CN116910664B - Cascade model-based flight ground guarantee dynamic prediction method - Google Patents
Cascade model-based flight ground guarantee dynamic prediction method Download PDFInfo
- Publication number
- CN116910664B CN116910664B CN202310852592.9A CN202310852592A CN116910664B CN 116910664 B CN116910664 B CN 116910664B CN 202310852592 A CN202310852592 A CN 202310852592A CN 116910664 B CN116910664 B CN 116910664B
- Authority
- CN
- China
- Prior art keywords
- prediction
- guarantee
- level
- model
- ground
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 230000008569 process Effects 0.000 claims abstract description 26
- 238000012546 transfer Methods 0.000 claims abstract description 22
- 230000005540 biological transmission Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 5
- 239000010410 layer Substances 0.000 claims description 3
- 239000002356 single layer Substances 0.000 claims description 3
- 238000010276 construction Methods 0.000 description 12
- 230000010006 flight Effects 0.000 description 6
- 238000007726 management method Methods 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000004140 cleaning Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 235000012054 meals Nutrition 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 244000309464 bull Species 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000035784 germination Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 210000003899 penis Anatomy 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a cascade model-based flight ground guarantee dynamic prediction method, wherein a cascade structure is used as a model frame in a prediction model, a multi-output gradient lifting regression tree is used as a basic learner, and input features comprise basic attribute features and hierarchical information transfer features. And judging a ground guarantee key route, determining a prediction level on the basis, and constructing a cascade multi-output gradient lifting regression tree prediction model and a prediction feature set. And (3) acquiring the initial prediction result and the hierarchy information transfer characteristic in the next hierarchy prediction characteristic set at the beginning of the guarantee, wherein each hierarchy prediction takes the newly acquired hierarchy information transfer characteristic and the basic attribute characteristic as input, and takes the latest prediction result and the hierarchy information transfer characteristic of the next hierarchy as output until the guarantee is finished along with the advance of the guarantee process. The invention realizes the dynamic prediction of the ground guarantee multi-node completion time, builds a comprehensive feature set and effectively improves the ground guarantee prediction precision.
Description
Technical Field
The invention belongs to the field of airport ground guarantee optimization, and particularly relates to a multi-node dynamic prediction method for flight ground guarantee based on a cascade multi-output gradient lifting regression tree.
Background
The current airport is mainly based on an A-CDM (Airport Collaborative Decision Making) system for the whole flow management of the ground operation of the flight, the A-CDM system is implemented by tens of millions of airports and part of medium and small airports in the country by the end of 2022, the construction of the system breaks through the information barriers among the operation main bodies such as the air traffic control, the airports, the airlines and the like, but the system has not been substantially developed in the aspects of refinement and intellectualization of the flight management, and the system is particularly obvious in the ground guarantee stage.
The civil aviation bureau technical specification of the airport collaborative decision system gives 45 milestones from landing to take off of flights, 37 of which are concentrated in the ground guarantee stage. In the airports with the built A-CDM system, most airports can only monitor the result of the 'wheel shift removal' of the final node of ground guarantee, belong to the category of post management, and cannot carry out prepositive prompt on the situation deviating from the normal guarantee completion time; although a few airports realize the active prediction of 'wheel withdrawal gear', the static prediction of a single node cannot guide a guarantee department to perform effective in-process intervention, and the prediction precision is difficult to meet the requirement, the essential reasons are that the features of ground guarantee delay germination period are relatively fuzzy, but along with the promotion of a guarantee process, the fluctuation of upstream operation can be accumulated at the downstream to cause 'bull penis effect' of the guarantee delay, and flights cannot finally finish the guarantee flow according to the set time target of an A-CDM system, thereby influencing the subsequent departure flow, and leading to the waste of take-off time slots and the reduction of the integral operation efficiency of the airports.
In order to promote the transition of the flight guarantee from post management to pre-prediction and in-process intervention, the transition of the ground guarantee prediction from static to dynamic and from a single node to the whole process node is required to be promoted, and the completion time of each key node of the flight in the ground guarantee process is predicted dynamically through data driving, so that the method has important significance for comprehensively improving the predictability and the fine management level of the ground guarantee process and improving the management efficiency of an A-CDM system.
The existing research on ground assurance prediction focuses on target gear withdrawal time prediction, and the proposed prediction method comprises distribution fitting, simulation modeling, network operation, heuristic algorithm and the like, so that the prerequisite influence of the ground assurance process on the final completion moment is ignored, the precision of a prediction model is difficult to be greatly improved, and the intervention on the ground assurance before delay occurs is not facilitated. In the aspect of influence factor research, the conventionally considered guarantee time prediction influence factors generally comprise factors such as machine type, machine position, operation period, airlines, route properties and the like, and intrinsic factors determined by airport flight schedules are less considered.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a cascade model-based flight ground guarantee dynamic prediction method.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
The method for dynamically predicting the ground guarantee of the flight based on the cascade model is characterized by comprising the following steps of:
Judging a key route of ground guarantee, and constructing a cascade multi-output gradient lifting regression tree prediction model according to a plurality of key operation nodes with sequential operation sequences on the key route;
Constructing basic attribute characteristics and hierarchical information transfer characteristics according to actual ground guarantee data, wherein the basic attribute characteristics correspond to original input characteristics of a cascade multi-output gradient lifting regression tree prediction model, and the hierarchical information transfer characteristics are used for information transfer between different layers of the cascade multi-output gradient lifting regression tree prediction model;
Training a cascade multi-output gradient lifting regression tree prediction model according to the basic attribute characteristics, the hierarchical information transfer characteristics and the completion time of the actual key operation nodes;
and predicting the completion time of each key operation node in the ground guarantee process according to the trained cascade multi-output gradient lifting regression tree prediction model.
In order to optimize the technical scheme, the specific measures adopted further comprise:
Further, before the cascade multi-output gradient lifting regression tree prediction model is constructed, a cascade frame is constructed, a prediction level of the cascade frame corresponds to key operation nodes on a key route, and each key operation node predicts the completion time of the unfinished key operation; the ground guarantee process has n+1 key operation nodes with sequential operation sequences, the finishing time of the previous n key operations is a dynamic prediction node, and the n-level linked frame is:
The target label of the I-th level prediction object is marked as Z i, the predicted value of the completion time of all unfinished key jobs at the I-th prediction node is shown as f i, the I-th level learner is marked as thetai ,, the original input characteristic is marked as X, the last level transmission information is marked as I i-1, and Z i＝fi({X,Ii-1};θi) is equal to or more than 1 and equal to or less than n.
Further, in the cascade multi-output gradient lifting regression tree prediction model, in each level of the constructed cascade framework, a multi-output gradient lifting regression tree is used as a single-layer learner, and the cascade multi-output gradient lifting regression tree prediction model is defined as follows:
In the case of n+1 key jobs, the ith hierarchy is denoted by T i, i key jobs have been completed, and also (n-i+1) key jobs have been completed, and the predicted value of the key jobs has been recorded as The original input feature is marked as X= (X 1,x2,…,xm), and the i-th-level multi-output gradient lifting regression tree model is marked as/>The parameters are denoted as/>The last level of transfer information is denoted as I i-1, then/>
Further, the basic attribute features are obtained according to actual ground assurance data fields, including several items of: machine type, whether towed, inbound, outbound, inbound type, outbound type, inbound terminal, outbound terminal, inbound passenger number, outbound passenger number, ground agent, inbound security period, outbound security period, inbound route properties, outbound route properties, minimum outbound time, planned outbound time, and inbound flight density.
Further, the hierarchical information transfer characteristic comprises a preamble guarantee prediction time and a preamble guarantee prediction deviation, and the preamble guarantee prediction time is obtained dynamically according to a guarantee process, wherein the preamble guarantee prediction time is a predicted value obtained at a T i -1 hierarchy at the completion time of the ith key operation, and the preamble guarantee prediction deviation is the difference between an actual value obtained at the beginning of the T i hierarchy at the completion time of the ith key operation and a predicted value obtained at the T i-1 hierarchy.
Further, the completion time of each key operation node in the ground guarantee process is predicted according to the trained cascade multi-output gradient lifting regression tree prediction model, and the method specifically comprises the following steps:
Firstly, initial prediction is carried out at a T 1 level, the time of the preface operation is taken as a level information transmission characteristic I 0, and an input_1 of the T 1 level is formed by the preface operation time and X, so that predicted values output_1 of all unfinished operations under the current guarantee progress and a level information transmission characteristic I 1 of the next level are obtained;
Predicting each level after the level T 1 in sequence until the level T n is predicted, wherein for the prediction of the level T i, the input_i of the level T i is a set of X and I i-1, the output_i is the latest prediction result, and the level information transmission characteristic I i of the next level is calculated;
and finally, obtaining dynamic prediction results of all levels, wherein the dynamic prediction results correspond to the completion time of each key operation node in the ground guarantee process.
In another aspect, the present invention also provides a computer readable storage medium storing a computer program, where the computer program causes a computer to execute the cascade model-based flight ground assurance dynamic prediction method as described above.
In another aspect, the present invention further provides an electronic device, including: the system comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the cascade model-based flight ground guarantee dynamic prediction method when executing the computer program.
The beneficial effects of the invention are as follows:
1) The invention provides a multi-main-body cooperative multi-operation serial-parallel complex process, which is characterized in that multi-operation prediction is guaranteed through a multi-output model, a multi-level model is constructed by combining a cascading frame, the effect of dynamic prediction is achieved through transmitting prediction information among level models, the limitation that only a single node of a wheel shield is removed for static prediction in the existing research is broken through, and the dynamic prediction of the completion time of multiple nodes of the ground guarantee is realized;
2) The multi-output model provided by the invention greatly reduces the complexity and time cost of model training, so that the prediction is more efficient, the gradient lifting regression tree can flexibly process continuous or discrete type data and fully utilize all characteristic information of the data, has better robustness on abnormal values, and can obtain higher prediction precision under relatively less operation time;
3) The invention constructs the ground guarantee dynamic prediction feature set comprising the basic attribute features and the hierarchical information transmission features, wherein the internal features determined by the airport flight schedule are considered, the feature construction is more comprehensive, and the ground guarantee prediction precision is effectively improved.
Drawings
FIG. 1 is a flow chart of a flight ground assurance operation in an embodiment.
FIG. 2 is a mapping relation diagram of a cascade frame and a ground-guaranteed multi-node dynamic prediction process in an embodiment.
FIG. 3 is a block diagram of a hierarchical multiple output gradient-lifting regression tree algorithm in an embodiment.
FIG. 4 is a schematic diagram of a hierarchical information transfer feature in an embodiment.
FIG. 5 is a graph of dynamic prediction results for all nodes in an embodiment.
FIG. 6 is a graph showing the variation trend of MAE and.+ -. 5min prediction accuracy in the examples.
Fig. 7 is a flowchart of a method for dynamically predicting a ground guarantee of a flight based on a cascading model in an embodiment.
Detailed Description
The invention will now be described in further detail with reference to the accompanying drawings.
The technical scheme provided by the invention comprises three parts of model construction, feature construction and prediction method design.
1. Model construction
The model building part specifically comprises cascade framework building and cascade multi-output gradient lifting regression tree model building.
1. Cascading frame construction
The flight ground guarantee comprises a plurality of operations with serial-parallel relation, strict operation sequence requirements, such as opening and closing cabin doors, loading and unloading passengers, meal allocation, cleaning, oiling and the like, the flight ground guarantee operation flow is shown in a figure 1, the ground guarantee multi-node dynamic prediction is realized, the completion time of the operation is required to be predicted in different stages of the guarantee, the early or late completion of the front operation on a key route influences the subsequent operation, and the prediction and the completion condition of the operation in the last stage are required to be considered for the multi-node prediction in different stages. Assuming that n+1 key operation nodes with sequential operation sequences are shared from the upper gear to the lower gear, taking the completion time of n key operations except the lower gear as a node for dynamic prediction, n levels are represented, so that an n-level linked frame n-cascaded corresponding to a ground-guaranteed multi-node dynamic prediction process is defined as follows:
The target label of the I-level prediction object is denoted as Zi , to represent the predicted value of the completion time of all outstanding key jobs at the I-level prediction node, the I-level learner is denoted as f i, the parameter is denoted as theta i, the original input characteristic is denoted as X, the last-level transmission information is denoted as I i-1, and Z i＝fi({X,Ii-1};θi), and 1 is equal to or more than I is equal to or less than n.
The mapping relation between the cascade frame and the ground-guaranteed multi-node dynamic prediction process is shown in fig. 2.
2. Cascaded multi-output gradient lifting regression tree model construction
After the cascade framework is built, a learner for multi-node prediction of each level is further built. By means of MultiOutputRegressor functions in python, a plurality of independent single-output gradient lifting regression tree (Gradient Boosting Regression Tree, GBRT) models are integrated and packaged, a multi-output gradient lifting regression tree model is formed to serve as a single-layer learner of a cascade framework, and finally a cascade multi-output gradient lifting regression tree model (CascadedMultiple Output-GradientBoosting Regression Tree, CMO-GBRT) is built, and is defined as follows:
The number of predicted objects of each level under the cascade frame is equal to the total number of key jobs minus the number of corresponding levels, in the case of n+1 key jobs, the ith level is represented by T i, when the i key jobs are completed, the (n-i+1) outstanding jobs are shared from the (i+1) th job to the (n+1) th shift removal job which are the forefront in sequence, and the predicted values of the i key jobs are recorded as Assuming m original input features, denoted as x= (X 1,x2,…,xm), the i-th level multiple output GBRT is denoted as/>The parameters are denoted as/>The information transmitted by the previous layer is recorded as I i-1
A framework of a cascaded multiple-output gradient-lifting regression tree algorithm is shown in fig. 3.
2. Feature construction
The feature construction part specifically comprises two types of features, namely basic attribute features and hierarchical information transfer features.
1. Basic attribute features
The basic attribute characteristics are obtained according to the actual ground guarantee data field and correspond to the original input characteristics, and specifically comprise a machine type, whether dragging, a port entering machine position, a port leaving machine position, a port entering machine position type, a port entering machine station building, a port leaving machine station building, a port entering passenger number, a port leaving passenger number, a ground agency, a port entering guarantee period, a port leaving guarantee period, port entering route properties, port leaving route properties, minimum station passing time, planned station passing time and arrival flight density. Wherein the planned transit time is calculated as formula (1):
Tx＝tSTD-tSTA (1)
Wherein: t S is the planned transit time, T STD is the planned departure time, and T STA is the planned arrival time.
The flight density is defined as the number of flights falling within the time window adjacent to the front and rear of the actual falling of the current guaranteed flight, and the following formula (2) is calculated:
Wherein: n d is the density of arriving flights, N b,Na respectively represents the number of landed flights in adjacent time windows before and after the actual landing of the current guaranteed flight, t represents the time window, and generally 15min is taken.
2. Hierarchical information delivery features
The construction level information transfer characteristic is used for transferring information between different levels of the cascade model, and is obtained through calculation according to actual ground guarantee data and a preamble prediction result, and specifically comprises preamble guarantee prediction time and preamble guarantee prediction deviation.
When the prediction is performed through the CMO-GBRT, except for the initial prediction before the start of the guarantee, a new level of predicted value is updated for the unfinished operation when each subsequent operation is completed, at this time, the completed operation has a previous level of predicted time and a new obtained actual time, the predicted value of the previous level is used as a preamble guarantee predicted time feature, and the deviation of the predicted value and the actual value is used as a preamble guarantee predicted deviation feature. Two types of hierarchical information transfer features are calculated as follows:
Wherein: l pre is the preamble guarantee prediction time characteristic, A predicted value of the completion time of the ith operation in the T i-1 level; l bias is the preface guarantee prediction deviation feature,/>The actual value of the i-th job completion time obtained at the beginning of T i.
3. Prediction method design
After model construction and feature construction are completed, ground guarantee multi-node dynamic prediction can be performed, and the prediction method specifically comprises the following steps:
S1: acquiring airport ground guarantee actual operation data, constructing a basic attribute feature set X, judging a ground guarantee key route, and constructing a cascade multi-output gradient lifting regression tree prediction model of n prediction levels;
S2: because the T 1 level is the 1 st level, the preface operation time is taken as the level information transmission characteristic I 0, a characteristic set input_1 of the T 1 level is formed by the preface operation time and x, a training optimal T 1 level CMO-GBRT _1 model is input, and the predicted values ouput _1 of all incomplete operations and the level information transmission characteristic I 1 of the next level under the current guarantee progress are obtained;
S3: the process of S2 is repeated until the T n -level prediction is completed. The CMO-GBRT _i model of the T i level inputs input_i into a set of x and I i-1, outputs ouput _i into the latest prediction result, and calculates the level information transmission characteristic of the next level to be I i. The hierarchical information transfer features are shown in fig. 4. And finally obtaining all dynamic prediction results result.
In an embodiment, the invention specifically provides a cascade model-based flight ground guarantee dynamic prediction method, as shown in fig. 7, which specifically comprises the following steps.
Step 1: judging a key route, and constructing a cascade multi-output gradient lifting regression tree prediction model.
Before the ground guarantee multi-node dynamic prediction is carried out, a guarantee key route needs to be judged in advance to design a prediction level. The ground guarantee can be closely related to the passenger guarantee flow or not on time, the processed data and first-line guarantee experience are combined, 7 operations related to the passenger guarantee, namely an upper gear, a passenger opening, a passenger closing end, a passenger opening, a passenger opening closing end and a passenger closing end, are selected as key operations, the operation flow formed by serially connecting the ground guarantee in sequence is taken as a key route, the 6 subsequent nodes are dynamically predicted from the upper gear node, so that a CMO-GBRT model with 6 levels is constructed, and the level numbers and the operation nodes corresponding to the prediction moments are shown in table 1.
Table 1 correspondence between hierarchy numbers and job nodes
Step 2: and constructing basic attribute characteristics and hierarchical information transfer characteristics by combining actual operation data.
The ground guarantee operation data of partial passenger transport flights of the Pudong airport 2022 are selected, and the data fields comprise flight basic information and main guarantee operation time data, as shown in table 2.
TABLE 2 flight ground assurance operation data
| Fields | Sample example | Fields | Sample example |
| Date of day | 2021-3-1 | Upper wheel block | 14:32 |
| Flight number | 9C6136|9C8569 | Door for opening passenger | 14:36 |
| Flight attributes | Work in front | Customer start/end | 14:38|14:43 |
| STA | 14:35 | Cleaning start/end | 14:51|15:02 |
| STD | 15:50 | Oil supply start/end | 15:24|15:37 |
| Model type | A320 | Meal start/end | 15:28|15:29 |
| Machine position | 55|61 | Boarding gate opening | 15:34 |
| Route properties | Domestic and international | Gate closing | 15:50 |
| Ground proxy | CQH | Closing door | 15:57 |
| Total number of passengers | 175|165 | Wheel removing block | 16:07 |
Because the time format data is not suitable for the regression prediction model, data conversion is required, and the time from when a certain job is completed is defined as the time period from when the ground assurance starts to when the job is completed, as shown in formula (5):
wherein: t task is the operation time, task is the operation name, For the actual completion time of the operation, t 0 is the guarantee actual start time, and the above gear time is replaced. Table 2 sample job completion times after conversion are shown in table 3.
TABLE 3 flight ground assurance operation completion time conversion results
| Work is carried out | Post conversion working time (min) | Work is carried out | Post conversion working time (min) |
| Upper wheel block | 0 | End of meal | 57 |
| Door for opening passenger | 4 | Boarding gate opening | 62 |
| End of getting off | 11 | Gate closing | 78 |
| End of cleaning | 30 | Closing door | 85 |
| End of oil supply | 65 | Wheel removing block | 95 |
According to the ground guarantee data of the Pudong airport, 18 basic attribute features and 2 level information transmission features are analyzed and obtained, and the final feature set is shown in table 4.
TABLE 4 final feature set
Step 3: based on the design of the prediction method, a model is trained and applied to the ground-guaranteed multi-node dynamic prediction.
Dividing the processed data set into a training set and a testing set according to the proportion of 7:3, and setting MAE, rw and prediction accuracy as model evaluation indexes, wherein the prediction accuracy indexes are the sample number proportion of the difference value of a prediction result and actual time within +/-3 min and +/-5 min.
The initial prediction is performed when the flight performs the up-shift operation, and at this time, the prediction result of the previous level is not available, and the level information transmission characteristic cannot be obtained, so that the time length of the forward operation entering the port is replaced by the time length of the forward operation entering the port, and the initial prediction result is shown in table 5. The T 1 -level prediction object is 6 nodes including a passenger opening door, a passenger discharging end, a passenger gate opening, a passenger gate closing door and a wheel removing stop, and MAE of the prediction results of all the nodes is lower than 3.2min, so that the error is small, and the requirements of airport collaborative decision-making technical specifications on the prediction accuracy of key operations are met; meanwhile, R 2 is higher and is larger than 0.7, which shows that the fitting degree of the model to sample data is better, and the effectiveness of the feature set is verified; in the aspect of prediction precision, the prediction precision of all nodes +/-5 min reaches more than 80%, the prediction precision of +/-3 min reaches more than 50%, wherein the prediction precision of +/-3 min of a starting node and a finishing node of a getting-off node reaches more than 85.2% and 67.32%, and the prediction precision of a T 1 level is good.
Table 5T 1 hierarchical prediction results
| Evaluation index | Door for opening passenger | End of getting off | Boarding gate opening | Gate closing | Closing door | Wheel removing block |
| MAE | 1.6 | 2.46 | 2.93 | 2.96 | 2.94 | 3.11 |
| R2 | 0.85 | 0.80 | 0.96 | 0.97 | 0.97 | 0.97 |
| ±3min | 85.20％ | 67.32％ | 56.36％ | 54.79％ | 55.44％ | 52.97％ |
| ±5min | 95.39％ | 87.98％ | 82.47％ | 82.37％ | 83.23％ | 80.04％ |
In order to verify the superiority of GBRT model, convolutional neural network CNN in the deep learning model commonly used for ground guarantee prediction is selected for comparison analysis, and the prediction result of the T 1 level gear removing node is taken as a comparison object, as shown in table 6, GBRT error is smaller, fitting degree is higher, prediction precision is better, so GBRT is selected as a reference model to be effective.
Table 6 GBRT compares the results of the withdrawal time prediction for CNN at T 1 level
| Evaluation index | CNN | GBRT |
| MAE | 5.77 | 3.11 |
| R2 | 0.87 | 0.97 |
| ±3min | 34.83％ | 52.97％ |
| ±5min | 53.64％ | 80.04％ |
With the advance of the guarantee process, the dynamic prediction results of all the nodes are shown in fig. 5, so that the prediction errors of all the nodes are gradually reduced, R2 is gradually increased, and the prediction accuracy is improved to different degrees. Taking two evaluation indexes of MAE and + -5 min prediction precision as an example, the change trend of the two evaluation indexes is shown in fig. 6, and besides the prediction performance of a boarding gate opening node does not change significantly, as the prediction level increases, the MAE of a boarding gate closing, a boarding gate closing and a wheel gate removing node presents a significant decrease trend, and + -5 min prediction precision presents a significant increase trend, so that a CMO-GBRT model can dynamically predict a plurality of nodes in the guarantee process, the prediction performance can be gradually improved along with the guarantee process, and the effectiveness of the model is verified. And because the prediction performance of the boarding gate closing, passenger gate closing and wheel stop removing nodes is obviously improved in the level T4, the boarding gate opening can be obtained as an optimal entry point for the intervention of a collaborative decision mechanism in the ground guarantee process and the avoidance of flight delay.
In another embodiment, the invention further provides a computer readable storage medium storing a computer program, wherein the computer program causes a computer to execute the cascade model-based flight ground guarantee dynamic prediction method according to the first embodiment.
In another embodiment, the present invention further provides an electronic device, including: the system comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to realize the cascade model-based flight ground guarantee dynamic prediction method according to the first embodiment.
In the disclosed embodiments, a computer storage medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a computer storage medium would include one or more wire-based electrical connections, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the concept of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to the invention without departing from the principles thereof are intended to be within the scope of the invention as set forth in the following claims.
Claims (6)
1. The method for dynamically predicting the ground guarantee of the flight based on the cascade model is characterized by comprising the following steps of:
Judging a key route of ground guarantee, and constructing a cascade multi-output gradient lifting regression tree prediction model according to a plurality of key operation nodes with sequential operation sequences on the key route;
Constructing basic attribute characteristics and hierarchical information transfer characteristics according to actual ground guarantee data, wherein the basic attribute characteristics correspond to original input characteristics of a cascade multi-output gradient lifting regression tree prediction model, and the hierarchical information transfer characteristics are used for information transfer between different layers of the cascade multi-output gradient lifting regression tree prediction model;
Training a cascade multi-output gradient lifting regression tree prediction model according to the basic attribute characteristics, the hierarchical information transfer characteristics and the completion time of the actual key operation nodes;
predicting the completion time of each key operation node in the ground guarantee process according to the trained cascade multi-output gradient lifting regression tree prediction model;
Before the cascade multi-output gradient lifting regression tree prediction model is built, a cascade frame is built, a prediction level of the cascade frame corresponds to key operation nodes on a key route, and each key operation node predicts the completion time of the unfinished key operation; the ground guarantee process has n+1 key operation nodes with sequential operation sequences, the finishing time of the previous n key operations is a dynamic prediction node, and the n-level linked frame is:
The target label of the I-th level prediction object is marked as Z i and represents the predicted value of the completion time of all unfinished key jobs at the I-th prediction node, the I-th level learner is marked as f i, the parameter is marked as theta i, the original input characteristic is marked as X, the last level transmission information is marked as I i-1, Z i＝fi({X,Ii-1};θi), and I is more than or equal to 1 and less than or equal to n;
in each level of the constructed cascade framework, the cascade multi-output gradient lifting regression tree prediction model uses a multi-output gradient lifting regression tree as a single-layer learner, and defines the cascade multi-output gradient lifting regression tree prediction model as follows:
In the case of n+1 key jobs, the ith hierarchy is denoted by T i, i key jobs have been completed, and also (n-i+1) key jobs have been completed, and the predicted value of the key jobs has been recorded as The original input feature is marked as X= (X 1,x2,…,xm), the i-th-level multi-output gradient lifting regression tree model is marked as f i GBRT, and the parameters are marked as/>The last level of transfer information is denoted as I i-1, then/>
2. The cascading model-based flight ground guarantee dynamic prediction method as claimed in claim 1, wherein: the basic attribute characteristics are obtained according to the actual ground guarantee data field and comprise a plurality of items as follows: machine type, whether towed, inbound, outbound, inbound type, outbound type, inbound terminal, outbound terminal, inbound passenger number, outbound passenger number, ground agent, inbound security period, outbound security period, inbound route properties, outbound route properties, minimum outbound time, planned outbound time, and inbound flight density.
3. The cascading model-based flight ground guarantee dynamic prediction method as claimed in claim 1, wherein: the hierarchical information transfer characteristic comprises a preamble guarantee prediction time and a preamble guarantee prediction deviation, and the preamble guarantee prediction time and the preamble guarantee prediction deviation are dynamically obtained according to a guarantee process, wherein the preamble guarantee prediction time is a predicted value obtained at a T i-1 level at the completion time of an ith key operation, and the preamble guarantee prediction deviation is the difference between an actual value obtained at the beginning of the T i level at the completion time of the ith key operation and a predicted value obtained at the T i-1 level.
4. The cascading model-based flight ground guarantee dynamic prediction method as claimed in claim 1, wherein: the completion time of each key operation node in the ground guarantee process is predicted according to the trained cascade multi-output gradient lifting regression tree prediction model, and the method specifically comprises the following steps:
Firstly, initial prediction is carried out at a T 1 level, the time of the preface operation is taken as a level information transmission characteristic I 0, and an input_1 of the T 1 level is formed by the preface operation time and X, so that predicted values output_1 of all unfinished operations under the current guarantee progress and a level information transmission characteristic I 1 of the next level are obtained;
Predicting each level after the level T 1 in sequence until the level T n is predicted, wherein for the prediction of the level T i, the input_i of the level T i is a set of X and I i-1, the output_i is the latest prediction result, and the level information transmission characteristic I i of the next level is calculated;
and finally, obtaining dynamic prediction results of all levels, wherein the dynamic prediction results correspond to the completion time of each key operation node in the ground guarantee process.
5. A computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute the cascade model-based flight ground assurance dynamic prediction method according to any one of claims 1 to 4.
6. An electronic device, comprising: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the cascading model-based flight ground-assurance dynamic prediction method of any one of claims 1-4 when the computer program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310852592.9A CN116910664B (en) | 2023-07-12 | 2023-07-12 | Cascade model-based flight ground guarantee dynamic prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310852592.9A CN116910664B (en) | 2023-07-12 | 2023-07-12 | Cascade model-based flight ground guarantee dynamic prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116910664A CN116910664A (en) | 2023-10-20 |
| CN116910664B true CN116910664B (en) | 2024-04-19 |
Family
ID=88366110
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310852592.9A Active CN116910664B (en) | 2023-07-12 | 2023-07-12 | Cascade model-based flight ground guarantee dynamic prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116910664B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106779222A (en) * | 2016-12-20 | 2017-05-31 | 中国人民解放军空军装备研究院雷达与电子对抗研究所 | Airport ground stand-by period Forecasting Methodology and device |
| CN106845856A (en) * | 2017-02-15 | 2017-06-13 | 民航成都信息技术有限公司 | Civil Aviation Airport ground service ensures real-time dynamic decision method |
| CN112580798A (en) * | 2020-12-03 | 2021-03-30 | 哈尔滨电站设备成套设计研究所有限公司 | Intelligent early warning method for equipment based on multi-input multi-output ResNet |
| CN112990437A (en) * | 2021-03-24 | 2021-06-18 | 厦门吉比特网络技术股份有限公司 | Reinforced learning neural network based on causal multiple outputs and construction method thereof |
| CN113095779A (en) * | 2021-04-09 | 2021-07-09 | 飞友科技有限公司 | Configurable flight guarantee time calculation method and system |
| CN113255207A (en) * | 2021-04-14 | 2021-08-13 | 杭州电子科技大学 | Iterative multi-output-Markov chain-based multi-step prediction method for argon fraction variable of air separation system |
| CN114254214A (en) * | 2021-12-30 | 2022-03-29 | 中山大学 | Traffic prediction method and system based on space-time hierarchical network |
| CN114611781A (en) * | 2022-03-07 | 2022-06-10 | 中国民航大学 | Flight plan-oriented outbound passenger gathering information prediction method and system |
| CN114781704A (en) * | 2022-04-08 | 2022-07-22 | 南京航空航天大学 | Flight delay prediction method based on station-passing flight guarantee process |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2018241119A1 (en) * | 2017-10-06 | 2019-05-02 | Tata Consultancy Services Limited | System and method for flight delay prediction |
| GB2584625A (en) * | 2019-05-28 | 2020-12-16 | Sita Information Networking Computing Uk Ltd | System and method for flight arrival time prediction |
| CA3109505A1 (en) * | 2020-02-18 | 2021-08-18 | Royal Bank Of Canada | System and method for weather dependent machine learning architecture |
-
2023
- 2023-07-12 CN CN202310852592.9A patent/CN116910664B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106779222A (en) * | 2016-12-20 | 2017-05-31 | 中国人民解放军空军装备研究院雷达与电子对抗研究所 | Airport ground stand-by period Forecasting Methodology and device |
| CN106845856A (en) * | 2017-02-15 | 2017-06-13 | 民航成都信息技术有限公司 | Civil Aviation Airport ground service ensures real-time dynamic decision method |
| CN112580798A (en) * | 2020-12-03 | 2021-03-30 | 哈尔滨电站设备成套设计研究所有限公司 | Intelligent early warning method for equipment based on multi-input multi-output ResNet |
| CN112990437A (en) * | 2021-03-24 | 2021-06-18 | 厦门吉比特网络技术股份有限公司 | Reinforced learning neural network based on causal multiple outputs and construction method thereof |
| CN113095779A (en) * | 2021-04-09 | 2021-07-09 | 飞友科技有限公司 | Configurable flight guarantee time calculation method and system |
| CN113255207A (en) * | 2021-04-14 | 2021-08-13 | 杭州电子科技大学 | Iterative multi-output-Markov chain-based multi-step prediction method for argon fraction variable of air separation system |
| CN114254214A (en) * | 2021-12-30 | 2022-03-29 | 中山大学 | Traffic prediction method and system based on space-time hierarchical network |
| CN114611781A (en) * | 2022-03-07 | 2022-06-10 | 中国民航大学 | Flight plan-oriented outbound passenger gathering information prediction method and system |
| CN114781704A (en) * | 2022-04-08 | 2022-07-22 | 南京航空航天大学 | Flight delay prediction method based on station-passing flight guarantee process |
Non-Patent Citations (3)
| Title |
|---|
| 基于XGBoost模型的炼油厂氢气网络动态多输出预测模型;王宁;华东理工大学学报（自然科学版）;77-83 * |
| 基于级联模型的航班撤轮挡时刻预测;丁杨;中国优秀硕士学位论文全文数据库工程科技Ⅱ辑;23-38 * |
| 枢纽机场航班保障服务时间估计;唐云霄;系统仿真学报;2856-2864+2874 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116910664A (en) | 2023-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110807525B (en) | Neural network flight guarantee service time estimation method based on improved genetic algorithm | |
| Bresina et al. | Activity Planning for the Mars Exploration Rovers. | |
| Camponogara et al. | Distributed optimization for model predictive control of linear dynamic networks with control-input and output constraints | |
| CN104504460A (en) | Method and device for predicating user loss of car calling platform | |
| CN112508398B (en) | Dynamic production scheduling method and device based on deep reinforcement learning and electronic equipment | |
| CN111582584B (en) | Air section operation time optimization method based on fairness of airlines | |
| CN109409811A (en) | A kind of Logistic Scheduling method, apparatus, computer equipment and storage medium | |
| Chenarani et al. | Simulating the impact of activity uncertainties and risk combinations in R&D projects. | |
| Andrade-Michel et al. | Vehicle and reliable driver scheduling for public bus transportation systems | |
| CN110942193A (en) | Vehicle scheduling method and storage medium | |
| CN111898867A (en) | Airplane final assembly production line productivity prediction method based on deep neural network | |
| CN115511212A (en) | Flight guarantee flow residual execution time prediction method based on deep learning | |
| CN116910664B (en) | Cascade model-based flight ground guarantee dynamic prediction method | |
| CN114330842A (en) | Airport ferry vehicle scheduling method based on two-stage optimization | |
| CN114819622A (en) | Carrier-based aircraft tripping operation scheduling method, device, equipment and medium | |
| Manafi et al. | A centroid opposition-based coral reefs algorithm for solving an automated guided vehicle routing problem with a recharging constraint | |
| Lv et al. | A variable neighborhood search algorithm for airport ferry vehicle scheduling problem | |
| CN112053033B (en) | Airport stand allocation method | |
| Zhang et al. | Applications of decision support system in aviation maintenance | |
| He et al. | A dynamic holding approach to stabilizing a bus line based on the q-learning algorithm with multistage look-ahead | |
| Grihon et al. | Surrogate modeling of buckling analysis in support of composite structure optimization | |
| CN110851961A (en) | Ship lock maintenance plan making method based on system availability | |
| US11694134B2 (en) | Optimization method and device of crew replanning for high-speed railway in typical scenarios | |
| Asadi et al. | Aircraft total turnaround time estimation using fuzzy critical path method | |
| CN113139344A (en) | Civil aircraft operation risk assessment method oriented to multiple failure modes |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |