CN116432806A - Rolling prediction method and system for flight ground guarantee node time - Google Patents
Rolling prediction method and system for flight ground guarantee node time Download PDFInfo
- Publication number
- CN116432806A CN116432806A CN202211625138.1A CN202211625138A CN116432806A CN 116432806 A CN116432806 A CN 116432806A CN 202211625138 A CN202211625138 A CN 202211625138A CN 116432806 A CN116432806 A CN 116432806A
- Authority
- CN
- China
- Prior art keywords
- time
- flight
- characteristic
- predicted
- airport
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000005096 rolling process Methods 0.000 title claims abstract description 31
- 230000010006 flight Effects 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims description 22
- 238000010801 machine learning Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 6
- 125000004122 cyclic group Chemical group 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000004140 cleaning Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000003066 decision tree Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 241001522296 Erithacus rubecula Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000003245 working effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Tourism & Hospitality (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Business, Economics & Management (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a rolling prediction method and a rolling prediction system for flight ground guarantee node time, comprising the following steps: acquiring a first air characteristic, a first airport operation characteristic and a flight characteristic; inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, wherein the first prediction model outputs ideal time of each flight service node; acquiring the completed actual time of the flight service node, and determining the time difference between the actual time and the ideal time; inputting a second weather feature, a second airport feature and the time difference into a second prediction model, wherein the second prediction model outputs the predicted time of the rest flight service nodes; repeatedly determining the predicted time of the rest flight service nodes until the flights leave the port; to further increase the operating efficiency of the airport.
Description
Technical Field
The invention relates to the technical field of flight support, in particular to a rolling prediction method and a rolling prediction system for flight ground support node time.
Background
In the process from the departure of a front station to the landing and then to the departure of a local station, a series of ground guarantee nodes need to be completed, for example: a series of ground guarantee nodes such as entering station, boarding, opening cabin door, unloading luggage, loading luggage, closing cabin door, feeding aviation food, feeding aviation oil, starting boarding and withdrawing wheel guard, and after finishing the necessary ground guarantee nodes, the flight can take off again. TAMS (Total Ai rport Management System) the system, also called as airport global management system, can predict the time of the ground guarantee nodes of the flights, and then makes decisions on the sequencing of the flights, the optimization of resources, the guidance of passengers and the like according to the prediction result, so that the airport operation efficiency is improved, the operation refinement degree is improved, and the traveling experience of passengers is improved. The prediction method commonly used at present comprises two modes of manual prediction and prediction based on a mathematical model. However, the prior art does not have the ability to roll prediction in real-time synchronization with the flight status and cannot characterize the impact of weather and airport operating conditions on the flight service node time. The existing mathematical model does not consider weather characteristics and airport operation characteristics, so that the weather characteristics and the airport operation characteristics cannot be associated with the time prediction of the flight service node, and in fact, the weather characteristics and the airport operation characteristics have great influence on the flight service node.
In view of the above, the invention provides a rolling prediction method and a system for time of a ground guarantee node of a flight, which are used for triggering a rolling prediction mechanism when the flight advances from a kth node to a k+1 node, and re-predicting and updating a time prediction value of a node to be processed so as to cope with changeable emergency situations of the airport, guide the airport to quickly make temporary adjustment and cope with plans, and further improve the operation efficiency of the airport.
Disclosure of Invention
The invention aims to provide a rolling prediction method for flight ground guarantee node time, which comprises the following steps: acquiring a first air characteristic, a first airport operation characteristic and a flight characteristic; inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, wherein the first prediction model outputs ideal time of each flight service node; the ideal time refers to the predicted time for completing the flight service node; acquiring the completed actual time of the flight service node, and determining the time difference between the actual time and the ideal time; the actual time refers to the time of completing the flight service node; inputting a second weather feature, a second airport feature and the time difference into a second prediction model, wherein the second prediction model outputs the predicted time of the rest flight service nodes; the predicted time refers to the predicted time for completing the rest flight service nodes; the remaining flight service nodes refer to incomplete flight service nodes; and repeatedly determining the predicted time of the rest flight service nodes until the flights leave the port.
Further, the first weather characteristic and the second weather characteristic include rainfall, cloud cover height, wind speed, wind direction, barometric pressure, and weather type.
Further, the first airport operational characteristic and the second airport operational characteristic include a flight normal rate and a mission timeout rate.
Further, the flight characteristics include planned departure time, planned arrival time, model, and airline.
Further, the inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, the first prediction model outputting ideal time of each flight service node includes: processing the first air characteristic, the first airport operation characteristic and the flight characteristic according to a tensor generation rule to generate a one-dimensional tensor I:
I=(x 1 ,x 2 ,x 3 ,x 4 ,x 5 …,x m )
wherein x is 1 ,x 2 ,x 3 ,x 4 ,x 5 …,x m Representing characteristic parameters, m representing the total number of characteristic parameters; the characteristic parameter is related to the first air characteristic, the first airport operational characteristic, or the flight characteristic;
calculating the ideal time S by taking the one-dimensional tensor as an input of the first prediction model i :
Wherein,,representing ideal times for each flight service node, n representing the total number of flight service nodes, and G (I) representing a loop call for a multiple output model or a single output model.
Further, the acquiring the actual time of the completed flight service node, determining a time difference between the actual time and the ideal time, includes: generating an actual time sequence S based on the actual time r :
S r =(t 1 ,t 2 ,…,t k )
Wherein t is 1 ,t 2 ,…,t k Representing the actual time of each completed flight service node, k representing the total number of completed flight service nodes, 1<k<n; n represents the total number of flight service nodes;
based on the actual time and the ideal time, a sequence Δ of the time differences is obtained:
Δ=S r -S i =(δ 1 ,δ 2 ,…,δ k ,0,…,0)
wherein S is i Representing the sequence of ideal times, delta 1 ,δ 2 ,…,δ k Indicating the time difference between the completed flight service node and the corresponding ideal time, "0" indicates zero padding terms, for a total of n-k.
Further, the inputting the second weather feature, the second airport feature and the time difference into a second prediction model, the second prediction model outputting the predicted time of the remaining flight service nodes, includes: inputting the second weather feature, the second airport feature and the sequence of time differences delta into the second predictive model to obtain a complete sequence of time differences delta ′ :
Wherein,,representing a predicted time difference, the predicted time difference being a predicted time difference between a time at which the remaining flight service node is completed and the ideal time, G (delta) representing a cycle of a multiple output model or a single output modelCalling;
summing the complete time difference sequence delta' with the ideal time sequence to obtain the predicted time sequence S p :
Further, the inputting the second weather feature, the second airport feature, and the sequence of time differences Δ into the second prediction model, to obtain a complete sequence of time differences Δ', includes: predicting a first predicted time difference based on the second weather feature, the second airport feature, and the time difference; the first predicted time difference is the time difference between the completion time of a first node of the predicted remaining flight service nodes and the corresponding ideal time; predicting a second predicted time difference based on the second weather feature, the second airport feature, the time difference, and the first predicted time difference; the second predicted time difference is the time difference between the completion time of the second node of the predicted remaining flight service nodes and the corresponding ideal time; repeatedly obtaining the predicted time difference of the rest flight service nodes based on the second weather feature, the second airport feature, the time difference and the predicted time difference; and obtaining the complete time difference sequence based on the time difference and the predicted time difference.
Further, the second prediction model is a machine learning model, and the second prediction model may be obtained by training an initial second prediction model, including: inputting a training sample into the initial second predictive model; constructing a loss function based on the output of the initial second prediction model and the label of the training sample; the training sample is obtained by summing a unit matrix with dimension of n multiplied by n, a second weather feature, a second airport feature and a transposed sequence of historical time differences, and the label is obtained by shifting the transposed sequence of historical time differences one bit later; iteratively updating parameters of the initial second model based on the loss function until the test precision of the initial second prediction model reaches a preset precision threshold or the iteration number is larger than a preset number threshold; and taking the initial second prediction model with completed iteration as the second prediction model.
The invention aims to provide a rolling prediction system for flight ground guarantee node time, which comprises a first acquisition module, an ideal time determination module, a time difference determination module, a prediction time determination module and a circulation module, wherein the first acquisition module is used for acquiring the time difference of the flight ground guarantee node time; the first acquisition module is used for acquiring first air characteristics, first airport operation characteristics and flight characteristics; the ideal time determining module is used for inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, and the first prediction model outputs ideal time of each flight service node; the ideal time refers to the predicted time for completing the flight service node; the time difference determining module is used for obtaining the completed actual time of the flight service node and determining the time difference between the actual time and the ideal time; the actual time refers to the time of completing the flight service node; the prediction time determining module is used for inputting a second weather feature, a second airport feature and the time difference into a second prediction model, and the second prediction model outputs the prediction time of the rest flight service nodes; the predicted time refers to the predicted time for completing the rest flight service nodes; the remaining flight service nodes refer to incomplete flight service nodes; the circulation module is used for repeatedly determining the predicted time of the rest flight service nodes until the flights come out.
The technical scheme of the embodiment of the invention has at least the following advantages and beneficial effects:
before a flight starts, the rolling prediction method and the rolling prediction system for the time of the ground guarantee node of the flight can conduct primary prediction on the ideal time of the flight service node, and the prediction result can provide important reference for airport scheduling plans; in the execution process of the flight, the invention can carry out rolling prediction on the time of the flight service node, can provide important reference for real-time scheduling of the airport, and simultaneously provides basis for overtime alarming when the predicted time deviates seriously.
The invention divides the flight support service nodes into two types of the incoming port direction and the outgoing port direction, carries out predictive modeling in a classified manner, solves the problem of predicting the time of the stop flight service nodes, is not limited to a specific flight service node set, and has the capability of popularizing the model to any flight support flow (any airport).
The invention adopts a method of combining a regression algorithm (a first prediction model) and a cyclic neural network (a second prediction model) to predict the time of the flight service node, uses the regression algorithm to predict an ideal time sequence, and uses the cyclic neural network to predict a deviation value sequence of the ideal time and the actual time. The short-term memory capacity of the cyclic neural network can better represent the influence of delay of the previous time node on the subsequent node, so that the prediction accuracy of the model on the time sequence is improved. The method combining the two algorithms can characterize the catch-up capability of airport staff to the planning time (target time) after delay occurs, so that the model has the evaluation capability to the working effect.
The invention brings the weather characteristics and the airport operation characteristics into the machine learning model, thereby having the capability of taking the weather characteristics and the airport operation characteristics as input prediction time and being capable of better representing the influence of weather and airport operation conditions.
In the invention, the regression model does not directly predict the absolute time (Beijing time), but predicts the time interval of a certain node relative to the previous node, and then restores the absolute time (Beijing time) through a preset recombination rule. The process solves the problem of poor scalability of absolute time (Beijing time), and can improve the upper limit of accuracy of the machine learning model.
Drawings
FIG. 1 is an exemplary flow chart of a rolling prediction method for flight ground assurance node time provided by some embodiments of the present invention;
fig. 2 is an exemplary block diagram of a rolling prediction system for flight ground assurance node time according to some embodiments of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Fig. 1 is an exemplary flowchart of a rolling prediction method for a flight ground assurance node time according to some embodiments of the present invention. In some embodiments, the process 100 may be performed by the system 200.
As shown in fig. 1, the process 100 includes the following:
The first weather characteristic may refer to a characteristic of weather of the predicted flight-executing flight-service node. In some embodiments, the first weather characteristic may include characteristics of predicted rainfall, cloud cover, wind speed, wind direction, barometric pressure, and weather type, among other data. For example, characteristics of weather during the predicted arrival and departure of flights. In some embodiments, the first weather characteristic may be obtained by weather forecast. For example, the first air data is acquired from the airport bus via an interface protocol.
The first airport operational characteristic may refer to a characteristic of flight operation in the terminal. The first airport operational characteristic may include a flight normal rate and a task timeout rate. In some embodiments, the first airport operational characteristic may be obtained from an airport bus via an interface protocol.
A flight characteristic may refer to a characteristic of flights entering and exiting a port. The flight characteristics may include planned departure time, planned arrival time, model and airline class, etc. In some embodiments, the flight characteristics may be obtained from the airport bus via an interface protocol.
The first predictive model may include, but is not limited to, machine learning models of regression algorithms such as LGBM, decision trees, random forests, adaboost, and MLP.
In some embodiments, an algorithm (LGBM, decision tree, random forest, adaboost, and/or MLP, etc.) of the first predictive model may be first determined to obtain an initial first predictive model.
A dataset for training an initial first predictive model is obtained, the dataset comprising a training set and a testing set. In some embodiments, the data set is derived from historical data of flights in the terminal. Wherein the dataset is input with weather features, airport features and flight features and the node time data is output:
D A (j,t)=[I(t),S(j)]
wherein D is A (j, t) represents a dataset, I (t) represents an input, i.e. weather features, airport features and flight features at time t, S (j) represents an output, i.e. real data that the flight service node of flight j at time t has completed, j represents a flight, t represents any time t within the lifecycle of flight j.
Training and iterating the initial first prediction model by using a large amount of historical data until the accuracy of the model in the test set reaches a certain threshold value or reaches a certain iteration number, for example, when the result ratio of the node time error within 15min exceeds 90%, or the iteration number exceeds 10000 times, stopping training, and taking the initial first prediction model after training as the first prediction model.
A flight service node may refer to a node where a flight performs a service. For example, the nodes of a given flight guarantee flow are divided into a departure direction and a departure direction, and then the result can be expressed as a link for reducing the total weight of the aircraft or serving the departure of the aircraft as a departure direction node; the link that represents the result as increasing the gross weight of the aircraft or that serves the departure of the aircraft is the departure node. Illustratively, the inbound flight service node may include inbound, up-gear and off-load, off-passenger, clean, etc. services that are strongly related to inbound time; the flight service nodes of departure can comprise services of feeding, feeding oil, loading, boarding and withdrawal, and the like which are strongly related to departure time. In some embodiments, the flight service node may be acquired based on the status of the flight.
The ideal time refers to the predicted time to complete the flight service node. For example, for inbound flights, the ideal time may include a predicted time for the flight to get in, a predicted time for boarding and alighting, a predicted time for cleaning, and the like.
In some embodiments, inputting the first air characteristic, the first airport operating characteristic, and the flight characteristic into a first predictive model, the first predictive model outputting ideal times for each flight service node, comprising:
processing the first air characteristic, the first airport operation characteristic and the flight characteristic according to a tensor generation rule to generate a one-dimensional tensor I:
I=(x 1 ,x 2 ,x 3 ,x 4 ,x 5 …,x m )
wherein x is 1 ,x 2 ,x 3 ,x 4 ,x 5 …,x m Representing characteristic parameters, m representing the total number of characteristic parameters; the characteristic parameter is associated with a first air characteristic, a first airport operating characteristic, or a flight characteristic;
calculating ideal time S by taking one-dimensional tensor as input of first prediction model i :
Wherein,,representing ideal time of each flight service node, n represents flight serviceThe total number of nodes, G (I), represents the round robin invocation of a multiple output model or a single output model.
In some embodiments, the method further includes classifying all the flight service nodes into two types of inbound and outbound according to a preset rule, selecting a planned inbound time as a reference point of the inbound node and a planned outbound time as a reference point of the outbound node; calculating the time difference between all nodes and the reference point, and taking the time difference as training set output data Y of the first prediction model in the training link; training and a first predictive model, the output at the time of application becomes relative time. In some embodiments, all nodes may also take their neighboring nodes as reference points. The adjacent node, the port entering node uses the front node in the flow as a reference point, the port leaving node uses the subsequent node in the flow as a reference point, the start point and the end point of the flow are two nodes of port entering and port leaving, and the planned port entering time and the planned port leaving time are used as reference points.
The actual time refers to the time at which the flight service node is completed. For example, for inbound flights, as time passes, the flight has completed inbound, the actual time of inbound may be obtained.
In some embodiments, obtaining the actual time of the completed flight service node, determining the time difference between the actual time and the ideal time, includes:
generating an actual time sequence S based on the actual time r :
S r =(t 1 ,t 2 ,…,t k )
Wherein t is 1 ,t 2 ,…,t k Representing the actual time of each completed flight service node, k representing the total number of completed flight service nodes, 1<k<n; n represents the total number of flight service nodes.
Based on the actual time and the ideal time, a sequence of time differences Δ is obtained:
Δ=S r -S i =(δ 1 ,δ 2 ,…,δ k ,0,…,0)
wherein S is i Representing the sequence of ideal times, delta 1 ,δ 2 ,…,δ k Indicating the time difference between the completed flight service node and the corresponding ideal time, "0" indicates zero padding terms, for a total of n-k.
And 140, inputting the second weather feature, the second airport feature and the time difference into a second prediction model, and outputting the predicted time of the rest flight service node by the second prediction model. In some embodiments, step 140 may be performed by the predicted time determination module 240.
When a flight proceeds to node k, weather and airport operating conditions may change, thus obtaining up-to-date weather and airport feature data. The second weather characteristic may refer to a characteristic of weather at the completion of the flight service node. In some embodiments, the second weather characteristic may include characteristics of data such as actual rainfall, cloud deck elevation, wind speed, wind direction, barometric pressure, and weather type. In some embodiments, the stored up-to-date weather data may be retrieved from the airport bus at intervals (e.g., one hour) via an interface protocol.
The second airport operational characteristic may refer to a characteristic of the actual operation of a flight in the terminal. The second airport operational characteristic may include the actual normal rate of flights and the task timeout rate. In some embodiments, the second airport feature may be acquired from the airport bus via an interface protocol.
The remaining flight service nodes refer to incomplete flight service nodes. In some embodiments, the remaining flight service nodes may be acquired in various possible ways. The predicted time refers to the predicted time to complete the remaining flight service nodes. For example, when a flight enters a port, the completed node is an inbound station, and the rest of the flight service nodes comprise boarding and unloading, alighting, cleaning and the like; the predicted time is the time for completing the business of gear loading, unloading, passenger getting off, cleaning and the like.
The second predictive model may be a machine learning model based on a Recurrent Neural Network (RNN) algorithm, including but not limited to RNN, LSTM, GRU, etc. algorithms.
In some embodiments, the second predictive model may be derived by training an initial second predictive model, comprising:
the training samples are input into an initial second predictive model. Wherein an algorithm such as initial second prediction mode selection RNN, LSTM, GRU and their variants.
Constructing a loss function based on the output of the initial second prediction model and the label of the training sample; the training sample is obtained by summing up a unit matrix with dimension of n multiplied by n, a second weather feature, a second airport feature and a transposed sequence of the historical time difference, and the label is obtained by shifting the transposed sequence of the historical time difference one bit.
Illustratively, historical meteorological data is obtained from meteorological monitoring stations near an airport, and weather characteristic data of a period of time is extracted at intervals of one hour, wherein the weather characteristic data comprise rainfall, cloud base height, wind speed, wind direction, air pressure, weather type and the like; acquiring historical operation data of an airport from a database of the airport, and calculating airport characteristic data for a period of time at intervals of one hour, wherein the airport characteristic data comprises an airport flight normal rate, a task overtime rate and the like; flight schedule data of flights involved in a period of time is extracted from a database of an airport, and the flight schedule data comprise scheduled departure time, scheduled arrival time, machine type, airline operators and the like as flight characteristic data. The three types of characteristics are combined in a time-sharing mode to generate input characteristic tensors at a plurality of moments:
I(t)=(x 1 (t),x 2 (t),x 3 (t),x 4 (t),x 5 (t)…,x m (t))
wherein t is the time number.
Extracting original data, extracting historical service node data of flights involved in a period of time from a database of an airport, and extracting only the flight service node data involved in a given guarantee flow. Then converting the time format, converting the absolute time (Beijing time) of the original data into the relative time (relative to a reference point), wherein the selection of the reference point is based on the classification of the flight service nodes, the arrival-to-arrival node is based on the planned arrival time (SI BT) as the reference point, the departure-to-arrival node is based on the planned departure time (SOBT), and the difference of the absolute time interval reference time is calculated, so that the service node relative time of a plurality of flights is obtained:
S(j)=(t 1 (j),t 2 (j),…,t n (j))
where j is the flight number.
And (3) making a difference between the actual time and the ideal time predicted by the first prediction model (20:00 of the day before time t is selected), so as to obtain a time difference sequence:
Δ=(δ 1 ,δ 2 ,…,δ n )
before using time difference data as input, the data needs to be preprocessed, firstly, a unit matrix with dimension of n multiplied by n, a second weather feature, a second airport feature and a transposed time difference sequence are combined, the generated matrix X (j) is the input data of a machine learning model, then the transposed time difference sequence is shifted one bit later to obtain the output Y (j, n) of the machine learning model, and a data set D is obtained by combining B (j, n) the data of the plurality of flights are longitudinally spliced to obtain a data set which is finally used for training:
D B (j,n)=[X(j)Y(j,n)]
and iteratively updating parameters of the initial second model based on the loss function until the test precision of the initial second prediction model reaches a preset precision threshold or the iteration number is larger than a preset number threshold. For example, the resulting ratio of node time errors within 5mi n exceeds 60%, or the number of iterations exceeds 100000, etc.
And taking the initial second prediction model with the completed iteration as a second prediction model.
In some embodiments, the second weather feature, the second airport feature, and the time difference are input into a second predictive model, the second predictive model outputting predicted times of remaining flight service nodes, comprising:
inputting a second weather feature, a second airport feature and a sequence delta of time differences into a second predictive model to obtain a complete time difference sequence delta':
wherein,,representing a predicted time difference, the predicted time difference being a predicted time difference between a time at which remaining flight service nodes are completed and the ideal time, G (delta) representing a cyclic call of a multiple output model or a single output model;
in some embodiments, inputting the second weather feature, the second airport feature, and the sequence of time differences Δ into a second predictive model, resulting in a complete sequence of time differences Δ', comprising:
predicting to obtain a first predicted time difference based on the second weather feature, the second airport feature and the time difference; the first predicted time difference is the time difference between the completion time of the first node of the predicted remaining flight service nodes and the corresponding ideal time;
predicting to obtain a second predicted time difference based on the second weather feature, the second airport feature, the time difference and the first predicted time difference; the second predicted time difference is a time difference between the predicted completion time of the second node of the remaining flight service nodes and the corresponding ideal time.
Repeating the steps based on the second weather feature, the second airport feature, the time difference and the predicted time difference to obtain the predicted time difference of the rest flight service nodes;
based on the time difference and the predicted time difference, a complete time difference sequence is obtained
Summing the complete time difference sequence delta' with the ideal time sequence to obtain a predicted time sequence S p :
And (3) monitoring the flight dynamics in real time, and when the progress of the flight service node is updated, or the real-time weather characteristics, the airport operation characteristics and the flight characteristic variation reach a certain threshold (for example, the task overtime rate is improved by 5%), or the time interval reaches a certain threshold (for example, one hour), acquiring data again, repeating the steps 130 and 140, outputting and updating the predicted time of the flight service node, and repeating the processes until the flight leaves.
Example 1
In practical application, weather data, airport operation data, flight information data and flight service node progress data can be obtained from an airport bus in real time through an interface protocol; then, by the same method used in the training process, the data are converted into the characteristic data used by the model and the time of the flight service node is converted into the relative time almost simultaneously, and the relative time is placed in a storage device for the program to take at any time.
The initial prediction may be triggered at a fixed time of the previous day (e.g., 20: 00), the rolling prediction may be triggered when the flight service node progress is updated, or the real-time weather characteristics, airport operational characteristics, and flight characteristic variation reach a certain threshold (e.g., 5% increase in task timeout rate), or the time interval reaches a certain threshold (e.g., one hour).
When a certain flight triggers the initial prediction or rolling prediction of service node time, the weather feature, the flight service feature and the flight information feature data are processed by using a first prediction model, and the ideal time sequence of all the ground guarantee nodes of the flight is output. If the current round of prediction is rolling prediction, calculating a time difference sequence between an ideal time sequence and the actual time of the flight service node, then using a second prediction model to process the time difference sequence, outputting a full time difference sequence after filling, and then summing the full time difference sequence and the ideal time sequence to obtain the predicted time of the current round of rolling prediction. And outputting the final result to the system.
The rolling prediction mechanism of a certain flight is normally operated in the life cycle of the certain flight, and the rolling prediction mechanism is used for system call at any time until the flight comes out of the port.
Fig. 2 is an exemplary block diagram of a rolling prediction system for flight ground assurance node time according to some embodiments of the present invention. As shown in fig. 2, the system 200 includes a first acquisition module 210, an ideal time determination module 220, a time difference determination module 230, a predicted time determination module 240, and a loop module 250.
The first acquisition module 210 is configured to acquire a first air characteristic, a first airport operating characteristic, and a flight characteristic.
The ideal time determining module 220 is configured to input the first air characteristic, the first airport operation characteristic, and the flight characteristic into a first prediction model, where the first prediction model outputs an ideal time of each flight service node; the ideal time refers to the predicted time to complete the flight service node.
The time difference determining module 230 is configured to obtain an actual time of the completed flight service node, and determine a time difference between the actual time and an ideal time; the actual time refers to the time at which the flight service node is completed.
The predicted time determining module 240 is configured to input the second weather feature, the second airport feature, and the time difference into a second prediction model, where the second prediction model outputs predicted times of remaining flight service nodes; the predicted time refers to the predicted time for completing the rest flight service nodes; the remaining flight service nodes refer to incomplete flight service nodes.
The circulation module 250 is configured to repeatedly determine the predicted time of the remaining flight service nodes until the flight exits.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The rolling prediction method for the flight ground guarantee node time is characterized by comprising the following steps of:
acquiring a first air characteristic, a first airport operation characteristic and a flight characteristic;
inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, wherein the first prediction model outputs ideal time of each flight service node;
the ideal time refers to the predicted time for completing the flight service node;
acquiring the completed actual time of the flight service node, and determining the time difference between the actual time and the ideal time; the actual time refers to the time of completing the flight service node;
inputting a second weather feature, a second airport feature and the time difference into a second prediction model, wherein the second prediction model outputs the predicted time of the rest flight service nodes; the predicted time refers to the predicted time for completing the rest flight service nodes; the remaining flight service nodes refer to incomplete flight service nodes;
and repeatedly determining the predicted time of the rest flight service nodes until the flights leave the port.
2. The method of roll prediction of flight ground assurance node time of claim 1, wherein the first and second weather features include rainfall, cloud cover high, wind speed, wind direction, barometric pressure, and weather type.
3. The method of roll prediction of flight ground assurance node time of claim 1, wherein the first airport operational characteristic and the second airport operational characteristic comprise a flight normal rate and a mission timeout rate.
4. The rolling forecast method of flight ground assurance node time of claim 1, wherein the flight characteristics include planned departure time, model and airline.
5. The rolling prediction method for flight ground assurance node time according to claim 1, wherein the inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, the first prediction model outputting ideal times for each flight service node comprises:
processing the first air characteristic, the first airport operation characteristic and the flight characteristic according to a tensor generation rule to generate a one-dimensional tensor I:
I=(x 1 ,x 2 ,x 3 ,x 4 ,x 5 …,x m )
wherein x is 1 ,x 2 ,x 3 ,x 4 ,x 5 …,x m Representing characteristic parameters, m representing the total number of characteristic parameters; the characteristic parameter is related to the first air characteristic, the first airport operational characteristic, or the flight characteristic;
calculating the ideal time S by taking the one-dimensional tensor as an input of the first prediction model i :
6. The rolling prediction method for flight ground assurance node time according to claim 1, wherein the acquiring the actual time of the completed flight service node, determining the time difference between the actual time and the ideal time, comprises:
generating an actual time sequence S based on the actual time r :
S r =(t 1 ,t 2 ,…,t k )
Wherein t is 1 ,t 2 ,…,t k Representing the actual time of each completed flight service node, k representing the total number of completed flight service nodes, 1<k<n; n represents the total number of flight service nodes;
based on the actual time and the ideal time, a sequence Δ of the time differences is obtained:
Δ=S r -S i =(δ 1 ,δ 2 ,…,δ k ,0,…,0)
wherein S is i Representing the sequence of ideal times, delta 1 ,δ 2 ,…,δ k Indicating the time difference between the completed flight service node and the corresponding ideal time, "0" indicates zero padding terms, for a total of n-k.
7. The rolling prediction method for flight ground assurance node time of claim 6, wherein the inputting of the second weather characteristic, the second airport characteristic and the time difference into a second prediction model, the second prediction model outputting predicted times of remaining flight service nodes comprises:
inputting the second weather feature, the second airport feature and the sequence of time differences delta into the second predictive model to obtain a complete sequence of time differences delta ′ :
Wherein,,representing a predicted time difference, the predicted time difference being a predicted time difference between a time at which remaining flight service nodes are completed and the ideal time, G (delta) representing a cyclic call of a multiple output model or a single output model;
the complete time difference sequence delta is processed ′ And the ideal is as describedSumming the sequences of times to obtain a sequence S of said predicted times p :
8. The rolling prediction method for flight ground assurance node time of claim 7, wherein the inputting the second weather feature, the second airport feature and the sequence of time differences delta into the second prediction model results in a complete sequence of time differences delta ′ Comprising:
predicting a first predicted time difference based on the second weather feature, the second airport feature, and the time difference; the first predicted time difference is the time difference between the completion time of a first node of the predicted remaining flight service nodes and the corresponding ideal time;
predicting a second predicted time difference based on the second weather feature, the second airport feature, the time difference, and the first predicted time difference; the second predicted time difference is the time difference between the completion time of the second node of the predicted remaining flight service nodes and the corresponding ideal time;
repeatedly obtaining the predicted time difference of the rest flight service nodes based on the second weather feature, the second airport feature, the time difference and the predicted time difference;
and obtaining the complete time difference sequence based on the time difference and the predicted time difference.
9. The rolling prediction method for flight ground assurance node time of claim 1, wherein the second prediction model is a machine learning model, the second prediction model being obtainable by training an initial second prediction model, comprising:
inputting a training sample into the initial second predictive model;
constructing a loss function based on the output of the initial second prediction model and the label of the training sample; the training sample is obtained by summing a unit matrix with dimension of n multiplied by n, a second weather feature, a second airport feature and a transposed sequence of historical time differences, and the label is obtained by shifting the transposed sequence of historical time differences one bit later;
iteratively updating parameters of the initial second model based on the loss function until the test precision of the initial second prediction model reaches a preset precision threshold or the iteration number is larger than a preset number threshold;
and taking the initial second prediction model with completed iteration as the second prediction model.
10. The rolling prediction system for the flight ground guarantee node time is characterized by comprising a first acquisition module, an ideal time determination module, a time difference determination module, a prediction time determination module and a circulation module;
the first acquisition module is used for acquiring first air characteristics, first airport operation characteristics and flight characteristics;
the ideal time determining module is used for inputting the first air characteristic, the first airport operation characteristic and the flight characteristic into a first prediction model, and the first prediction model outputs ideal time of each flight service node; the ideal time refers to the predicted time for completing the flight service node;
the time difference determining module is used for obtaining the completed actual time of the flight service node and determining the time difference between the actual time and the ideal time; the actual time refers to the time of completing the flight service node;
the prediction time determining module is used for inputting a second weather feature, a second airport feature and the time difference into a second prediction model, and the second prediction model outputs the prediction time of the rest flight service nodes; the predicted time refers to the predicted time for completing the rest flight service nodes; the remaining flight service nodes refer to incomplete flight service nodes; the circulation module is used for repeatedly determining the predicted time of the rest flight service nodes until the flights come out.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211625138.1A CN116432806A (en) | 2022-12-16 | 2022-12-16 | Rolling prediction method and system for flight ground guarantee node time |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211625138.1A CN116432806A (en) | 2022-12-16 | 2022-12-16 | Rolling prediction method and system for flight ground guarantee node time |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116432806A true CN116432806A (en) | 2023-07-14 |
Family
ID=87084317
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211625138.1A Pending CN116432806A (en) | 2022-12-16 | 2022-12-16 | Rolling prediction method and system for flight ground guarantee node time |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116432806A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117422320A (en) * | 2023-12-19 | 2024-01-19 | 长安大学 | Method for extracting influence factors of weather on flight toughness operation |
CN118095574A (en) * | 2024-04-23 | 2024-05-28 | 中国民航大学 | Flight arrival time prediction method, medium and device based on flight chain |
-
2022
- 2022-12-16 CN CN202211625138.1A patent/CN116432806A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117422320A (en) * | 2023-12-19 | 2024-01-19 | 长安大学 | Method for extracting influence factors of weather on flight toughness operation |
CN117422320B (en) * | 2023-12-19 | 2024-03-05 | 长安大学 | Method for extracting influence factors of weather on flight toughness operation |
CN118095574A (en) * | 2024-04-23 | 2024-05-28 | 中国民航大学 | Flight arrival time prediction method, medium and device based on flight chain |
CN118095574B (en) * | 2024-04-23 | 2024-07-12 | 中国民航大学 | Flight arrival time prediction method, medium and device based on flight chain |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116432806A (en) | Rolling prediction method and system for flight ground guarantee node time | |
CN111652427B (en) | Flight arrival time prediction method and system based on data mining analysis | |
Chen et al. | Uncertainty in urban mobility: Predicting waiting times for shared bicycles and parking lots | |
CN110119845A (en) | A kind of application method of track traffic for passenger flow prediction | |
CN113222387B (en) | Multi-objective scheduling and collaborative optimization method for hydrogen fuel vehicle | |
CN107704971A (en) | A kind of data processing method and device of real-time estimate airport security number | |
CN111178628B (en) | Luggage arrival time prediction method and device | |
CN112288197B (en) | Intelligent scheduling method and device for station vehicles | |
CN111507762A (en) | Urban taxi demand prediction method based on multi-task co-prediction neural network | |
CN112132366A (en) | Prediction system for flight clearance rate | |
Dalmau et al. | A machine learning approach to predict the evolution of air traffic flow management delay | |
CN112257914B (en) | Aviation safety causal prediction method based on random forest | |
CN117995027A (en) | Flight plan drop-preparation recommendation method, system, equipment and medium | |
WO2021172249A1 (en) | Wear state prediction method, wear state prediction device, wear state prediction program, and prediction model generation method | |
Natarajan et al. | A novel approach: Airline delay prediction using machine learning | |
CN114611753A (en) | Flight airplane trajectory data prediction system and method | |
EP4113095A1 (en) | Wear state prediction method, wear state prediction device and wear state prediction program | |
SAHADEVAN et al. | Predictability improvement of scheduled flights departure time variation using supervised machine learning | |
CN113807704A (en) | Intelligent algorithm platform construction method for urban rail transit data | |
CN112905659A (en) | Urban rail transit data analysis method based on BIM and artificial intelligence | |
EP4290327A1 (en) | Component maintenance prediction system with behavior modeling | |
Dursun | Air-traffic flow prediction with deep learning: a case study for Diyarbakır airport | |
CN117494879B (en) | Aircraft taxiing time prediction method, device, equipment and readable storage medium | |
Sharma et al. | Error calculation of flight delay prediction using various machine learning approaches | |
CN117252402B (en) | Airport value machine counter airline allocation planning method, device, equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |