CN110210648B - Gray long-short term memory network-based control airspace strategic flow prediction method - Google Patents
Gray long-short term memory network-based control airspace strategic flow prediction method Download PDFInfo
- Publication number
- CN110210648B CN110210648B CN201910360715.0A CN201910360715A CN110210648B CN 110210648 B CN110210648 B CN 110210648B CN 201910360715 A CN201910360715 A CN 201910360715A CN 110210648 B CN110210648 B CN 110210648B
- Authority
- CN
- China
- Prior art keywords
- main feature
- air traffic
- year
- value
- traffic flow
- 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 47
- 230000015654 memory Effects 0.000 title claims abstract description 23
- 238000012216 screening Methods 0.000 claims abstract description 20
- 230000007787 long-term memory Effects 0.000 claims abstract description 11
- 230000006403 short-term memory Effects 0.000 claims abstract description 11
- 238000010219 correlation analysis Methods 0.000 claims abstract description 10
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 239000013598 vector Substances 0.000 claims description 62
- 238000012549 training Methods 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 23
- 238000012360 testing method Methods 0.000 claims description 15
- 238000009825 accumulation Methods 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 230000000694 effects Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000005316 response function Methods 0.000 claims description 6
- 238000000513 principal component analysis Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000008676 import Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000007726 management method Methods 0.000 abstract description 10
- 238000005457 optimization Methods 0.000 abstract description 2
- 238000011161 development Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000012847 principal component analysis method Methods 0.000 description 2
- 101100134058 Caenorhabditis elegans nth-1 gene Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- 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
- G06N3/045—Combinations of 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
- G06N3/08—Learning methods
-
- 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
- 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)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a control airspace strategic flow prediction method based on a gray long-short term memory network, and belongs to the technical field of air traffic flow management. The invention comprises the following steps: step 1: reading data; step 2: preprocessing data; step 3: preliminary screening influence factors by gray correlation analysis; step 4: extracting main features by using a main component analysis method; step 5: establishing a grey strategic flow prediction model; step 6: establishing a strategy flow prediction model of the long-term and short-term memory network; step 7: and establishing a grey long-short-term memory network combined prediction model. The invention can provide scientific basis for the optimization of airspace structures such as sector planning, airway adjustment and the like of the control area, realize the effective utilization of airspace resources, and provide basis for the resource demand distribution such as future personnel investment, financial investment, fixed asset investment and the like of the control area.
Description
Technical Field
The invention relates to a control airspace strategic flow prediction method based on a gray long-short-term memory network, and belongs to the technical field of air traffic flow management.
Background
In recent years, the national aviation industry has developed rapidly, and the aviation transportation demands are gradually vigorous. At present, the air traffic flow management service in China is always in a lag and low-efficiency state, and the contradiction between the air traffic flow management service and the air traffic flow which is growing at a high speed is increasingly prominent. The unreasonable airspace planning, low degree of systemization and automation of flow management, low guarantee capability of an air traffic control system and the randomness of flow control lead to the fact that the existing airspace resources and management means are difficult to adapt to the rapid increase of air traffic flow, traffic congestion, flight conflict and other conditions occur in airports, terminal areas, air way intersections and the like, the bottleneck of an air traffic network is formed, the ground waiting before flight and the conditions of air waiting, flight change, yaw and the like in flight are directly caused, so that flight safety is influenced, flight fuel consumption is increased, and flight normality is reduced. Therefore, a set of scientific and effective air traffic flow management system needs to be established to scientifically manage the air traffic flow, and the premise and the basis of the flow management are that the distribution and the development trend of the air traffic flow need to be accurately counted and predicted.
The prediction of the number of aircrafts in a certain airspace and a certain time period is a part of the prediction of air traffic flow, is an important basis and decision basis for air traffic planning and management, and provides a basis for improving the operation efficiency of the whole country and region. According to statistics of national and regional distribution of air traffic flow in the earlier stage, areas and waypoints where congestion possibly occurs in a period of time in the future are predicted, so that control personnel can perform corresponding strategic deployment, a deployment scheme is made in advance, and effective control measures are adopted in the following flight operation stage, so that the problem of flight delay caused by air traffic congestion is greatly relieved, flight safety is guaranteed, the reasonable utilization rate of airspace resources is improved, and the operation efficiency and economic benefit of airlines are greatly improved. In addition, the flow prediction can also improve the rationality of the schedule of the flight time, eliminate a series of unreasonable and unsafe factors caused by traffic jam, and promote the reasonable distribution and effective monitoring of the flight time. The air traffic flow prediction has instructive effects on construction and planning of facilities such as airports, so that resources such as airspace and airlines are effectively and fully utilized, the allocation of aviation resources in China is promoted to better adapt to the continuously increased traffic and transportation demands in the future, powerful support is provided for effective and smooth operation of air traffic, and sustainable development of civil aviation transportation industry in China is facilitated.
Air traffic flow predictions can be classified into strategic predictions and tactical predictions. Strategic prediction focuses on analysis of medium-long term flow space-time variation trend, and has important guiding function on adjusting national airspace structure, relieving airspace congestion, controlling staff and equipment allocation and other problems. Tactical prediction considers the conditions of airspace structure, control rule, airway configuration and the like, and according to the flight plan of the current day and the data of airspace data, telegrams, radars, weather and the like updated in real time, the number of aircrafts in future time is statistically predicted for airspace units such as airway points, airway sections, sectors and the like, so that a controller can be gracefully corresponding to the air traffic condition, and meanwhile, data support is provided for an air traffic management decision maker.
Currently, most air traffic flow prediction studies are directed to short-term flow prediction at the tactical level.
Disclosure of Invention
In order to more accurately predict the air traffic flow, the invention provides a control airspace strategic flow prediction method based on a gray long-short-term memory network, which can provide scientific basis for airspace structure optimization such as sector planning, airway adjustment and the like of a control area, realize effective utilization of airspace resources and provide basis for resource demand allocation such as future personnel investment, financial investment, fixed asset investment and the like of the control area.
The invention adopts the following technical scheme for solving the technical problems:
an air traffic flow strategic prediction method based on a gray long-short term memory neural network comprises the following steps:
step 1: reading data
Reading an air traffic flow data set comprising tower, near and regional annual flight times, and then reading a factor data set influencing strategic flow in corresponding years, wherein the factor data set comprises various indexes of national and regional economy, population, consumption level, various traffic mode transportation volumes, transportation employment personnel and fixed asset investment, travel, import and export amount, line number and airplane frame number;
step 2: data preprocessing
Combining the air traffic flow data set and the influence factor data set according to years, and then processing missing values and abnormal values of the air traffic flow data set and the influence factor data set;
step 3: preliminary screening influence factors by gray correlation analysis;
step 4: extracting main features by using a main component analysis method;
step 5: establishing a grey strategic flow prediction model;
step 6: establishing a strategy flow prediction model of the long-term and short-term memory network;
step 7: and establishing a grey long-short-term memory network combined prediction model.
Step 2, filling the data record of the gap by adopting a mean value method for the missing values in the data set; and replacing the abnormal value in the data by using an interpolation method.
The specific process of the step 3 is as follows:
step 3.1: firstly, taking a flow data set as a reference vector, taking all influence factor data sets as comparison vectors, and carrying out dimensionless treatment on the reference vector and the comparison vectors by adopting initial conversion;
step 3.2: the gray correlation coefficient vector xi between the reference vector and any comparison vector is calculated by the following formula:
wherein, delta is the absolute value of the difference between all the comparison vectors and the reference vector, M is the minimum value in delta, M is the maximum value in delta, ρ [ E [0,1] is the resolution coefficient;
step 3.3: calculating the average value of the elements in the gray correlation coefficient vector to obtain gray correlation degree between the reference vector and all the comparison vectors; and finally, screening the influence factors according to the degree of association to obtain the influence factors after preliminary screening.
The specific process of the step 4 is as follows:
step 4.1, zero-mean normalization is carried out on influence factors obtained by preliminary screening;
step 4.2: calculating a correlation coefficient matrix of the influence factors obtained by screening;
step 4.3: calculating the characteristic values and characteristic vectors of the correlation coefficient matrix, wherein the number of the characteristic values and the characteristic vectors is the same as that of the influence factors obtained by screening;
step 4.4: calculating the principal component contribution rate and the accumulated contribution rate according to the characteristic value: the contribution rate of each principal component is the ratio of the corresponding characteristic value to the sum of all characteristic values, and the accumulated contribution rate is the sum of the accumulated principal component contribution rates.
The specific process of the step 5 is as follows:
step 5.1: the air traffic flow sequence in the original data sequence is as follows:
wherein 
Is the original data sequence of air traffic flow, +.>
Is the air traffic flow value of the first year,
air traffic flow value for the next year, < > is given>
An air traffic flow value of the nth year;
the main characteristic sequences of the influencing factors are as follows:
wherein ：
raw data sequence for the first main feature, < +.>
The first main feature value of the first year, +.>
The value of the main feature of the second year, which is the first main feature,/, is given by->
The first main feature value of the nth year,/main feature value of the first main feature>
Original data sequence for N-1 th main feature,/the first data sequence is the first data sequence>
The first year of the N-1 th main feature is the main feature value,/for the first year>
The value of the main feature of the second year, N-1 th main feature,/for the second year>
A main feature value of the nth year which is the N-1 th main feature;
step 5.2: performing one-time accumulation generation processing on the original data sequence to obtain a generation sequence as follows:
wherein ,
generating a sequence for air traffic flow, +.>
Accumulating values for the air traffic flow of the previous year, < >>
Accumulating values for the air traffic flow of the first two years, < >>
The values are accumulated for the air traffic flow for the previous n years,
generating a sequence for the first main feature, < >>
The values are accumulated for the main feature of the year preceding the first main feature,
accumulating values for the first two years of the first main feature, < >>
Accumulating values for the first main feature n years before the first main feature, +>
Generating sequence for the N-1 th main feature,/a. Sup..sup.>
Accumulating values for the N-1 th main feature of the year before,/for the main feature>
Accumulating values for the N-1 th main feature two years before the main feature, +.>
Accumulating values for the main features of the first n years;
step 5.3: generating a series fitting differential equation by one-time accumulation, namely:
wherein 
Generating sequence representing air traffic flow derives time t +.>
The second main feature generation sequence is represented by a least square method to obtain parameter sequences a, b 2 ,b 3 ,…,b N The method comprises the steps of carrying out a first treatment on the surface of the Solving the differential equation to obtain a time response function, namely: />
wherein ：
accumulated predicted value representing air traffic flow of the first year,/->
An accumulated predicted value representing the air traffic flow of the next year, b j Represents the j-th parameter solved in differential equation,/->
A principal characteristic accumulation value representing the first two years of the j-1 th principal characteristic, +.>
A principal characteristic accumulation value indicating n years before the j-1 st principal characteristic, +.>
Representing the air traffic flow of the nth yearIs used for accumulating predicted values;
step 5.4: deriving and restoring the time response function to obtain a predictive equation, and finally obtaining a predictive value of air traffic flow in historical years
wherein ：
predictive value representing air traffic flow of the first year, < >>
Predictive value representing air traffic flow for the next year, < >>
Accumulated predicted value representing air traffic flow of n-1 th year,/->
A predicted value of the air traffic flow in the nth year is indicated.
The specific process of the step 6 is as follows:
step 6.1: initializing long-term and short-term memory network parameters
Step 6.1.1: input layer, output layer arrangement
When an input layer and an output layer of the flow prediction neural network are constructed, the number of samples for each batch of training is initially set to be 1; time step, initially setting to 1; inputting characteristics; outputting characteristics; initializing input layer weights and biases; the weight and bias of the initialized output layer;
step 6.1.2: long-short term memory network layer arrangement
When a long-term and short-term memory network layer is built, initializing an activation function to uniformly select a tanh function and a sigmoid function; initializing the number of network layers to be 1, and setting the number of network nodes to be 10; initializing a network layer weight and bias;
step 6.1.3: loss function setting
Selecting a square difference loss function with higher convergence speed in a regression model from a flow prediction neural network;
step 6.2: setting network input layer characteristics and output characteristics
Step 6.2.1: constructing input sample vectors
The input sample is the information of the auxiliary variable obtained by the gray correlation analysis and the principal component analysis at a certain moment;
step 6.2.2: building output sample vectors
The corresponding output sample vector is information of the flying flow at a certain moment later;
step 6.2.3: data normalization
For an input sample vector and an output sample vector, a zero-mean normalization method is to be used for generating a dimensionless training data set;
step 6.3: model training and evaluation
Dividing samples according to 80% of samples serving as a training set and 20% of samples serving as a testing set, then training the neural network according to different sample numbers, time steps, training cycle numbers, hidden layer numbers and node numbers of each batch of samples of the training set, wherein the input sample vector and the output sample vector are the same as the training set: and finally, evaluating the model prediction effect according to the test set, and selecting an average absolute percentage error as an evaluation index.
The specific process of the step 7 is as follows:
step 7.1: firstly, respectively predicting by adopting a gray prediction model and a neural network model to obtain a prediction result of the model on a test set, and evaluating the prediction effect of the prediction model by using an average absolute percentage error to obtain errors of the two models on the test set;
step 7.2: and (3) carrying out weighted combination on the two models, taking the weighted sum of errors of the two models on the test set in the step (6.1) as an objective function, solving by minimizing the objective function value and taking the sum of the weights as a constraint condition to obtain an optimal weighted coefficient, wherein the prediction result of the combined model is the weighted sum of the prediction values of the two models.
The beneficial effects of the invention are as follows:
1. the invention provides a method for combining gray correlation analysis and a principal component analysis method, which comprises the steps of initially screening influence factors of flow by using the gray correlation analysis method, and extracting principal features by using the principal component analysis method so as to analyze and obtain auxiliary variables influencing strategic flow prediction.
2. The invention combines the prediction model by using the gray long-short-term memory network, combines the advantages of the long-short-term memory network and the gray prediction model, establishes a new strategic flow prediction model, and realizes the high-precision prediction of strategic flow of the controlled area.
3. The invention perfects the research of the civil aviation industry in the aspect of air traffic flow strategic prediction, promotes the intelligent development of air traffic and provides scientific basis for future strategic planning of control areas.
Drawings
FIG. 1 is a flow chart of a strategic flow prediction method based on a gray long-short term memory neural network.
Detailed Description
The invention will be described in further detail with reference to the accompanying drawings.
A control airspace strategic flow prediction method based on a gray long-short term memory network is shown in fig. 1, and comprises the following steps:
step 1: reading data:
the method comprises the steps of reading an air traffic flow data set comprising a tower, a near-year flight frame and a regional year flight frame, and then reading a factor data set influencing strategic flow in corresponding years, wherein the factor data set comprises various indexes of national and regional economy, population, consumption level, various traffic mode transportation volumes, transportation employment personnel and fixed asset investment, travel, import and export amount, line number and airplane frame number.
Step 2: data preprocessing:
and merging the air traffic flow data set and the influence factor data set according to the year, and then processing the missing value and the abnormal value of the air traffic flow data set and the influence factor data set. And filling the missing data record by adopting a mean value method for the missing values in the data set. And replacing the abnormal value in the data by using an interpolation method.
Step 3: the gray correlation analysis is utilized to preliminarily screen influencing factors, and the specific method comprises the following steps:
step 3.1: firstly, taking a flow data set as a reference vector, taking all influence factor data sets as comparison vectors, and carrying out dimensionless processing on the reference vector and the comparison vectors by adopting initial value transformation.
Step 3.2: the gray correlation coefficient vector xi between the reference vector and any comparison vector is calculated by the following formula:
wherein, delta is the absolute value of the difference between all the comparison vectors and the reference vector, M is the minimum value in delta, M is the maximum value in delta, ρ [ E ] 0,1 is the resolution coefficient, and according to the minimum information principle, the resolution coefficient of the patent is 0.5.
Step 3.3: and calculating the average value of the elements in the gray correlation coefficient vector to obtain the gray correlation degree between the reference vector and all the comparison vectors. Finally, screening the influence factors according to the degree of association to obtain the influence factors after preliminary screening
Step 4: the main characteristic is extracted by using a main component analysis method, and the specific method is as follows:
carrying out principal component analysis on the influence factors obtained by screening by the gray correlation analysis method in the step 2, and extracting main features according to analysis results, wherein the specific method comprises the following steps:
and 4.1, carrying out zero-mean normalization on the influence factors obtained by the primary screening.
Step 4.2: and calculating a correlation coefficient matrix of the influence factors obtained by screening.
Step 4.3: and calculating the eigenvalues and eigenvectors of the correlation coefficient matrix. The number of the characteristic values and the characteristic vectors is the same as that of the influence factors obtained by screening.
Step 4.4: and calculating the contribution rate of the principal component and the accumulated contribution rate according to the characteristic value. The contribution rate of each principal component is the ratio of the corresponding eigenvalue to the sum of all eigenvalues. The cumulative contribution rate is the sum of the cumulative principal component contribution rates. The main component corresponding to the characteristic value with the accumulated contribution rate exceeding 95% is taken as the main characteristic of the influence factor.
Step 5: the grey strategic flow prediction model is established by the specific method that:
step 5.1: the air traffic flow sequence in the original data sequence is as follows:
wherein 
Is the original data sequence of air traffic flow, +.>
Is the air traffic flow value of the first year,
air traffic flow value for the next year, < > is given>
Is the air traffic flow value of the nth year.
The main characteristic sequences of the influencing factors are as follows:
wherein ：
raw data sequence for the first main feature, < +.>
The first main feature value of the first year, +.>
The value of the main feature of the second year, which is the first main feature,/, is given by->
The first main feature value of the nth year,/main feature value of the first main feature>
Original data sequence for N-1 th main feature,/the first data sequence is the first data sequence>
The first year of the N-1 th main feature is the main feature value,/for the first year>
The value of the main feature of the second year, N-1 th main feature,/for the second year>
Is the main characteristic value of the nth-1 main characteristic in the nth year.
Step 5.2: performing one-time accumulation generation processing on the original data sequence to obtain a generation sequence as follows:
wherein ,
generating a sequence for air traffic flow, +.>
Accumulating the air traffic flow of the previous yearValue of->
Accumulating values for the air traffic flow of the first two years, < >>
The values are accumulated for the air traffic flow for the previous n years,
generating a sequence for the first main feature, < >>
The values are accumulated for the main feature of the year preceding the first main feature,
accumulating values for the first two years of the first main feature, < >>
Accumulating values for the first main feature n years before the first main feature, +>
Generating sequence for the N-1 th main feature,/a. Sup..sup.>
Accumulating values for the N-1 th main feature of the year before,/for the main feature>
Accumulating values for the N-1 th main feature two years before the main feature, +.>
The values are accumulated for the main features of the first n years.
Step 5.3: generating a series fitting differential equation by one-time accumulation, namely:
wherein 
Generating sequence representing air traffic flow derives time t +.>
The second main feature generation sequence is represented by a least square method to obtain parameter sequences a, b 2 ,b 3 ,…,b N 。
Solving the differential equation to obtain a time response function, namely:
wherein ：
accumulated predicted value representing air traffic flow of the first year,/->
An accumulated predicted value representing the air traffic flow of the next year, b j Represents the j-th parameter solved in differential equation,/->
A principal characteristic accumulation value representing the first two years of the j-1 th principal characteristic, +.>
A principal characteristic accumulation value indicating n years before the j-1 st principal characteristic, +.>
And the accumulated predicted value of the air traffic flow in the nth year is represented.
Step 5.4: deriving and restoring the time response function to obtain a predictive equation, and finally obtaining a predictive value of air traffic flow in historical years
wherein ：
predictive value representing air traffic flow of the first year, < >>
Predictive value representing air traffic flow for the next year, < >>
Accumulated predicted value representing air traffic flow of n-1 th year,/->
A predicted value of the air traffic flow in the nth year is indicated.
Step 6: the method for establishing the long-term and short-term memory network strategic flow prediction model comprises the following steps:
step 6.1: initializing long-term and short-term memory network parameters
Step 6.1.1: and the input layer and the output layer are arranged. When constructing the input layer and the output layer of the traffic prediction neural network, the following parameter settings are considered: the number of samples for each batch of training is initially set to 1; time step, initially setting to 1; inputting characteristics; outputting characteristics; initializing input layer weights and biases; the weight and bias of the initialized output layer.
Step 6.1.2: and setting a long-term and short-term memory network layer. When a long-term and short-term memory network layer is built, initializing an activation function to uniformly select a tanh function and a sigmoid function; when the number of network layers and the number of network nodes in each layer are considered, the number of the initialized network layers is set to be 1, and the number of the network nodes is set to be 10; network layer weights and biases are initialized.
Step 6.1.3: and (5) loss function setting. The loss function is a main basis for model parameter correction, and a square difference loss function with higher convergence speed in a regression model is selected in the flow prediction neural network.
Step 6.2: setting network input layer characteristics and output characteristics
Step 6.2.1: an input sample vector is constructed. The input sample is the information of the auxiliary variable obtained according to gray correlation analysis and principal component analysis.
Step 6.2.2: an output sample vector is constructed. The corresponding output sample vector is information of the flying flow at a certain time later.
Step 6.2.3: data normalization. And for the input sample vector and the output sample vector, a zero-mean normalization method is used for generating a dimensionless training data set.
Step 6.3: model training and evaluation. Samples were partitioned according to 80% of the samples as training sets and 20% of the samples as test sets. Then training the neural network according to the different sample numbers, time steps, training cycle numbers, hidden layer numbers and node numbers of each batch of samples of the training set by the input sample vector and the output sample vector of the training set: the number of samples per batch is selected from the {1,2,3,4,5,6} set, the time step is selected from the {1,2,3,4} set, the training period number is selected from the {50, 100, 150, 200, 250} set, the number of hidden layers is selected from the {1,2,3,4} set, and the number of hidden layer nodes is selected from the {10, 20, 30, 40} set. The model prediction effect is better by minimizing the loss function formed by the prediction output and the actual output and continuously updating various weights and offset values. And finally, evaluating the model prediction effect according to the test set, and selecting the average absolute percentage error as an evaluation index.
Step 7: the method for establishing the grey long-short-term memory network combined prediction model comprises the following steps of:
the invention mainly adopts a parallel combination mode, processes respectively through a gray model and a long-short-term memory network strategic flow prediction model, and combines the processing results, and the specific method is as follows:
step 7.1: firstly, respectively predicting by adopting a gray prediction model and a neural network model to obtain a prediction result of the model on a test set, and evaluating the prediction effect of the prediction model by using an average absolute percentage error to obtain errors of the two models on the test set.
Step 7.2: and (3) carrying out weighted combination on the two models, taking the weighted sum of errors of the two models on the test set in the step (6.1) as an objective function, solving by taking the sum of the weights as a constraint condition to obtain an optimal weighted coefficient, and enabling the average absolute error and the mean square error of the weighted combination model prediction to be minimum, so as to achieve the optimal prediction effect. The prediction result of the combined model is the weighted sum of the two model prediction values.
Claims (6)
1. A control airspace strategic flow prediction method based on a gray long-short term memory network is characterized by comprising the following steps:
step 1: reading data
Reading an air traffic flow data set comprising tower, near and regional annual flight times, and then reading a factor data set influencing strategic flow in corresponding years, wherein the factor data set comprises various indexes of national and regional economy, population, consumption level, various traffic mode transportation volumes, transportation employment personnel and fixed asset investment, travel, import and export amount, line number and airplane frame number;
step 2: data preprocessing
Combining the air traffic flow data set and the influence factor data set according to years, and then processing missing values and abnormal values of the air traffic flow data set and the influence factor data set;
step 3: preliminary screening influence factors by gray correlation analysis;
step 4: extracting main features by using a main component analysis method;
step 5: establishing a grey strategic flow prediction model; the specific process of the step 5 is as follows:
step 5.1: the air traffic flow sequence in the original data sequence is as follows:
wherein 
Is the original data sequence of air traffic flow, +.>
Is the air traffic flow value of the first year,
air traffic flow value for the next year, < > is given>
An air traffic flow value of the nth year;
the main characteristic sequences of the influencing factors are as follows:
wherein ：
raw data sequence for the first main feature, < +.>
The first main feature value of the first year, +.>
The value of the main feature of the second year, which is the first main feature,/, is given by->
The first main feature value of the nth year,/main feature value of the first main feature>
Original data sequence for N-1 th main feature,/the first data sequence is the first data sequence>
The first year of the N-1 th main feature is the main feature value,/for the first year>
The value of the main feature of the second year, N-1 th main feature,/for the second year>
A main feature value of the nth year which is the N-1 th main feature;
step 5.2: performing one-time accumulation generation processing on the original data sequence to obtain a generation sequence as follows:
wherein ,
generating a sequence for air traffic flow, +.>
The values are accumulated for the air traffic flow of the previous year,
accumulating values for the air traffic flow of the first two years, < >>
For the first n years of air trafficFlow rate accumulated value, < >>
Generating a sequence for the first main feature, < >>
Accumulating values for the first main feature of the year before the first main feature, +>
Accumulating values for the first two years of the first main feature, < >>
Accumulating values for the first main feature n years before the first main feature, +>
Generating sequence for the N-1 th main feature,/a. Sup..sup.>
Accumulating values for the N-1 th main feature of the year before,/for the main feature>
Accumulating values for the N-1 th main feature two years before the main feature, +.>
Accumulating values for the main features of the first n years;
step 5.3: generating a series fitting differential equation by one-time accumulation, namely:
wherein 
Generating sequence representing air traffic flow derives time t +.>
The second main feature generation sequence is represented by a least square method to obtain parameter sequences a, b 2 ,b 3 ,…,b N The method comprises the steps of carrying out a first treatment on the surface of the Solving the differential equation to obtain a time response function, namely:
wherein ：
accumulated predicted value representing air traffic flow of the first year,/->
An accumulated predicted value representing the air traffic flow of the next year, b j Represents the j-th parameter solved in differential equation,/->
A principal characteristic accumulation value representing the first two years of the j-1 th principal characteristic, +.>
A principal characteristic accumulation value indicating n years before the j-1 st principal characteristic, +.>
An accumulated predicted value representing an nth year air traffic flow;
step 5.4: deriving and restoring the time response function to obtain a predictive equation, and finally obtaining a predictive value of air traffic flow in historical years
wherein ：
predictive value representing air traffic flow of the first year, < >>
Predictive value representing air traffic flow for the next year, < >>
Accumulated predicted value representing air traffic flow of n-1 th year,/->
A predicted value representing an air traffic flow of the nth year;
step 6: establishing a strategy flow prediction model of the long-term and short-term memory network;
step 7: and establishing a grey long-short-term memory network combined prediction model.
2. The method for predicting the strategic flow rate of a controlled airspace based on a gray long-short term memory network according to claim 1, wherein in the step 2, for missing values in a data set, a mean method is adopted to fill in the data records of the missing; and replacing the abnormal value in the data by using an interpolation method.
3. The method for predicting strategic flow rate of a controlled airspace based on a gray long-short term memory network according to claim 1, wherein the specific process of step 3 is as follows:
step 3.1: firstly, taking a flow data set as a reference vector, taking all influence factor data sets as comparison vectors, and carrying out dimensionless treatment on the reference vector and the comparison vectors by adopting initial conversion;
step 3.2: the gray correlation coefficient vector xi between the reference vector and any comparison vector is calculated by the following formula:
wherein, delta is the absolute value of the difference between all the comparison vectors and the reference vector, M is the minimum value in delta, M is the maximum value in delta, ρ [ E [0,1] is the resolution coefficient;
step 3.3: calculating the average value of the elements in the gray correlation coefficient vector to obtain gray correlation degree between the reference vector and all the comparison vectors; and finally, screening the influence factors according to the degree of association to obtain the influence factors after preliminary screening.
4. The method for predicting strategic flow rate of controlled airspace based on gray long-short term memory network according to claim 1, wherein the specific process of step 4 is as follows:
step 4.1, zero-mean normalization is carried out on influence factors obtained by preliminary screening;
step 4.2: calculating a correlation coefficient matrix of the influence factors obtained by screening; step 4.3: calculating the characteristic values and characteristic vectors of the correlation coefficient matrix, wherein the number of the characteristic values and the characteristic vectors is the same as that of the influence factors obtained by screening;
step 4.4: calculating the principal component contribution rate and the accumulated contribution rate according to the characteristic value: the contribution rate of each principal component is the ratio of the corresponding characteristic value to the sum of all characteristic values, and the accumulated contribution rate is the sum of the accumulated principal component contribution rates.
5. The method for predicting strategic flow rate of a controlled airspace based on a gray long-short term memory network according to claim 1, wherein the specific process of step 6 is as follows:
step 6.1: initializing long-term and short-term memory network parameters
Step 6.1.1: input layer, output layer arrangement
When an input layer and an output layer of the flow prediction neural network are constructed, the number of samples for each batch of training is initially set to be 1; time step, initially setting to 1; inputting characteristics; outputting characteristics; initializing input layer weights and biases; the weight and bias of the initialized output layer;
step 6.1.2: long-short term memory network layer arrangement
When a long-term and short-term memory network layer is built, initializing an activation function to uniformly select a tanh function and a sigmoid function; initializing the number of network layers to be 1, and setting the number of network nodes to be 10; initializing a network layer weight and bias;
step 6.1.3: loss function setting
Selecting a square difference loss function with higher convergence speed in a regression model from a flow prediction neural network;
step 6.2: setting network input layer characteristics and output characteristics
Step 6.2.1: constructing input sample vectors
The input sample is the information of the auxiliary variable obtained by the gray correlation analysis and the principal component analysis at a certain moment;
step 6.2.2: building output sample vectors
The corresponding output sample vector is information of the flying flow at a certain moment later;
step 6.2.3: data normalization
For an input sample vector and an output sample vector, a zero-mean normalization method is to be used for generating a dimensionless training data set;
step 6.3: model training and evaluation
Dividing samples according to 80% of samples serving as a training set and 20% of samples serving as a testing set, then training the neural network according to different sample numbers, time steps, training cycle numbers, hidden layer numbers and node numbers of each batch of samples of the training set, wherein the input sample vector and the output sample vector are the same as the training set: and finally, evaluating the model prediction effect according to the test set, and selecting an average absolute percentage error as an evaluation index.
6. The method for predicting airspace strategic flow rate of a controlled airspace based on a gray long-short term memory network of claim 5, which is characterized in that the specific process of step 7 is as follows:
step 7.1: firstly, respectively predicting by adopting a gray prediction model and a neural network model to obtain a prediction result of the model on a test set, and evaluating the prediction effect of the prediction model by using an average absolute percentage error to obtain errors of the two models on the test set;
step 7.2: and (3) carrying out weighted combination on the two models, taking the weighted sum of errors of the two models on the test set in the step (6.1) as an objective function, solving by minimizing the objective function value and taking the sum of the weights as a constraint condition to obtain an optimal weighted coefficient, wherein the prediction result of the combined model is the weighted sum of the prediction values of the two models.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910360715.0A CN110210648B (en) | 2019-04-30 | 2019-04-30 | Gray long-short term memory network-based control airspace strategic flow prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910360715.0A CN110210648B (en) | 2019-04-30 | 2019-04-30 | Gray long-short term memory network-based control airspace strategic flow prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110210648A CN110210648A (en) | 2019-09-06 |
| CN110210648B true CN110210648B (en) | 2023-05-23 |
Family
ID=67785406
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910360715.0A Active CN110210648B (en) | 2019-04-30 | 2019-04-30 | Gray long-short term memory network-based control airspace strategic flow prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110210648B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111191842A (en) * | 2019-12-30 | 2020-05-22 | 中国民用航空飞行学院 | Civil aviation transport airport fire station site selection optimization method |
| CN111237134B (en) * | 2020-01-14 | 2022-04-01 | 上海电力大学 | Offshore double-fed wind driven generator fault diagnosis method based on GRA-LSTM-stacking model |
| CN111428932B (en) * | 2020-03-27 | 2022-12-06 | 中国民航大学 | Medium-and-long-term air traffic flow prediction method based on wavelet transformation and gray prediction |
| CN112419131B (en) * | 2020-11-20 | 2022-07-08 | 中南大学 | Method for estimating traffic origin-destination demand |
| CN112787882A (en) * | 2020-12-25 | 2021-05-11 | 国网河北省电力有限公司信息通信分公司 | Internet of things edge traffic prediction method, device and equipment |
| CN117434486B (en) * | 2023-12-20 | 2024-03-08 | 智联信通科技股份有限公司 | DC shunt metering error analysis processing method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251002A (en) * | 2016-07-18 | 2016-12-21 | 华南理工大学 | Association analysis method for the meteorological big data of overhead transmission line load-bearing capacity assessment |
| CN107045785A (en) * | 2017-02-08 | 2017-08-15 | 河南理工大学 | A kind of method of the short-term traffic flow forecast based on grey ELM neutral nets |
| CN108694476A (en) * | 2018-06-29 | 2018-10-23 | 山东财经大学 | A kind of convolutional neural networks Stock Price Fluctuation prediction technique of combination financial and economic news |
| CN109215349A (en) * | 2018-10-26 | 2019-01-15 | 同济大学 | Traffic flow forecasting method when long based on deep learning |
| CN109508812A (en) * | 2018-10-09 | 2019-03-22 | 南京航空航天大学 | A kind of aircraft Trajectory Prediction method based on profound memory network |
| CN109615169A (en) * | 2018-11-08 | 2019-04-12 | 国家电网有限公司 | A kind of distribution network reliability evaluation method based on MEA-IElman neural network |
-
2019
- 2019-04-30 CN CN201910360715.0A patent/CN110210648B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251002A (en) * | 2016-07-18 | 2016-12-21 | 华南理工大学 | Association analysis method for the meteorological big data of overhead transmission line load-bearing capacity assessment |
| CN107045785A (en) * | 2017-02-08 | 2017-08-15 | 河南理工大学 | A kind of method of the short-term traffic flow forecast based on grey ELM neutral nets |
| CN108694476A (en) * | 2018-06-29 | 2018-10-23 | 山东财经大学 | A kind of convolutional neural networks Stock Price Fluctuation prediction technique of combination financial and economic news |
| CN109508812A (en) * | 2018-10-09 | 2019-03-22 | 南京航空航天大学 | A kind of aircraft Trajectory Prediction method based on profound memory network |
| CN109215349A (en) * | 2018-10-26 | 2019-01-15 | 同济大学 | Traffic flow forecasting method when long based on deep learning |
| CN109615169A (en) * | 2018-11-08 | 2019-04-12 | 国家电网有限公司 | A kind of distribution network reliability evaluation method based on MEA-IElman neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110210648A (en) | 2019-09-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110210648B (en) | Gray long-short term memory network-based control airspace strategic flow prediction method | |
| WO2021082393A1 (en) | Airport surface variable slide-out time prediction method based on big data deep learning | |
| CN103226899B (en) | Based on the space domain sector method for dynamically partitioning of air traffic feature | |
| CN110766212A (en) | Ultra-short-term photovoltaic power prediction method for historical data missing electric field | |
| WO2020248228A1 (en) | Computing node load prediction method in a hadoop platform | |
| CN105512745A (en) | Wind power section prediction method based on particle swarm-BP neural network | |
| CN101551884A (en) | A fast CVR electric load forecast method for large samples | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| CN109858700A (en) | BP neural network heating system energy consumption prediction technique based on similar screening sample | |
| CN115271186B (en) | Reservoir water level prediction and early warning method based on delay factor and PSO RNN Attention model | |
| CN114777192B (en) | Secondary network heat supply autonomous optimization regulation and control method based on data association and deep learning | |
| CN114444660A (en) | Short-term power load prediction method based on attention mechanism and LSTM | |
| Wu et al. | An improved svm model for flight delay prediction | |
| Xie et al. | Short-term power load forecasting model based on fuzzy neural network using improved decision tree | |
| CN111882114B (en) | Short-time traffic flow prediction model construction method and prediction method | |
| CN110717581A (en) | Short-term load prediction method based on temperature fuzzy processing and DBN | |
| CN111311905A (en) | Particle swarm optimization wavelet neural network-based expressway travel time prediction method | |
| CN116109212B (en) | Airport operation efficiency evaluation index design and monitoring method | |
| CN111008661B (en) | Croston-XGboost prediction method for reserve demand of aircraft engine | |
| CN111612227A (en) | Load prediction method based on K-means clustering and bat optimization neural network | |
| CN114463978B (en) | Data monitoring method based on track traffic information processing terminal | |
| CN108345996B (en) | System and method for reducing wind power assessment electric quantity | |
| Su et al. | Robust modeling for fleet assignment problem based on GASVR forecast | |
| CN110084516B (en) | Method for revising civil aviation segment operation time standard | |
| CN103020455B (en) | Multi-target model updating method for optimizing operation of coaxial cable sheath machine |
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 |