CN107480824A - Urban Rail Transit Stations passenger flow estimation system and method in short-term - Google Patents
Urban Rail Transit Stations passenger flow estimation system and method in short-term Download PDFInfo
- Publication number
- CN107480824A CN107480824A CN201710706219.7A CN201710706219A CN107480824A CN 107480824 A CN107480824 A CN 107480824A CN 201710706219 A CN201710706219 A CN 201710706219A CN 107480824 A CN107480824 A CN 107480824A
- Authority
- CN
- China
- Prior art keywords
- module
- data
- afc
- video
- historical
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000001914 filtration Methods 0.000 claims abstract description 126
- 238000007781 pre-processing Methods 0.000 claims abstract description 117
- 238000013528 artificial neural network Methods 0.000 claims abstract description 64
- 238000012216 screening Methods 0.000 claims abstract description 63
- 241001269238 Data Species 0.000 claims description 120
- 230000004927 fusion Effects 0.000 claims description 41
- 241000894007 species Species 0.000 claims description 39
- 238000012545 processing Methods 0.000 claims description 35
- 230000002159 abnormal effect Effects 0.000 claims description 9
- 238000004321 preservation Methods 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 5
- 235000013399 edible fruits Nutrition 0.000 claims description 4
- 238000002156 mixing Methods 0.000 claims description 3
- 238000007499 fusion processing Methods 0.000 claims description 2
- 210000005036 nerve Anatomy 0.000 claims 2
- 230000007935 neutral effect Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000003825 pressing Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Urban Rail Transit Stations passenger flow estimation system and method in short-term, wherein, system is counted equipment, data screening module, data preprocessing module, Federated Kalman Filtering module, neural network prediction module and database module and formed by AFC equipment, video passenger flow；The method have the benefit that：A kind of Urban Rail Transit Stations passenger flow estimation system and method in short-term is proposed, the present invention can improve the diversity of data source, comprehensive and accuracy, while can also improve the accuracy of data, finally make it that prediction result is more accurate.
Description
Technical field
The present invention relates to a kind of a kind of passenger flow estimation technology, more particularly to passenger flow is pre- in short-term for Urban Rail Transit Stations in short-term
Examining system and method.
Background technology
The characteristics of there is the current urban track traffic in China passenger traffic total amount to be continuously increased, passenger flow intensity grows steadily, close
Manage and carry out passenger flow induction, safety management and operation tissue that passenger flow estimation is advantageous to track traffic exactly.
It is different according to demand, passenger flow estimation can be divided into medium- and long-term forecasting, short-term forecast and short-term prediction；It is medium-term and long-term pre-
Survey and (be often referred to following 10~25 years) and be mainly used to Auxiliary Track transit's routes development plan and Station Design etc.；Short-term forecast
(being often referred in following 1 week or 1 month) is then mainly used in traffic behavior assessment；If for the purpose of real-time management, need to rely on
Short-term prediction (is often referred in following 5,15 or 30 minutes), and it is the pass for realizing track traffic security control and orderly passenger organization
Key.
Nerual network technique is very suitable for for handling foregoing passenger flow estimation problem in short-term, existing because of its own characteristic
Having also has the report of correlation in document.But there is following defect for the existing technology of passenger flow estimation in short-term based on neutral net：
The passenger flow variation characteristic of track traffic station is except showing as the cycle under normality (working day and two-day weekend)
Outside property and peak property, otherness and particularity can be also shown because of abnormal factors such as festivals or holidays, city large-scale activities, by very
Passenger flow change caused by state factor influences great on the safe operation of urban track traffic；In the prior art, more ripe base
In the passenger flow estimation technology in short-term of neutral net, the passenger flow estimation being directed to mostly under normality, shorter mention normality and very
The combination problem of state, this allow for prior art under the conditions of abnormal to website in short-term passenger flow estimation not system, comprehensively.
Meanwhile in the prior art, common volume of the flow of passengers information collecting device mainly has AFC equipment (AFC equipment)
With video image processing apparatus (video passenger flow statistics equipment), when carrying out passenger flow estimation in short-term, its volume of the flow of passengers information is only from
A certain equipment, information source is more single, and data are not necessarily reliable, accurate；In addition, prior art is for the volume of the flow of passengers that collects
Information is not dealt with, is just directly used in neutral net, and this, which allows for some abnormal datas, can also act on neutral net, from
And influence the accuracy of prediction.
The content of the invention
The problem of for background technology, the present invention propose a kind of Urban Rail Transit Stations passenger flow estimation system in short-term,
Its innovation is：Passenger flow estimation system counts equipment, number to the Urban Rail Transit Stations by AFC equipment, video passenger flow in short-term
According to screening module, data preprocessing module, Federated Kalman Filtering module, neural network prediction module and database module group
Into；
The AFC equipment and video passenger flow statistics equipment are connected with data screening module；Data screening module respectively with
Data preprocessing module, Federated Kalman Filtering module connect with database module；Data preprocessing module is blocked with federation respectively
Kalman Filtering module connects with database module；Federated Kalman Filtering module respectively with neural network prediction module and database
Module connects；Database module is connected with neural network prediction module；
The AFC equipment can export the volume of the flow of passengers data counted in real time to data screening module, the output of AFC equipment
Volume of the flow of passengers data are designated as AFC data；
The video passenger flow statistics equipment can be by the volume of the flow of passengers data counted on output in real time to data screening module, video
The volume of the flow of passengers data of passenger flow statisticses equipment output are designated as video data；
Current date species can be identified for the data screening module, after identifying date species, the data
The AFC data received and video data transmission to database module are carried out classification preservation by screening module；It is stored in database mould
AFC data in block are designated as AFC historical datas, and the video data being stored in database module is designated as video historical data；So
Afterwards, data screening module is carried out according to the date species identified to data preprocessing module and Federated Kalman Filtering module
Control：The date species includes normal day and special day, and working day and two-day weekend belong to normal day, festivals or holidays and active day category
In special day；If the date species of current date is identified as normal day, data screening module is defeated to data preprocessing module
Go out control signal one, and to Federated Kalman Filtering module output control signal two；If the date species of current date is identified
For special day, then data screening module is to data preprocessing module output control signal three, and to Federated Kalman Filtering module
Output control signal four；
When the data preprocessing module receives control signal three, data preprocessing module does not work；Data prediction mould
Block receives control signal for the moment, and data preprocessing module calls multiple AFC historical datas and multiple videos to go through from database module
History data, then data preprocessing module AFC historical datas and video historical data are handled respectively；
When data preprocessing module is handled AFC historical datas, the average of multiple AFC historical datas is first calculatedWith
Standard deviation sA, obtain first threshold scopeThen judge each AFC historical datas whether in first threshold
In the range of：AFC historical datas in the range of first threshold are designated as being worth, the AFC historical datas not in the range of first threshold
Bad value is designated as, if all AFC historical datas are preferably worth, data preprocessing module exports to Federated Kalman Filtering moduleIf all AFC historical datas are bad value, data preprocessing module by multiple AFC historical datas withDifference minimum
That one is exported to Federated Kalman Filtering module；If existing good value has bad value in multiple AFC historical datas again, bad value is picked
Remove and recalculate the remaining average being worth wellThen willExport to Federated Kalman Filtering module；
When data preprocessing module is handled video historical data, the average of multiple video historical datas is first calculated
With standard deviation sB, obtain Second Threshold scopeThen judge each video historical data whether second
In threshold range：Video historical data in the range of Second Threshold is designated as virtual value, the video not in the range of Second Threshold
Historical data is designated as invalid value, if all AFC historical datas are virtual value, data preprocessing module is filtered to federal Kalman
Ripple module exportsIf all video historical datas are invalid value, data preprocessing module is by multiple video historical datas
In withThat minimum one of difference is exported to Federated Kalman Filtering module；If existing virtual value in multiple video historical datas
There is invalid value again, then invalid value is rejected and recalculate the average of remaining virtual valueThen willOutput to federation blocks
Kalman Filtering module；
In the output signal of data preprocessing module, output signal corresponding with AFC historical datas is designated as signal one, with regarding
Output signal corresponding to frequency historical data is designated as signal two；
When the Federated Kalman Filtering module receives control signal two, to the He of signal one of data preprocessing module output
Signal two makees use processing, then exports result to neural network prediction module；The Federated Kalman Filtering
When module receives control signal four, Federated Kalman Filtering module is called newest AFC historical datas from database module and regarded
Frequency historical data, then exchanges the AFC historical datas used and video historical data makees use processing, then ties processing
Fruit is exported to neural network prediction module；
After the neural network prediction module receives the result, neural network prediction module is adjusted from database module
With multiple previous data, the result and multiple previous data are then subjected to passenger flow forecast together as input vector
Processing, obtains passenger flow forecast data, then outwards exports passenger flow forecast data；Passenger flow forecast data outwards export
Meanwhile passenger flow forecast data are also sent database module to preserve by neural network prediction module.
The present invention principle be：In the present invention, volume of the flow of passengers information source in AFC equipment (AFC system) and regards
Two kinds of equipment of frequency passenger flow statisticses equipment, the AFC data and video data obtained by both equipment need to be through Federated Kalman Filtering
Module is just used in neural network prediction module after carrying out fusion treatment, carries out fusion treatment to two kinds of data, can both reduce
Because of AFC equipment skip, misread and mix the factors such as ticket caused by data error, and can solve video passenger flow statistics equipment by light and
Environment influence it is big, when particularly passenger flow is intensive, overlapped, the problem of its accuracy of detection declines, so as to more accurately reflect
Go out the actual volume of the flow of passengers, data are obtained compared to using single equipment, detection coverage during data is obtained with two kinds of equipment
Wider, data source is more comprehensive；For the normality described in background technology and abnormal problem, inventor is set in scheme
Data screening module and data preprocessing module, data screening module is used to sort data and classification storage；Together
When, data screening module can also will belong to the data of special day by the data output for belonging to normal day to data preprocessing module
It is directly output to Federated Kalman Filtering module；The passenger flow data of normal day has the changing rule using week as the cycle, different weeks
The regularity of distribution is basically identical in time for passenger flow on the same day, and existing conventional process is to take the average passenger flow of last week or history
Amount.When taking volume of the flow of passengers last week, last week will also be inputted to nerve net simultaneously due to influence of fluctuations caused by some accidentalia
Network fallout predictor.Therefore, the volume of the flow of passengers of last week will be no longer suitably as prediction input；And when using the average volume of the flow of passengers of history,
It is but a kind of mean method for using and averaging.Therefore, the present invention is provided with a data preprocessing module, will can go through
Abnormality value removing in history data, to exclude because the abnormal datas caused by accidentalia such as interference impact to system, finally
So that the data for neural network prediction module are more reasonable, enhancing prediction stationarity, precision of prediction is improved.
Based on aforementioned schemes, following preferred scheme is proposed present invention is alternatively directed to Federated Kalman Filtering module：It is described
Federated Kalman Filtering module includes part filter module one, part filter module two and information fusion module；Described information is melted
Matched moulds block has two inputs and three output ends, and it is defeated that three output ends of information fusion module are designated as the first feedback signal respectively
Go out end, the second feedback signal output and main output end；The output end of part filter module one and local filtration module two it is defeated
Go out end to be connected correspondingly with two inputs of information fusion module, first feedback signal output and part filter
The feedback signal reception end connection of module one, second feedback signal output and the feedback signal of part filter module two connect
Receiving end is connected, and the main output end is connected with neural network prediction module；Data preprocessing module is to Federated Kalman Filtering mould
During block output signal, data preprocessing module exports signal one to part filter module one, and data preprocessing module is by signal
Two export to part filter module two；Federated Kalman Filtering module calls AFC historical datas and video to go through from database module
During history data, in corresponding AFC historical datas input part filter module one, corresponding video historical data input part filter module
In two.
Based on aforementioned system, the invention also provides a kind of Urban Rail Transit Stations passenger flow forecasting in short-term, the party
The Urban Rail Transit Stations that method is based in short-term passenger flow estimation system as it was previously stated, specific method is：The city rail
Passenger flow forecasting includes traffic website in short-term：
For Urban Rail Transit Stations in short-term in passenger flow estimation system operation, AFC equipment and video passenger flow count equipment
Periodically AFC data and video data are exported to data screening module；After receiving AFC data and video data every time, city
Passenger flow estimation system carries out passenger flow forecast operation to track traffic website with regard to pressing step such as in short-term：
1) current date species is identified data screening module, the AFC data and video data that then will be received
Transmit to database module and carry out classification preservation；Then：If the date species of current date is identified as normal day, data sieve
Modeling block enters to data preprocessing module output control signal one, and to Federated Kalman Filtering module output control signal two
Enter step 2)；If the date species of current date is identified as special day, data screening module is defeated to data preprocessing module
Go out control signal three, and to Federated Kalman Filtering module output control signal four, into step 3)；
2) data preprocessing module calls multiple AFC historical datas and multiple video historical datas from database module, so
Data preprocessing module is handled AFC historical datas and video historical data respectively afterwards, generates corresponding signal one and letter
Numbers two, into step 4)；
3) Federated Kalman Filtering module calls newest AFC historical datas and video historical data from database module,
Then the AFC historical datas used are exchanged and video historical data makees use processing, then export result to god
Through neural network forecast module, into step 5)；
4) signal one and signal two that Federated Kalman Filtering module exports to data preprocessing module are made at information fusion
Reason, then exports result to neural network prediction module, into step 5)；
5) neural network prediction module calls multiple previous data from database module, then by the result and more
Individual previous data carry out passenger flow forecast processing together as input vector, passenger flow forecast data are obtained, then by the volume of the flow of passengers
Prediction data outwards exports, and while passenger flow forecast data outwards export, neural network prediction module is also by passenger flow forecast
Data send database module to preserve.
Based on foregoing method, proposed present invention is alternatively directed to Federated Kalman Filtering module and its to the processing mode of data
Following preferred scheme, the concrete structure of Federated Kalman Filtering module as it was previously stated, Federated Kalman Filtering module by such as
Under type carries out Federated Kalman Filtering processing：
(1) part filter module one is handled signal one using linear Kalman filter method, obtains filter result
One, filter result one is exported to information fusion module, meanwhile, part filter module two uses linear Kalman filter method pair
Signal two is handled, and obtains filter result two, and filter result two is exported to information fusion module；
(2) information fusion module carries out use processing to filter result one and filter result two, then ties processing
Fruit is exported to neural network prediction module；
Information fusion module make the blending algorithm that is used during use processing for：
Wherein, k represents step number,For global best estimates value, Pg(k) it is the variance of global best estimates value,
For the estimate corresponding to filter result one, PZ(k) variance corresponding to filter result one,Represent to PZ(k) invert,For the estimate corresponding to filter result two, PV(k) variance corresponding to filter result two,Represent to PV(k)
Invert；
Information fusion module willAnd Pg(k)Part filter module one is fed back to, information fusion module willAnd Pg(k)Feed back to part filter module two；βZ(k) the distribution system of part filter module one is corresponded to when being walked for kth
Number, βV(k) distribution coefficient of part filter module two, β are corresponded to when being walked for kthZ(k)+βV(k)=1,Represent to βZ(k)
Invert,Represent to βV(k) invert.
The method have the benefit that：Propose a kind of Urban Rail Transit Stations passenger flow estimation system and side in short-term
Method, the present invention can improve the diversity of data source, comprehensive and accuracy, while can also improve the accuracy of data, most
Cause that prediction result is more accurate eventually.
Brief description of the drawings
Fig. 1, the present invention principle schematic；
Fig. 2, Federated Kalman Filtering module principle schematic；
Title in figure corresponding to each mark is respectively：AFC equipment 1, video passenger flow statistics equipment 2, data screening mould
Block 3, data preprocessing module 4, Federated Kalman Filtering module 5, the 5-1 of part filter module one, the 5-2 of part filter module two,
Information fusion module 5-3, neural network prediction module 6, database module 7.
Embodiment
Passenger flow estimation system, its innovation are a kind of Urban Rail Transit Stations in short-term：The Urban Rail Transit Stations
Passenger flow estimation system counts equipment 2, data screening module 3, data preprocessing module 4, connection by AFC equipment 1, video passenger flow in short-term
Nation's Kalman filtering module 5, neural network prediction module 6 and database module 7 form；
The AFC equipment 1 and video passenger flow statistics equipment 2 are connected with data screening module 3；Data screening module 3 is divided
It is not connected with data preprocessing module 4, Federated Kalman Filtering module 5 and database module 7；Data preprocessing module 4 is distinguished
It is connected with Federated Kalman Filtering module 5 and database module 7；Federated Kalman Filtering module 5 respectively with neural network prediction
Module 6 and database module 7 connect；Database module 7 is connected with neural network prediction module 6；
The AFC equipment 1 can output be defeated to data screening module 3, AFC equipment 1 in real time by the volume of the flow of passengers data counted on
The volume of the flow of passengers data gone out are designated as AFC data；
The video passenger flow statistics equipment 2 can export the volume of the flow of passengers data counted on to data screening module 3 in real time, depending on
The volume of the flow of passengers data that frequency passenger flow statisticses equipment 2 exports are designated as video data；
Current date species can be identified for the data screening module 3, after identifying date species, the data
The AFC data received and video data transmission to database module 7 are carried out classification preservation by screening module 3；It is stored in database
AFC data in module 7 are designated as AFC historical datas, and the video data being stored in database module 7 is designated as video history number
According to；Then, data screening module 3 is according to the date species identified, to data preprocessing module 4 and Federated Kalman Filtering mould
Block 5 is controlled：The date species includes normal day and special day, and working day and two-day weekend belong to normal day, festivals or holidays and
Active day belongs to special day；If the date species of current date is identified as normal day, data screening module 3 is located in advance to data
The output control signal one of module 4 is managed, and to the output control signal two of Federated Kalman Filtering module 5；If the date of current date
Species is identified as special day, then data screening module 3 is blocked to the output control signal three of data preprocessing module 4, and to federation
The output control signal four of Kalman Filtering module 5；
When the data preprocessing module 4 receives control signal three, data preprocessing module 4 does not work；Data prediction
Module 4 receives control signal for the moment, and data preprocessing module 4 is called multiple AFC historical datas from database module 7 and multiple regarded
Frequency historical data, then data preprocessing module 4 AFC historical datas and video historical data are handled respectively；
When data preprocessing module 4 is handled AFC historical datas, the average of multiple AFC historical datas is first calculated
With standard deviation sA, obtain first threshold scopeThen judge each AFC historical datas whether in the first threshold
In the range of value：AFC historical datas in the range of first threshold are designated as being worth, the AFC history numbers not in the range of first threshold
According to bad value is designated as, if all AFC historical datas are preferably worth, data preprocessing module 4 is defeated to Federated Kalman Filtering module 5
Go outIf all AFC historical datas are bad value, data preprocessing module 4 by multiple AFC historical datas withDifference is most
That small one is exported to Federated Kalman Filtering module 5；If existing good value has bad value again in multiple AFC historical datas, will
Abnormal data erasing simultaneously recalculates the remaining average being worth wellThen willExport to Federated Kalman Filtering module 5；
When data preprocessing module 4 is handled video historical data, the average of multiple video historical datas is first calculatedWith standard deviation sB, obtain Second Threshold scopeThen judge each video historical data whether
In two threshold ranges：Video historical data in the range of Second Threshold is designated as virtual value, regarding not in the range of Second Threshold
Frequency historical data is designated as invalid value, if all AFC historical datas are virtual value, data preprocessing module 4 is to federal karr
Graceful filtration module 5 exportsIf all video historical datas are invalid value, data preprocessing module 4 goes through multiple videos
In history data withThat minimum one of difference is exported to Federated Kalman Filtering module 5；If in multiple video historical datas both
There is virtual value to have invalid value again, then invalid value is rejected and recalculate the average of remaining virtual valueThen willOutput
To Federated Kalman Filtering module 5；
In the output signal of data preprocessing module 4, output signal corresponding with AFC historical datas is designated as signal one, with
Output signal corresponding to video historical data is designated as signal two；
When the Federated Kalman Filtering module 5 receives control signal two, to the signal one of the output of data preprocessing module 4
Make use processing with signal two, then export result to neural network prediction module 6；Federal Kalman's filter
When ripple module 5 receives control signal four, Federated Kalman Filtering module 5 calls newest AFC historical datas from database module 7
With video historical data, then exchange the AFC historical datas used and video historical data makees use processing, then will place
Reason result is exported to neural network prediction module 6；
After the neural network prediction module 6 receives the result, neural network prediction module 6 is from database module
7 call multiple previous data, and the result and multiple previous data then are carried out into the volume of the flow of passengers together as input vector
Prediction is handled, and is obtained passenger flow forecast data, is then outwards exported passenger flow forecast data, passenger flow forecast data are outwards defeated
While going out, passenger flow forecast data are also sent database module 7 to preserve by neural network prediction module 6.
Further, the Federated Kalman Filtering module 5 includes the 5-1 of part filter module one, part filter module two
5-2 and information fusion module 5-3；Described information Fusion Module 5-3 has two inputs and three output ends, information fusion module
5-3 three output ends are designated as the first feedback signal output, the second feedback signal output and main output end respectively；Local filter
The 5-1 of ripple module one output end and the 5-2 of local filtration module two output end and information fusion module 5-3 two inputs one
One is correspondingly connected, and first feedback signal output is connected with the 5-1 of part filter module one feedback signal reception end, institute
The feedback signal reception end that the second feedback signal output is stated with the 5-2 of part filter module two is connected, the main output end and god
Connected through neural network forecast module 6；When data preprocessing module 4 is to Federated Kalman Filtering 5 output signal of module, data prediction
Signal one is exported to the 5-1 of part filter module one, data preprocessing module 4 and exports signal two to part filter mould by module 4
The 5-2 of block two；Federated Kalman Filtering module 5 calls AFC historical datas and during video historical data from database module 7, accordingly
In the 5-1 of AFC historical datas input part filter module one, in the 5-2 of corresponding video historical data input part filter module two.
A kind of Urban Rail Transit Stations passenger flow forecasting in short-term, including Urban Rail Transit Stations passenger flow estimation in short-term
System, passenger flow estimation system counts equipment 2 to the Urban Rail Transit Stations by AFC equipment 1, video passenger flow in short-term, data are sieved
7 groups of modeling block 3, data preprocessing module 4, Federated Kalman Filtering module 5, neural network prediction module 6 and database module
Into；
The AFC equipment 1 and video passenger flow statistics equipment 2 are connected with data screening module 3；Data screening module 3 is divided
It is not connected with data preprocessing module 4, Federated Kalman Filtering module 5 and database module 7；Data preprocessing module 4 is distinguished
It is connected with Federated Kalman Filtering module 5 and database module 7；Federated Kalman Filtering module 5 respectively with neural network prediction
Module 6 and database module 7 connect；Database module 7 is connected with neural network prediction module 6；
The AFC equipment 1 can output be defeated to data screening module 3, AFC equipment 1 in real time by the volume of the flow of passengers data counted on
The volume of the flow of passengers data gone out are designated as AFC data；
The video passenger flow statistics equipment 2 can export the volume of the flow of passengers data counted on to data screening module 3 in real time, depending on
The volume of the flow of passengers data that frequency passenger flow statisticses equipment 2 exports are designated as video data；
Current date species can be identified for the data screening module 3, after identifying date species, the data
The AFC data received and video data transmission to database module 7 are carried out classification preservation by screening module 3；It is stored in database
AFC data in module 7 are designated as AFC historical datas, and the video data being stored in database module 7 is designated as video history number
According to；Then, data screening module 3 is according to the date species identified, to data preprocessing module 4 and Federated Kalman Filtering mould
Block 5 is controlled：The date species includes normal day and special day, and working day and two-day weekend belong to normal day, festivals or holidays and
Active day belongs to special day；If the date species of current date is identified as normal day, data screening module 3 is located in advance to data
The output control signal one of module 4 is managed, and to the output control signal two of Federated Kalman Filtering module 5；If the date of current date
Species is identified as special day, then data screening module 3 is blocked to the output control signal three of data preprocessing module 4, and to federation
The output control signal four of Kalman Filtering module 5；
When the data preprocessing module 4 receives control signal three, data preprocessing module 4 does not work；Data prediction
Module 4 receives control signal for the moment, and data preprocessing module 4 is called multiple AFC historical datas from database module 7 and multiple regarded
Frequency historical data, then data preprocessing module 4 AFC historical datas and video historical data are handled respectively；
When data preprocessing module 4 is handled AFC historical datas, the average of multiple AFC historical datas is first calculated
With standard deviation sA, obtain first threshold scopeThen judge each AFC historical datas whether in the first threshold
In the range of value：AFC historical datas in the range of first threshold are designated as being worth, the AFC history numbers not in the range of first threshold
According to bad value is designated as, if all AFC historical datas are preferably worth, data preprocessing module 4 is defeated to Federated Kalman Filtering module 5
Go outIf all AFC historical datas are bad value, data preprocessing module 4 by multiple AFC historical datas withDifference is most
That small one is exported to Federated Kalman Filtering module 5；If existing good value has bad value again in multiple AFC historical datas, will
Abnormal data erasing simultaneously recalculates the remaining average being worth wellThen willExport to Federated Kalman Filtering module 5；
When data preprocessing module 4 is handled video historical data, the average of multiple video historical datas is first calculatedWith standard deviation sB, obtain Second Threshold scopeThen judge each video historical data whether second
In threshold range：Video historical data in the range of Second Threshold is designated as virtual value, the video not in the range of Second Threshold
Historical data is designated as invalid value, if all AFC historical datas are virtual value, data preprocessing module 4 is to federal Kalman
Filtration module 5 exportsIf all video historical datas are invalid value, data preprocessing module 4 is by multiple video history
In data withThat minimum one of difference is exported to Federated Kalman Filtering module 5；It is if existing in multiple video historical datas
Virtual value has invalid value again, then rejects invalid value and recalculate the average of remaining virtual valueThen willOutput is extremely
Federated Kalman Filtering module 5；
In the output signal of data preprocessing module 4, output signal corresponding with AFC historical datas is designated as signal one, with
Output signal corresponding to video historical data is designated as signal two；
When the Federated Kalman Filtering module 5 receives control signal two, to the signal one of the output of data preprocessing module 4
Make use processing with signal two, then export result to neural network prediction module 6；Federal Kalman's filter
When ripple module 5 receives control signal four, Federated Kalman Filtering module 5 calls newest AFC historical datas from database module 7
With video historical data, then exchange the AFC historical datas used and video historical data makees use processing, then will place
Reason result is exported to neural network prediction module 6；
After the neural network prediction module 6 receives the result, neural network prediction module 6 is from database module
7 call multiple previous data (previous data are the actual volume of the flow of passengers data got using technological means), then by the place
Manage result and multiple previous data and carry out passenger flow forecast processing together as input vector, obtain passenger flow forecast data, so
Passenger flow forecast data are outwards exported afterwards, while passenger flow forecast data outwards export, neural network prediction module 6 will also
Passenger flow forecast data send database module 7 to preserve；
Its innovation is：Passenger flow forecasting includes the Urban Rail Transit Stations in short-term：
For Urban Rail Transit Stations in short-term in passenger flow estimation system operation, AFC equipment 1 and video passenger flow count equipment
2 periodically export AFC data and video data to data screening module 3；After receiving AFC data and video data every time, city
Passenger flow estimation system carries out passenger flow forecast operation to city's track traffic website with regard to pressing step such as in short-term：
1) current date species is identified data screening module 3, the AFC data and video data that then will be received
Transmit to database module 7 and carry out classification preservation；Then：If the date species of current date is identified as normal day, data
Screening module 3 is to the output control signal one of data preprocessing module 4, and to the output control signal of Federated Kalman Filtering module 5
Two, into step 2)；If the date species of current date is identified as special day, data screening module 3 is to data prediction
The output control signal three of module 4, and to the output control signal four of Federated Kalman Filtering module 5, into step 3)；
2) data preprocessing module 4 calls multiple AFC historical datas and multiple video historical datas from database module 7,
Then data preprocessing module 4 is handled AFC historical datas and video historical data respectively, generates the corresponding He of signal one
Signal two, into step 4)；
3) Federated Kalman Filtering module 5 calls newest AFC historical datas and video history number from database module 7
According to, then exchange the AFC historical datas used and video historical data make use processing, then by result export to
Neural network prediction module 6, into step 5)；
4) signal one and signal two that Federated Kalman Filtering module 5 exports to data preprocessing module 4 make information fusion
Processing, then exports result to neural network prediction module 6, into step 5)；
5) neural network prediction module 6 calls multiple previous data from database module 7, then by the result and
Multiple previous data carry out passenger flow forecast processing together as input vector, passenger flow forecast data are obtained, then by passenger flow
Amount prediction data outwards exports, and while passenger flow forecast data outwards export, neural network prediction module 6 is also pre- by the volume of the flow of passengers
Surveying data send database module 7 to preserve.
Further, the Federated Kalman Filtering module 5 includes the 5-1 of part filter module one, part filter module two
5-2 and information fusion module 5-3；Described information Fusion Module 5-3 has two inputs and three output ends, information fusion module
5-3 three output ends are designated as the first feedback signal output, the second feedback signal output and main output end respectively；Local filter
The 5-1 of ripple module one output end and the 5-2 of local filtration module two output end and information fusion module 5-3 two inputs one
One is correspondingly connected, and first feedback signal output is connected with the 5-1 of part filter module one feedback signal reception end, institute
The feedback signal reception end that the second feedback signal output is stated with the 5-2 of part filter module two is connected, the main output end and god
Connected through neural network forecast module 6；When data preprocessing module 4 is to Federated Kalman Filtering 5 output signal of module, data prediction
Signal one is exported to the 5-1 of part filter module one, data preprocessing module 4 and exports signal two to part filter mould by module 4
The 5-2 of block two；Federated Kalman Filtering module 5 calls AFC historical datas and during video historical data from database module 7, accordingly
In the 5-1 of AFC historical datas input part filter module one, in the 5-2 of corresponding video historical data input part filter module two；
The Federated Kalman Filtering module 5 carries out Federated Kalman Filtering processing as follows：
(1) 5-1 of part filter module one is handled signal one using linear Kalman filter method, obtains filtering knot
Fruit one, filter result one is exported to information fusion module 5-3, meanwhile, the 5-2 of part filter module two is filtered using linear Kalman
Wave method is handled signal two, obtains filter result two, and filter result two is exported to information fusion module 5-3；
(2) information fusion module 5-3 carries out use processing to filter result one and filter result two, then will processing
As a result export to neural network prediction module 6；
Information fusion module 5-3 make the blending algorithm that is used during use processing for：
Wherein, k represents step number,For global best estimates value, Pg(k) it is the variance of global best estimates value,
For the estimate corresponding to filter result one, PZ(k) variance corresponding to filter result one,Represent to PZ(k) ask
It is inverse,For the estimate corresponding to filter result two, PV(k) variance corresponding to filter result two,Expression pair
PV(k) invert；
Information fusion module 5-3 willAnd Pg(k)Feed back to the 5-1 of part filter module one, information fusion mould
Block 5-3 willAnd Pg(k)Feed back to the 5-2 of part filter module two；βZ(k) part filter module is corresponded to when being walked for kth
One 5-1 distribution coefficient, βV(k) 5-2 of part filter module two distribution coefficient, β are corresponded to when being walked for kthZ(k)+βV(k)=1,Represent to βZ(k) invert,Represent to βV(k) invert.
Claims (4)
- A kind of 1. Urban Rail Transit Stations passenger flow estimation system in short-term, it is characterised in that：The Urban Rail Transit Stations are short When passenger flow estimation system by AFC equipment (1), video passenger flow count equipment (2), data screening module (3), data preprocessing module (4), Federated Kalman Filtering module (5), neural network prediction module (6) and database module (7) composition；The AFC equipment (1) and video passenger flow statistics equipment (2) are connected with data screening module (3)；Data screening module (3) it is connected respectively with data preprocessing module (4), Federated Kalman Filtering module (5) and database module (7)；Data are located in advance Reason module (4) is connected with Federated Kalman Filtering module (5) and database module (7) respectively；Federated Kalman Filtering module (5) It is connected respectively with neural network prediction module (6) and database module (7)；Database module (7) and neural network prediction module (6) connect；The AFC equipment (1) can be by the volume of the flow of passengers data counted on output in real time to data screening module (3), AFC equipment (1) The volume of the flow of passengers data of output are designated as AFC data；The video passenger flow statistics equipment (2) can export the volume of the flow of passengers data counted on to data screening module (3) in real time, depending on The volume of the flow of passengers data of frequency passenger flow statisticses equipment (2) output are designated as video data；Current date species can be identified for the data screening module (3), after identifying date species, the data sieve The AFC data received and video data transmission to database module (7) are carried out classification preservation by modeling block (3)；It is stored in data AFC data in library module (7) are designated as AFC historical datas, and the video data being stored in database module (7) is designated as video and gone through History data；Then, data screening module (3) is according to the date species identified, to data preprocessing module (4) and federal karr Graceful filtration module (5) is controlled：The date species includes normal day and special day, and working day and two-day weekend belong to normal Day, festivals or holidays and active day belong to special day；If the date species of current date is identified as normal day, data screening module (3) to data preprocessing module (4) output control signal one, and to Federated Kalman Filtering module (5) output control signal two； If the date species of current date is identified as special day, data screening module (3) is exported to data preprocessing module (4) and controlled Signal three processed, and to Federated Kalman Filtering module (5) output control signal four；When the data preprocessing module (4) receives control signal three, data preprocessing module (4) does not work；Data prediction Module (4) receives control signal for the moment, data preprocessing module (4) from database module (7) call multiple AFC historical datas and Multiple video historical datas, then data preprocessing module (4) respectively to AFC historical datas and video historical data at Reason；When data preprocessing module (4) is handled AFC historical datas, the average of multiple AFC historical datas is first calculatedAnd mark Accurate poor sA, obtain first threshold scopeThen judge each AFC historical datas whether in first threshold model In enclosing：AFC historical datas in the range of first threshold are designated as being worth, the AFC historical datas note not in the range of first threshold For bad value, if all AFC historical datas are preferably worth, data preprocessing module (4) is defeated to Federated Kalman Filtering module (5) Go outIf all AFC historical datas are bad value, data preprocessing module (4) by multiple AFC historical datas withDifference That minimum one is exported to Federated Kalman Filtering module (5)；If existing good value has bad value in multiple AFC historical datas again, Then by abnormal data erasing and recalculate the remaining average being worth wellThen willExport to Federated Kalman Filtering module (5)；When data preprocessing module (4) is handled video historical data, the average of multiple video historical datas is first calculated With standard deviation sB, obtain Second Threshold scopeThen judge each video historical data whether second In threshold range：Video historical data in the range of Second Threshold is designated as virtual value, the video not in the range of Second Threshold Historical data is designated as invalid value, if all AFC historical datas are virtual value, data preprocessing module (4) is to federal karr Graceful filtration module (5) outputIf all video historical datas are invalid value, data preprocessing module (4) regards multiple In frequency historical data withThat minimum one of difference is exported to Federated Kalman Filtering module (5)；If multiple video history numbers Existing virtual value has invalid value again in, then rejects invalid value and recalculate the average of remaining virtual valueThen willExport to Federated Kalman Filtering module (5)；In the output signal of data preprocessing module (4), output signal corresponding with AFC historical datas is designated as signal one, with regarding Output signal corresponding to frequency historical data is designated as signal two；When the Federated Kalman Filtering module (5) receives control signal two, to the signal one of data preprocessing module (4) output Make use processing with signal two, then export result to neural network prediction module (6)；The federal Kalman When filtration module (5) receives control signal four, Federated Kalman Filtering module (5) calls newest AFC from database module (7) Historical data and video historical data, then exchange the AFC historical datas used and video historical data makees use processing, Then result is exported to neural network prediction module (6)；After the neural network prediction module (6) receives the result, neural network prediction module (6) is from database module (7) multiple previous data are called, the result and multiple previous data are then subjected to passenger flow together as input vector Prediction processing is measured, passenger flow forecast data is obtained, then outwards exports passenger flow forecast data, passenger flow forecast data are outside While output, passenger flow forecast data are also sent database module (7) to preserve by neural network prediction module (6).
- 2. Urban Rail Transit Stations according to claim 1 passenger flow estimation system in short-term, it is characterised in that：The federation Kalman filtering module (5) includes part filter module one (5-1), part filter module two (5-2) and information fusion module (5- 3)；Described information Fusion Module (5-3) has two inputs and three output ends, three outputs of information fusion module (5-3) End is designated as the first feedback signal output, the second feedback signal output and main output end respectively；Part filter module one (5-1) Output end and two inputs of output end and information fusion module (5-3) of local filtration module two (5-2) correspond Ground connects, and first feedback signal output is connected with the feedback signal reception end of part filter module one (5-1), and described the Two feedback signal outputs are connected with the feedback signal reception end of part filter module two (5-2), the main output end and nerve Neural network forecast module (6) connects；When data preprocessing module (4) is to Federated Kalman Filtering module (5) output signal, data are pre- Processing module (4) exports signal one to part filter module one (5-1), data preprocessing module (4) by signal two export to Part filter module two (5-2)；Federated Kalman Filtering module (5) calls AFC historical datas and video from database module (7) During historical data, in corresponding AFC historical datas input part filter module one (5-1), the input of corresponding video historical data is local In filtration module two (5-2).
- 3. a kind of Urban Rail Transit Stations passenger flow forecasting in short-term, including Urban Rail Transit Stations passenger flow estimation system in short-term System, passenger flow estimation system counts equipment (2), data to the Urban Rail Transit Stations by AFC equipment (1), video passenger flow in short-term Screening module (3), data preprocessing module (4), Federated Kalman Filtering module (5), neural network prediction module (6) and data Library module (7) forms；The AFC equipment (1) and video passenger flow statistics equipment (2) are connected with data screening module (3)；Data screening module (3) it is connected respectively with data preprocessing module (4), Federated Kalman Filtering module (5) and database module (7)；Data are located in advance Reason module (4) is connected with Federated Kalman Filtering module (5) and database module (7) respectively；Federated Kalman Filtering module (5) It is connected respectively with neural network prediction module (6) and database module (7)；Database module (7) and neural network prediction module (6) connect；The AFC equipment (1) can be by the volume of the flow of passengers data counted on output in real time to data screening module (3), AFC equipment (1) The volume of the flow of passengers data of output are designated as AFC data；The video passenger flow statistics equipment (2) can export the volume of the flow of passengers data counted on to data screening module (3) in real time, depending on The volume of the flow of passengers data of frequency passenger flow statisticses equipment (2) output are designated as video data；Current date species can be identified for the data screening module (3), after identifying date species, the data sieve The AFC data received and video data transmission to database module (7) are carried out classification preservation by modeling block (3)；It is stored in data AFC data in library module (7) are designated as AFC historical datas, and the video data being stored in database module (7) is designated as video and gone through History data；Then, data screening module (3) is according to the date species identified, to data preprocessing module (4) and federal karr Graceful filtration module (5) is controlled：The date species includes normal day and special day, and working day and two-day weekend belong to normal Day, festivals or holidays and active day belong to special day；If the date species of current date is identified as normal day, data screening module (3) to data preprocessing module (4) output control signal one, and to Federated Kalman Filtering module (5) output control signal two； If the date species of current date is identified as special day, data screening module (3) is exported to data preprocessing module (4) and controlled Signal three processed, and to Federated Kalman Filtering module (5) output control signal four；When the data preprocessing module (4) receives control signal three, data preprocessing module (4) does not work；Data prediction Module (4) receives control signal for the moment, data preprocessing module (4) from database module (7) call multiple AFC historical datas and Multiple video historical datas, then data preprocessing module (4) respectively to AFC historical datas and video historical data at Reason；When data preprocessing module (4) is handled AFC historical datas, the average of multiple AFC historical datas is first calculatedAnd mark Accurate poor sA, obtain first threshold scopeThen judge each AFC historical datas whether in first threshold model In enclosing：AFC historical datas in the range of first threshold are designated as being worth, the AFC historical datas note not in the range of first threshold For bad value, if all AFC historical datas are preferably worth, data preprocessing module (4) is defeated to Federated Kalman Filtering module (5) Go outIf all AFC historical datas are bad value, data preprocessing module (4) by multiple AFC historical datas withDifference That minimum one is exported to Federated Kalman Filtering module (5)；If existing good value has bad value in multiple AFC historical datas again, Then by abnormal data erasing and recalculate the remaining average being worth wellThen willExport to Federated Kalman Filtering module (5)；When data preprocessing module (4) is handled video historical data, the average of multiple video historical datas is first calculatedWith Standard deviation sB, obtain Second Threshold scopeThen judge each video historical data whether in Second Threshold In the range of：Video historical data in the range of Second Threshold is designated as virtual value, the video history not in the range of Second Threshold Data are designated as invalid value, if all AFC historical datas are virtual value, data preprocessing module (4) is filtered to federal Kalman Ripple module (5) exportsIf all video historical datas are invalid value, data preprocessing module (4) goes through multiple videos In history data withThat minimum one of difference is exported to Federated Kalman Filtering module (5)；If in multiple video historical datas Existing virtual value has invalid value again, then rejects invalid value and recalculate the average of remaining virtual valueThen willIt is defeated Go out to Federated Kalman Filtering module (5)；In the output signal of data preprocessing module (4), output signal corresponding with AFC historical datas is designated as signal one, with regarding Output signal corresponding to frequency historical data is designated as signal two；When the Federated Kalman Filtering module (5) receives control signal two, to the signal one of data preprocessing module (4) output Make use processing with signal two, then export result to neural network prediction module (6)；The federal Kalman When filtration module (5) receives control signal four, Federated Kalman Filtering module (5) calls newest AFC from database module (7) Historical data and video historical data, then exchange the AFC historical datas used and video historical data makees use processing, Then result is exported to neural network prediction module (6)；After the neural network prediction module (6) receives the result, neural network prediction module (6) is from database module (7) multiple previous data are called, the result and multiple previous data are then subjected to passenger flow together as input vector Prediction processing is measured, passenger flow forecast data is obtained, then outwards exports passenger flow forecast data, passenger flow forecast data are outside While output, passenger flow forecast data are also sent database module (7) to preserve by neural network prediction module (6)；It is characterized in that：Passenger flow forecasting includes the Urban Rail Transit Stations in short-term：For Urban Rail Transit Stations in short-term in passenger flow estimation system operation, AFC equipment (1) and video passenger flow count equipment (2) AFC data and video data periodically are exported to data screening module (3)；Receive AFC data and video data every time Afterwards, Urban Rail Transit Stations in short-term passenger flow estimation system with regard to as follows carry out passenger flow forecast operation：1) current date species is identified data screening module (3), then passes the AFC data and video data that receive Transport to database module (7) and carry out classification preservation；Then：If the date species of current date is identified as normal day, data Screening module (3) is exported to Federated Kalman Filtering module (5) and controlled to data preprocessing module (4) output control signal one Signal two processed, into step 2)；If the date species of current date is identified as special day, data screening module (3) is to number Data preprocess module (4) output control signal three, and to Federated Kalman Filtering module (5) output control signal four, into step It is rapid 3)；2) data preprocessing module (4) calls multiple AFC historical datas and multiple video historical datas from database module (7), Then data preprocessing module (4) is handled AFC historical datas and video historical data respectively, generates corresponding signal one With signal two, into step 4)；3) Federated Kalman Filtering module (5) calls newest AFC historical datas and video history number from database module (7) According to, then exchange the AFC historical datas used and video historical data make use processing, then by result export to Neural network prediction module (6), into step 5)；4) signal one and signal two that Federated Kalman Filtering module (5) exports to data preprocessing module (4) make information fusion Processing, then exports result to neural network prediction module (6), into step 5)；5) neural network prediction module (6) calls multiple previous data from database module (7), then by the result and Multiple previous data carry out passenger flow forecast processing together as input vector, passenger flow forecast data are obtained, then by passenger flow Amount prediction data outwards exports, and while passenger flow forecast data outwards export, neural network prediction module (6) is also by the volume of the flow of passengers Prediction data send database module (7) to preserve.
- 4. Urban Rail Transit Stations according to claim 3 passenger flow forecasting in short-term, it is characterised in that：The federation Kalman filtering module (5) includes part filter module one (5-1), part filter module two (5-2) and information fusion module (5- 3)；Described information Fusion Module (5-3) has two inputs and three output ends, three outputs of information fusion module (5-3) End is designated as the first feedback signal output, the second feedback signal output and main output end respectively；Part filter module one (5-1) Output end and two inputs of output end and information fusion module (5-3) of local filtration module two (5-2) correspond Ground connects, and first feedback signal output is connected with the feedback signal reception end of part filter module one (5-1), and described the Two feedback signal outputs are connected with the feedback signal reception end of part filter module two (5-2), the main output end and nerve Neural network forecast module (6) connects；When data preprocessing module (4) is to Federated Kalman Filtering module (5) output signal, data are pre- Processing module (4) exports signal one to part filter module one (5-1), data preprocessing module (4) by signal two export to Part filter module two (5-2)；Federated Kalman Filtering module (5) calls AFC historical datas and video from database module (7) During historical data, in corresponding AFC historical datas input part filter module one (5-1), the input of corresponding video historical data is local In filtration module two (5-2)；The Federated Kalman Filtering module (5) carries out Federated Kalman Filtering processing as follows：(1) part filter module one (5-1) is handled signal one using linear Kalman filter method, obtains filter result One, filter result one is exported to information fusion module (5-3), meanwhile, part filter module two (5-2) uses linear Kalman Filtering method is handled signal two, obtains filter result two, and filter result two is exported to information fusion module (5-3)；(2) information fusion module (5-3) carries out use processing to filter result one and filter result two, then ties processing Fruit is exported to neural network prediction module (6)；Information fusion module make the blending algorithm that is used during use processing for：<mrow> <msub> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>g</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>=</mo> <msub> <mi>P</mi> <mi>g</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>&lsqb;</mo> <msubsup> <mi>P</mi> <mi>Z</mi> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <msub> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>Z</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>+</mo> <msubsup> <mi>P</mi> <mi>V</mi> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <msub> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>V</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>&rsqb;</mo> </mrow><mrow> <msub> <mi>P</mi> <mi>g</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>=</mo> <msup> <mrow> <mo>&lsqb;</mo> <msubsup> <mi>P</mi> <mi>Z</mi> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>+</mo> <msubsup> <mi>P</mi> <mi>V</mi> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>&rsqb;</mo> </mrow> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msup> </mrow>Wherein, k represents step number,For global best estimates value, Pg(k) it is the variance of global best estimates value,For filter Estimate corresponding to ripple result one, PZ(k) variance corresponding to filter result one,Represent to PZ(k) invert,For the estimate corresponding to filter result two, PV(k) variance corresponding to filter result two,Represent to PV(k) Invert；Information fusion module (5-3) willWithFeed back to part filter module one (5-1), information fusion module (5-3) willWithFeed back to part filter module two (5-2)；βZ(k) part filter is corresponded to when being walked for kth The distribution coefficient of module one (5-1), βV(k) part filter module two (5-2) distribution coefficient, β are corresponded to when being walked for kthZ(k)+βV (k)=1,Represent to βZ(k) invert,Represent to βV(k) invert.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710706219.7A CN107480824B (en) | 2017-08-17 | 2017-08-17 | Short-time passenger flow prediction system and method for urban rail transit station |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710706219.7A CN107480824B (en) | 2017-08-17 | 2017-08-17 | Short-time passenger flow prediction system and method for urban rail transit station |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107480824A true CN107480824A (en) | 2017-12-15 |
| CN107480824B CN107480824B (en) | 2020-06-05 |
Family
ID=60600739
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710706219.7A Active CN107480824B (en) | 2017-08-17 | 2017-08-17 | Short-time passenger flow prediction system and method for urban rail transit station |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107480824B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108694463A (en) * | 2018-04-25 | 2018-10-23 | 东南大学 | A kind of Urban Rail Transit Stations passenger flow forecasting out of the station |
| CN112434859A (en) * | 2020-11-26 | 2021-03-02 | 国电南瑞科技股份有限公司 | Rail transit underground station environment-friendly regulation method combined with passenger flow prediction technology |
| CN113822462A (en) * | 2021-08-06 | 2021-12-21 | 上海申铁信息工程有限公司 | Station emergency command method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004023114A (en) * | 2002-06-12 | 2004-01-22 | Nippon Telegr & Teleph Corp <Ntt> | Method and system for predicting communication traffic |
| CN103632212A (en) * | 2013-12-11 | 2014-03-12 | 北京交通大学 | System and method for predicating time-varying user dynamic equilibrium network-evolved passenger flow |
| CN105224999A (en) * | 2015-09-10 | 2016-01-06 | 北京市交通信息中心 | The real-time passenger flow forecasting of urban track traffic based on AFC data and system |
| CN107067076A (en) * | 2017-05-27 | 2017-08-18 | 重庆大学 | A kind of passenger flow forecasting based on time lag NARX neutral nets |
-
2017
- 2017-08-17 CN CN201710706219.7A patent/CN107480824B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004023114A (en) * | 2002-06-12 | 2004-01-22 | Nippon Telegr & Teleph Corp <Ntt> | Method and system for predicting communication traffic |
| CN103632212A (en) * | 2013-12-11 | 2014-03-12 | 北京交通大学 | System and method for predicating time-varying user dynamic equilibrium network-evolved passenger flow |
| CN105224999A (en) * | 2015-09-10 | 2016-01-06 | 北京市交通信息中心 | The real-time passenger flow forecasting of urban track traffic based on AFC data and system |
| CN107067076A (en) * | 2017-05-27 | 2017-08-18 | 重庆大学 | A kind of passenger flow forecasting based on time lag NARX neutral nets |
Non-Patent Citations (1)
| Title |
|---|
| 史文雯: "城市轨道交通短时客流预测与最优客运能力调配问题的研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108694463A (en) * | 2018-04-25 | 2018-10-23 | 东南大学 | A kind of Urban Rail Transit Stations passenger flow forecasting out of the station |
| CN108694463B (en) * | 2018-04-25 | 2021-01-05 | 东南大学 | Urban rail transit station entrance and exit passenger flow prediction method |
| CN112434859A (en) * | 2020-11-26 | 2021-03-02 | 国电南瑞科技股份有限公司 | Rail transit underground station environment-friendly regulation method combined with passenger flow prediction technology |
| CN112434859B (en) * | 2020-11-26 | 2022-08-26 | 国电南瑞科技股份有限公司 | Rail transit underground station environment-friendly regulation method combined with passenger flow prediction technology |
| CN113822462A (en) * | 2021-08-06 | 2021-12-21 | 上海申铁信息工程有限公司 | Station emergency command method and device |
| CN113822462B (en) * | 2021-08-06 | 2023-10-13 | 上海申铁信息工程有限公司 | Station emergency command method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107480824B (en) | 2020-06-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107480824A (en) | Urban Rail Transit Stations passenger flow estimation system and method in short-term | |
| CN107330547A (en) | A kind of city bus dynamic dispatching optimization method and system | |
| CN104517040B (en) | One kind is based on IC-card data public transit vehicle in-car degree of crowding computational methods | |
| CN108230698A (en) | traffic flow forecasting method and device | |
| CN107633674A (en) | A kind of emphasis commerial vehicle exception tracing point elimination method and system | |
| CN112288374B (en) | Intelligent supply chain logistics intelligent optimization management system based on cloud computing | |
| CN107270921A (en) | A kind of generation dimension polling path method and device for planning | |
| CN108108897B (en) | Rail transit passenger flow clearing method and system and electronic equipment | |
| CN105592487A (en) | Evaluation method and apparatus for LTE network service flow | |
| CN106373399B (en) | Bottleneck identification system is run between a kind of public bus network website | |
| CN103888961B (en) | A kind of user monitoring method and device | |
| CN103809586A (en) | Fault analyzing and diagnosing system and method of railway dispatching orders | |
| CN109936813A (en) | Recognition methods, device, equipment and the medium of high-speed railway user | |
| CN107730081A (en) | A kind of data acquisition analysis method and system based on RFID technique | |
| CN116090836A (en) | Global risk determination method, device, equipment and medium applied to rail transit | |
| CN105679028A (en) | Bus route setting reasonability and bus stop efficiency data analysis method | |
| CN111016976B (en) | Train operation deviation propagation condition identification method based on multilayer coupling relation | |
| EP1972894A1 (en) | Method, computer based device and system for determining a route | |
| CN106683029A (en) | Evaluation method and system for first and last bus on-time rate | |
| CN104766474A (en) | Urban comprehensive transportation junction transfer passenger flow volume detecting method based on mobile phone terminals | |
| Yang et al. | A human trajectory estimate based on individual mobility pattern library | |
| CN116933958B (en) | Subway transfer station transfer passenger flow estimation method based on card swiping data | |
| Nyström | Punctuality and railway maintenance | |
| CN113378338B (en) | Modeling method of high-speed railway networked train operation scheme diagram | |
| EP4343731A1 (en) | Method and system for delay prediction for scheduled public transport using multi architectural deep learning |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: No. 1, Fuxing Avenue, Shuang Fu town, Jiangjin District, Chongqing Applicant after: chongqing jiaotong university Address before: 400074 Chongqing Nan'an District University Avenue, No. 66 Applicant before: chongqing jiaotong university |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201230 Address after: 251700 Pengjia village, Jianglou Town, Huimin County, Binzhou City, Shandong Province (West of Jiechen Road) Patentee after: Binzhou Xuanhua rope net Co.,Ltd. Address before: No. 1, Fuxing Avenue, Shuang Fu town, Jiangjin District, Chongqing Patentee before: CHONGQING JIAOTONG University |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210430 Address after: 102200 423, 4 / F, block a, Xinhua future city building, 175 Litang Road, Changping District, Beijing Patentee after: Li Qiannan Address before: 251700 Pengjia village, Jianglou Town, Huimin County, Binzhou City, Shandong Province (West of Jiechen Road) Patentee before: Binzhou Xuanhua rope net Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221228 Address after: No. 422, Siming South Road, Xiamen, Fujian, 361001 Patentee after: XIAMEN University Patentee after: Wanchong Construction Co.,Ltd. Patentee after: THE NO.3 ENGINEERING CO.,LTD. OF CHINA RAILWAY 22TH BUREAU Group Patentee after: Xiamen Zhuoyi Construction Engineering Co.,Ltd. Address before: 102200 423, 4 / F, block a, Xinhua future city building, 175 Litang Road, Changping District, Beijing Patentee before: Li Qiannan |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230103 Address after: No. 422, Siming South Road, Xiamen, Fujian, 361001 Patentee after: XIAMEN University Patentee after: THE NO.3 ENGINEERING CO.,LTD. OF CHINA RAILWAY 22TH BUREAU Group Patentee after: HAIHUAN TECHNOLOGY GROUP CO.,LTD. Patentee after: Xiamen Zhuoyi Construction Engineering Co.,Ltd. Address before: No. 422, Siming South Road, Xiamen, Fujian, 361001 Patentee before: XIAMEN University Patentee before: Wanchong Construction Co.,Ltd. Patentee before: THE NO.3 ENGINEERING CO.,LTD. OF CHINA RAILWAY 22TH BUREAU Group Patentee before: Xiamen Zhuoyi Construction Engineering Co.,Ltd. |