function, in the macro-apron allocation model, for flights... Execute action The corresponding instant reward function Represented as: in, , , , as well as These represent the target weight of apron load, the weight of taxiing path, the weight of apron conflict, the target feedback weight of micro-aircraft position allocation, and the weight of delay, respectively. Indicates the target load on the apron. Indicates the target taxiing path outside the apron. Indicates the target of the apron conflict. This represents the target value for micro-aircraft position allocation fed back by the micro-aircraft position allocation model. Indicates the delay time caused by the taxiing conflict. This represents a penalty term introduced when there is no feasible solution in the micro-machine allocation model.

[0052] Furthermore, define the state space. To the action space The mapping strategy is With the goal of maximizing long-term cumulative discount returns, we seek the optimal apron allocation strategy. It is represented as: in, This represents the discount factor, used to characterize the trade-off between current rewards and subsequent rewards.

[0053] Based on this, the macro-level apron allocation model The optimization problem can be structured as follows: The solution of the macro-level apron allocation model is the apron assignment result for the flights to be reassigned, and the apron assignment result is then passed to the micro-level gate allocation model.

[0054] A micro-level aircraft parking allocation model is constructed. After determining the apron to which the flight to be reassigned belongs, a micro-level aircraft parking allocation model is constructed for the candidate parking positions within the determined apron to achieve optimal matching between flights and parking positions. By decomposing the objective function, for any apron... Define the objective function of its microscopic aircraft parking allocation model within the apron. for: in, and These represent the target weights for near-stand configuration and the weights for taxiing paths within the apron, respectively. During the optimization process of this model, all relevant constraints should be followed, including uniqueness constraints, minimum guarantee time constraints, aircraft attribute matching constraints, buffer time constraints, and potential conflict constraints.

[0055] By solving the micro-level parking space allocation model, the target parking spaces within the target apron for flights to be reassigned are obtained. The objective function value corresponding to the target parking space is then fed back to the immediate reward function in the macro-level apron allocation model, serving as... The value of the parameter is determined to achieve the linkage optimization between macro-level apron allocation and micro-level aircraft position allocation.

[0056] Specifically, a macro-level apron allocation reinforcement learning algorithm is constructed. The constructed macro-level apron allocation agent model is used as the macro-level apron allocation decision environment. An apron allocation agent is established, enabling the agent to allocate aprons based on the current state. Select flights to be reassigned apron allocation actions And obtain immediate rewards based on the environmental transition results after the action is executed. To achieve dynamic allocation decisions for flights awaiting reallocation at the apron level; a deep Q-network is used to approximate the state-action value function, constructing an online Q-network. With the target Q network The macro-level apron allocation strategy is updated based on the temporal difference error, where the loss function of the macro-level apron allocation agent is... Represented as: The target Q value is expressed as: in, Indicates the parameters of the online Q network. Indicates the target Q-network parameters. Indicates the discount factor. Indicates the target Q value. To select candidate actions that maximize the state-action value function output by the target Q-network, the parameters of the online Q-network are iteratively updated based on the loss function. Target aprons for flights to be reallocated are then generated based on the updated state-action value function, and the target aprons are passed to the micro-aircraft stand allocation algorithm for solution. Simultaneously, the feedback from the micro-aircraft stand allocation algorithm is used as part of the macro-aircraft stand allocation real-time reward to guide the iterative optimization of subsequent apron allocation strategies.

[0057] Specifically, a micro-level gate allocation optimization algorithm is constructed. The target apron for the flight to be reassigned is used as the input for micro-level gate allocation. Within the allocated apron, the constructed micro-level gate allocation model is invoked to optimize the matching relationship between the flight to be reassigned and the candidate parking positions, thereby obtaining the optimal target parking position for the flight to be reassigned within the corresponding apron. The micro-level gate allocation optimization algorithm integrates the current occupancy status, availability information, and flight support requirements of each parking position within the allocated apron, while satisfying the constraints of uniqueness, minimum support time, aircraft attribute matching, buffer time, and potential conflict constraints. Under these conditions, common commercial solvers (such as GUROBI, CPLEX, etc.) are used to solve for the gate allocation schemes of the flights to be reallocated. The objective function value, feasibility results, and environmental update information corresponding to the gate allocation schemes are fed back to the macro-level apron allocation reinforcement learning algorithm. The feedback information includes at least the micro-level gate allocation optimization objective value, gate occupancy status update results, and candidate gate set update results. The macro-level apron allocation reinforcement learning algorithm updates the reward value and environmental status based on the feedback information to achieve a two-layer linkage iterative optimization between apron allocation decision and gate allocation decision.

[0058] Using actual operational data from this busy airport, and combining iterative solutions with the constructed algorithm and model, the convergence curve is shown below. Figure 3 As shown. By Figure 3 It can be seen that the solution method proposed in this invention can effectively improve the convergence problem of existing methods during the solution process, and can obtain better optimization results under the same solution time conditions. Furthermore, the optimization results are as follows: Figure 4 As shown. By Figure 4 As can be seen, compared with the first-come, first-served model, the optimization using this invention can significantly improve the efficiency of parking space allocation. It demonstrates a clear optimization effect in several selected typical day scenarios, verifying the applicability and stability of this invention under different operating conditions. Figure 5 As shown, analysis of partial rack occupancy reveals that this invention also demonstrates good effectiveness in controlling continuous occupancy, effectively preventing unauthorized continuous service events and thus improving the feasibility and operational safety of rack allocation schemes. Regarding parameter sensitivity analysis, such as... Figure 6 As shown, the present invention can flexibly adjust the weights among multiple optimization objectives according to the actual airport operation and management needs, thereby achieving differentiated optimization under different application scenarios and meeting the specific needs of actual application units for operational efficiency, resource utilization and constraint control.

[0059] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0060] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0061] The present invention has been further described above with reference to specific embodiments. However, it should be understood that the specific description herein should not be construed as limiting the nature and scope of the present invention. Various modifications made to the above embodiments by those skilled in the art after reading this specification are all within the scope of protection of the present invention.

Claims

1. An optimization method for airport parking space reallocation under abnormal operational scenarios, characterized in that, Includes the following steps: S1. Obtain information on flights to be reassigned and airport parking space resources under abnormal operating scenarios, and establish a parking space allocation model based on the information on flights to be reassigned and airport parking space resources. S2. Based on the parking space allocation model, construct a macro apron allocation model and a micro parking space allocation model, wherein the macro apron allocation model is used to determine the target apron corresponding to the flight to be reassigned, and the micro parking space allocation model is used to determine the target parking space according to the target apron; S3. Construct a reinforcement learning model based on the macro-level apron allocation model, and perform policy iteration on the macro-level apron allocation model through the reinforcement learning model to obtain the target apron; S4. Input the target apron into the micro-aircraft parking allocation model, optimize and solve the candidate parking positions within the target apron, and obtain the target parking position; S5. Obtain feedback information corresponding to the target parking position and feed the feedback information back to the reinforcement learning model to update the reinforcement learning model so as to continue to execute the apron allocation for subsequent flights to be reassigned.

2. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 1, characterized in that, The parking space allocation model includes near-parking space configuration targets, apron load targets, taxiing distance targets, and conflict avoidance targets; The target for near-gate allocation is determined based on the number of flights to be reassigned to near-gate positions. The target for near-gate allocation is as follows: in, apron The target value for the near-station configuration; Indicates the camera position The near and far camera position attributes; Indicates aircraft Have you been assigned to a work station? The aircraft collection within the planned time period is The target airport apron assembly point is The set of available aircraft stands at the target airport is ; The taxiing distance target includes taxiing distance targets within the apron and taxiing distance targets outside the apron. The taxiing distance target within the apron is represented as follows: The target taxiing distance outside the apron is represented as: in, apron Target gliding distance within; This represents the target taxiing distance outside the apron; and They represent aircraft Taxis entry and exit distances within the apron; and They represent aircraft The taxiing distance and taxiing distance outside the apron; Indicates the camera position Does it belong to the apron? ; The apron load target is determined based on the number of flights allocated to each apron and the number of parking spaces in each apron. The apron load target is expressed as follows: in, Indicates the target value of the apron load; Indicates allocation to the apron The number of aircraft; apron The number of available workstations; The taxiing distance target is determined based on the taxiing in distance within the apron, taxiing out distance within the apron, taxiing in distance outside the apron, and taxiing out distance outside the apron corresponding to the flight to be reassigned. The taxiing distance target within the apron is expressed as: The target taxiing distance outside the apron is represented as: in, apron Target gliding distance within; This represents the target taxiing distance outside the apron; and They represent aircraft Taxis entry and exit distances within the apron; and They represent aircraft The taxiing distance and taxiing distance outside the apron; Indicates the camera position Does it belong to the apron? ; The conflict avoidance objective is determined based on the number of potential conflicts during the taxiing in and taxiing out processes of the flights to be reassigned on the apron. The conflict avoidance objective is expressed as follows: in, This indicates the conflict avoidance target value; Indicates aircraft On the tarmac Does it collide with the aircraft during the taxiing process? There is a potential conflict; Indicates aircraft On the tarmac Did it collide with the aircraft during the taxiing process? There is a potential conflict; potential conflicts are identified through a conflict determination time window. and minimum safe interval time Perform constraint discrimination.

3. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 2, characterized in that, A multi-objective integrated optimization function is constructed, which integrates near-gate allocation objectives, apron load objectives, taxiing objectives, and conflict avoidance objectives to establish a multi-objective parking space allocation optimization function: in, Represents a multi-objective comprehensive optimization function; , , and These represent the weighting coefficients for the near-gate configuration target, apron load target, transit taxiing target, and conflict avoidance target, respectively, and each weighting coefficient is a non-negative number.

4. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 1, characterized in that, The parking space allocation model includes uniqueness constraints, minimum guarantee time constraints, aircraft attribute matching constraints, buffer time constraints, and potential conflict constraints. The uniqueness constraint is used to limit the allocation of each flight to be reassigned to a single parking position for any target flight. It can only be assigned to one parking position, and for any given parking position Under the same allocation criteria, a flight can be assigned to at most one target flight. The minimum guarantee time constraint is used to limit the duration of time that the flight to be reassigned remains parked at the parking position for any target flight. This ensures that the minimum support time requirement is met between the aircraft's push-out time and push-in time. The aircraft attribute matching constraints are used to limit the matching relationship between parking positions and aircraft type, airline affiliation, and parking position support capabilities. For any parking position... With any target flight The aircraft attribute matching relationship satisfies: in, Indicates the parking position With the target flight Corresponding attribute matching parameters between aircraft; when parking positions Meet the target flight When corresponding to the parking conditions of the aircraft ;otherwise, ; The buffer time constraint is used to limit the time interval between adjacent aircraft assigned to the same parking position; The potential conflict constraints are used to limit the taxiing time conflict relationships between flights awaiting reassignment and other flights within the same apron.

5. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 1, characterized in that, The macro-level apron allocation model is a Markov decision process model. The state vector of the Markov decision process model includes: the flight status of each apron. Occupancy of various camera positions at arrival time With available quantity ,flight and before and after Aircraft type information ,flight landing runway information ,flight Before and after Arrival interval of aircraft And whether it is the same as the runway indicator vector The state vector of the Markov decision process model is represented as: Flight Choose a parking apron As a macroscopic apron allocation action; in the state vector Next action Then transition to the next state vector Establish a system targeting flights Execute action The corresponding instant reward function: in, , , , as well as These represent the target weight of apron load, the weight of taxiing path, the weight of apron conflict, the weight of lower-level target feedback, and the weight of delay, respectively. Indicates the target load on the apron. Indicates the target taxiing path outside the apron. Indicates the target of the apron conflict. This represents the lower-level target value fed back by the micro-machine position allocation model. Indicates the delay time caused by the taxiing conflict. This represents a penalty term introduced when there is no feasible solution in the lower-level model.

6. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 5, characterized in that, The macro-apron allocation model defines a mapping strategy between the state space and the action space. Using the maximization of long-term cumulative discount returns as the optimization objective, the optimal apron allocation strategy for the target flight is obtained by solving the problem. : in, This represents the discount factor, used to characterize the trade-off between current rewards and subsequent rewards; The macro-apron allocation model outputs the apron assignment result corresponding to the target flight according to the optimal apron allocation strategy, and transmits the apron assignment result to the micro-apron allocation model.

7. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 5, characterized in that, The micro-level parking space allocation model determines a set of candidate parking spaces within the target apron. Based on the candidate parking space set, the near-parking space configuration target, the apron taxiing distance target, the uniqueness constraint, the minimum guarantee time constraint, the aircraft attribute matching constraint, the buffer time constraint, and the potential conflict constraint, it optimizes the matching relationship between the flight to be reassigned and the candidate parking spaces in the candidate parking space set. For any apron... Define the objective function of its microscopic aircraft parking allocation model within the apron. for: and These represent the target weight for near-gate configuration and the weight for taxiing paths within the apron, respectively; the parking position allocation result for the target flight within the target apron is obtained, and the objective function value corresponding to the parking position allocation result is fed back to the reward function of the macro apron allocation model.

8. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 5, characterized in that, The reinforcement learning model includes an online Q-network. With the target Q network , Using the macro-level apron allocation model as a reinforcement learning environment, an apron allocation agent is established. The apron allocation agent selects the apron allocation action corresponding to the target flight based on the current state, and obtains an immediate reward based on the environmental transition result after the apron allocation action is executed. The online Q-network is used to learn the state-action value function, and the target Q-network is used to calculate the target Q-value. Construct a loss function based on the immediate reward, the next state, and the target Q value: in Indicates the parameters of the online Q network. Indicates the target Q-network parameters. Indicates the discount factor. Candidate actions to maximize the output function of the target Q-network; and updating the parameters of the online Q-network based on the loss function; The target apron corresponding to the target flight is output according to the updated online Q network, and the target apron is input into the micro-aircraft stand allocation model.

9. The airport parking space reallocation optimization method for abnormal operation scenarios according to claim 7, characterized in that, The micro-aircraft stand allocation model determines a set of candidate parking stands based on the target apron, and combines the occupancy status, availability status, and flight support requirements of each parking stand in the candidate parking stand set. Under the conditions of uniqueness constraint, minimum support time constraint, aircraft attribute matching constraint, buffer time constraint, and potential conflict constraint, the model optimizes the matching relationship between the target flight and the candidate parking stands to obtain the target parking stand. The feedback information corresponding to the target parking position is obtained and fed back to the reinforcement learning model to update the immediate reward and the environmental state. The feedback information includes the lower-level optimization target value, the parking position occupancy status update result, and the candidate parking position set update result.

10. An airport parking stand reallocation optimization system for abnormal operational scenarios, characterized in that: include: The data acquisition module is used to acquire information on flights to be reassigned and airport parking space resources under abnormal operating scenarios. The parking stand allocation model construction module is used to build a parking stand allocation model based on the flight information to be reassigned and the airport parking stand resource information; The hierarchical model construction module is used to build a macro apron allocation model and a micro apron allocation model based on the parking stand allocation model. The macro apron allocation model is used to determine the target apron corresponding to the flight to be reassigned, and the micro apron allocation model is used to determine the target parking stand based on the target apron. The reinforcement learning module is used to build a reinforcement learning model based on the macro-level apron allocation model, and to iterate the macro-level apron allocation model through the reinforcement learning model to obtain the target apron. The aircraft parking position optimization solution module is used to input the target apron into the micro-aircraft parking position allocation model, optimize and solve the candidate parking positions within the target apron, and obtain the target parking position; The feedback update module is used to obtain feedback information corresponding to the target parking position and feed the feedback information back to the reinforcement learning model to update the reinforcement learning model so as to continue to execute the apron allocation for subsequent flights to be reallocated.

Images