Urban rail transit network multi-level passenger flow state estimation system

Abstract

The invention belongs to the field of passenger flow simulation of an urban rail transit network, and particularly relates to a multi-level passenger flow state estimation system of the urban rail transit network. The method comprises the following steps: defining a trip chain of passengers by acquiring, analyzing and processing related data of urban rail transit and mobile phone signaling; random stratified sampling is carried out according to a proportion by using real-time OD passenger flow estimated by a BP neural network to obtain an OD path passenger flow distribution result, a passenger trip chain reconstruction algorithm model is established, the time similarity of a passenger trip sub-chain and a train operation sub-chain on a feasible path between different path OD pairs is calculated, passengers are matched to a train, and a complete trip chain of the passengers is reconstructed; and a multi-level passenger flow state index system of the urban rail transit system is constructed, so that the system can output microscopic, mesoscopic and macroscopic passenger flow indexes of the urban rail transit system in real time.

Description

Technical Field

[0001] The invention belongs to the field of passenger flow simulation of urban rail transit network, and in particular relates to a multi-level passenger flow state estimation system of an urban rail transit network. Background Art

[0002] As the backbone network in the comprehensive urban public transportation network, rail transit has always been an important means of travel to alleviate urban traffic congestion and is also the focus of urban transportation development. In recent years, rail transit in many cities has entered the stage of network operation. In order to better understand the passenger flow operation status of the network, optimize the driving organization, and accurately clear the ticketing and daily operation management, it is necessary to clarify the real-time passenger capacity and cross-sectional passenger flow on the trains in the network operation. In order to estimate the passenger flow status of the urban rail transit line network in real time, it is necessary to reconstruct the travel trajectory of passengers in the urban rail network, locate the trains that passengers take in the network, and locate the time and space coordinates of passengers getting on and off the train when the train arrives at the station. By constructing a network simulation model, according to the matching of the transaction data of the automatic ticket vending and checking (AFC) system and the train operation timetable, the passenger travel chain is calculated, and then the number of passengers on the trains issued on different lines in the network is counted according to the passenger boarding and alighting nodes in the passenger travel chain, and then the cross-sectional passenger flow of the network is estimated based on the real-time passenger capacity of the train. In the existing research at home and abroad, most of the estimation of passenger travel characteristics is done through manual surveys, field surveys, or calculations using station layouts and assumed travel speeds. Manual surveys and field surveys consume a lot of manpower and material resources and can only partially sample. The method of assuming passenger travel speeds will reduce the accuracy of the estimation results due to the deviation between the assumption and the actual situation due to differences in passenger age and gender. For the estimation of the probability of OD to feasible path selection, most of the existing methods only use AFC transaction data and train timetables to build a probability model or calibrate passenger travel preference parameters through traffic surveys, and fail to make full use of multi-source data in the network for analysis. For the reconstruction of passenger travel chains, existing methods fail to consider passengers who are not fully OD travelers, and cannot perform real-time estimation. Regarding the estimation of train passenger capacity, most of the existing research at home and abroad uses manual surveys, perception, and other methods to estimate the probability of OD to feasible path selection. Equipment or weighing system is used to count the number of passengers. The manual survey method consumes a lot of manpower and material resources and can only partially sample. The deployment of sensing equipment such as infrared radar or camera will cost a lot of money. The weighing system can only obtain the weight of passengers on the train, and the different weights of passengers or the luggage they carry will also reduce the accuracy of the estimation results. As for the estimation of the passenger flow of the line network section, the existing methods cannot infer the travel path of passengers due to the above reasons, nor can they accurately estimate the passenger flow of the section under a given time and space state; similarly, the current existing methods cannot accurately estimate the number of passengers waiting on the platform; the number of passengers in the transfer channel is usually counted by video recognition or manual survey. The former needs to call a large amount of video data to identify the direction of passengers' travel and then count them. The manual survey method requires a lot of manpower and material resources and the survey results need to be repeatedly updated according to network expansion or passenger flow changes. Summary of the invention

[0003] Purpose of the invention: In view of the above problems, the present invention proposes a multi-level passenger flow state estimation system for urban rail transit network, which can provide a basis for urban rail transit network simulation, timetable optimization, ticket clearing, real-time train passenger capacity and cross-sectional passenger flow calculation, and emergency response to sudden incidents.

[0004] The present invention adopts the following technical solution:

[0005] A multi-level passenger flow state estimation system for an urban rail transit network comprises the following steps:

[0006] S1, obtain and analyze the basic data of urban rail transit network stations and lines, passenger entry and exit transaction data recorded by the AFC system, train operation timetable data, basic information data of signaling base stations, and mobile phone signaling data; clean the input data and obtain the specified data format;

[0007] S2, construct a spatiotemporal topological model of the road network and a constrained breadth-first algorithm to search for a feasible path set between OD (Origin-Destination: starting point, destination) pairs. According to the train operation schedule information loaded in the road network topological model, a spatiotemporal topological model of the rail transit road network is constructed, and a constrained breadth-first algorithm is used to search for a feasible path set between OD pairs that meets the maximum number of transfer constraints;

[0008] S3, construct a passenger route selection probability estimation model based on mobile phone signaling data, calculate the initial probability of route selection with the mobile phone user travel trajectory reconstructed from the mobile phone signaling data, and define the passenger's travel chain;

[0009] The passenger entry and exit time and transfer time distribution are analyzed by using the travel data of single-path without transfer and single-path with transfer, so as to obtain the travel time distribution of feasible paths. Then, the travel data distribution between OD pairs is estimated by using the mixed Gaussian distribution theory, and then a nonlinear optimization model is constructed with the KL divergence of the estimated and observed travel time distribution between OD pairs as the optimization target. The path selection probability of all passengers in urban rail transit is estimated by expanding the sample using the path selection probability of mobile phone users.

[0010] S4, using BP neural network to estimate the OD passenger flow of urban rail transit network, using monthly historical AFC transaction data as training data, training BP neural network model, and using the trained BP neural network to estimate the real-time OD passenger flow of the network;

[0011] Combined with the feasible path selection probability obtained in step S3, the real-time OD passenger flow estimated by BP neural network is randomly sampled in proportion to obtain the OD path passenger flow distribution result, and the passenger exit time is estimated according to the passenger entry time and the path travel time distribution estimated in S3, and the real-time estimated passenger exit time between OD pairs is obtained;

[0012] S5, build a passenger travel chain reconstruction algorithm model, match the passenger flow on the OD feasible path with the train, and locate the spatiotemporal information of passengers getting on and off the train during the trip;

[0013] Taking the minimum entry walking time and transfer walking time of passengers inferred from the entry and exit time of passengers traveling on a single path without transfer and a single path with a single transfer in S3 as constraints, for a certain station, the travel chain of passengers traveling on a single path without transfer at the station is reconstructed, and then the entry time is calculated based on the boarding time of passengers in the travel chain. The minimum value of this value among this group of passengers is the minimum entry walking time; the minimum transfer walking time is the transfer time calculated based on the travel chain reconstructed by passengers traveling on a single path with a single transfer in a certain transfer direction passing through the station, and the minimum value of the transfer time is the minimum transfer walking time;

[0014] By using the real-time estimated passenger exit time between OD pairs in S4, the time similarity between the passenger travel subchain and the train operation subchain on the feasible paths between different OD pairs is calculated, and the passengers are matched to the trains, thereby reconstructing the complete travel chain of the passengers and realizing the spatial-temporal location query function of the passengers in the urban rail transit system.

[0015] The different paths mentioned above refer to: feasible paths between OD pairs.

[0016] S6, according to the matching of the nodes where passengers get on and off the train in the complete travel chain of the passengers with the train timetable, the number of passengers getting on and off the train at the station is counted, and the real-time train passenger capacity is obtained by using the train passenger capacity statistical model, thereby obtaining the real-time train passenger capacity of the urban rail transit line network;

[0017] S7, based on the real-time passenger capacity of the train inferred in S6, estimate the real-time passenger flow of the line section at different time granularities within the analysis time period to obtain the real-time passenger flow of the urban rail transit line network;

[0018] According to the real-time passenger volume of trains under different time states during the analysis period inferred in S6;

[0019] The analysis time period can be one day or the morning and evening peaks, and the time status is a specific time range, such as: half an hour, 15 minutes or 5 minutes.

[0020] S8. According to the matching of the nodes where passengers get on and off the train in the complete travel chain of the passengers with the train timetable, the number of passengers getting on the train at the station is calculated, and the number of passengers waiting on the platform is obtained by using the estimation model of the number of passengers waiting on the platform, and the number of passengers waiting on the platform at the urban rail transit line is obtained;

[0021] S9, construct a real-time estimation model for the number of passengers in the transfer channel, and infer the number of passengers in the transfer channel according to the nodes where passengers get on and off the bus in the complete travel chain of the passengers and the average walking time of the passengers inferred in S3, so as to obtain the real-time number of passengers in the transfer channel of the urban rail transit network;

[0022] S10. Construct a multi-level passenger flow status indicator system for urban rail transit systems, including micro, meso and macro passenger flow indicators.

[0023] Micro passenger flow status indicators include passenger boarding and alighting time, location and arrival time at the platform, and individual passenger walking time or waiting time;

[0024] Meso-level passenger flow indicators include the real-time passenger capacity of trains, the real-time number of people waiting for trains on platforms, and the number of people in transfer passages;

[0025] Macro passenger flow indicators include real-time route flow;

[0026] Based on steps S1-S9, the system can output micro, meso and macro passenger flow indicators of the urban rail transit system in real time.

[0027] The multi-level passenger flow state estimation system for urban rail transit network of the present invention obtains basic data of rail transit network stations and lines in S1, which is as follows:

[0028] S1.1, the urban rail transit network station data includes four attribute values: station number, station longitude, station latitude and the line to which it belongs, expressed as:

[0029] VS={[v1,Lon1,Lat 1, l1],…,[v i ,Lon i ,Lat i ,l j ]}

[0030] VS represents the basic data of network sites and lines, i represents the site number, and v i represents the site number, Lon represents the longitude of the site, Lat represents the latitude of the site, k represents the line number, l j Indicates the line number to which the station belongs;

[0031] N = {φ(L,V)} represents the spatial model of urban rail transit network.

[0032] N represents the road network, L represents the set of lines, V represents the set of stations, and the lines and stations form a network through a nonlinear relationship φ;

[0033] The basic properties of the network are derived from the logical relationship between lines, stations, and network in the network N = {φ(L,V)}:

[0034] At the line level, line l j The site collection is represented as in, Indicates line l j The i-th station in, i=1,2,…,n, n is the line l j The number of sites in

[0035] At the network level, the line set is represented by L = {l1,l2,…,l m}, where l j ,j=1,2,…,m represents a certain line in the road network, and m is the total number of lines in the road network;

[0036] The set of all sites in the road network is represented as

[0037] Using φ(L,V) to represent the spatial relationship between stations, routes, and road networks, the road network spatial model is expressed as follows:

[0038] N={φ(L,V)}

[0039] In the model, line l j Contains two attribute values: line name and line rail transit clearing system ACC number; station Contains five attribute values: station name, line number, station ACC number, station line number, and whether it is a transfer station;

[0040] S1.2, obtain AFC transaction data. The transaction data recorded by the urban rail transit automatic ticket vending and checking system includes the passenger's travel starting and ending points and entry and exit times. The AFC transaction data of a complete trip of passenger x is expressed as:

[0041] Trade x ={ID x ,v o ,ta,v d ,te}

[0042] Where, ID x represents the transaction number of passenger x recorded in the automatic ticket vending and checking system, v o ,v d They represent the starting and ending stations of passengers in the urban rail transit system, and ta and te represent the time when passengers pass through the entry and exit gates respectively;

[0043] S1.3, obtain train timetable data. The train operation timetable data includes the arrival and departure time information of the operating trains on the line, which is expressed as:

[0044] T(L,TR)

[0045] In the formula, L is the set of lines in the subway network, TR is the set of trains in the subway network;

[0046] TR={tr1,tr2,…,tr k ,…,tr K}, k = 1, 2, ..., K, K is the total number of departures in the road network;

[0047] Line j The set of trains dispatched from Indicates line l j The g-th train departs from the line l. j The number of departures on

[0048] S1.4, obtain the basic information data of the signaling base station, the steps are as follows:

[0049] The basic information data of the signaling base station includes the base station number, base station longitude, base station latitude and whether it belongs to the subway system, which is expressed as:

[0050] BS={[BSID1,Lon1,Lat 1, 1],…,[BSID i ,Lon i ,Lat i ,0]}

[0051] Among them, BSID i Indicates the mobile base station code, Lon i Indicates the longitude of the base station, Lat i Indicates the latitude of the base station, i = 1, 2, ..., n, n is the number of base stations, 0 means the base station does not belong to the subway network, 1 means the base station belongs to the subway network;

[0052] S1.5, obtain mobile phone signaling data, the steps are as follows:

[0053] Mobile phone signaling data includes mobile station identification code, timestamp, mobile base station code and signaling event code. According to mobile phone signaling data, the continuous spatiotemporal position of passengers in the network is obtained. The mobile phone signaling data collected by passenger x in a trip in the subway network is expressed as:

[0054] MS x ={[MSID x ,t1,BSID1,EVID1],…,[MSID x ,t i ,BSID i ,EVID i ]}

[0055] Among them, MSID x Indicates the mobile station identification code MSID, ti Indicates timestamp, EVID i Indicates the signaling event code.

[0056] In the urban rail transit network multi-level passenger flow state estimation system of the present invention, the construction of the rail transit network spatiotemporal topology model in S2 specifically includes:

[0057] S2.1, construct a road network spatial topology model to search for the shortest path between OD pairs in the road network;

[0058] According to the logical relationship between stations and lines in the road network N = {φ(L,V)}, the rail transit network is represented as a directed graph G(V,A).

[0059] Where V is the set of stations in the rail transit network, A is the set of directed edges in the rail transit network, and the directed edges represent the train track areas between stations;

[0060] Construct the adjacency matrix M of G(V,A) c and the distance matrix M d ,

[0061]

[0062] In the formula, the adjacency matrix M c The matrix element c ij =1 indicates that sites i and j are directly connected, otherwise, they are not directly connected, which is represented by 0; the distance matrix M d The matrix element d ij Represents the distance of the directed edge from site i to site j. If two sites are directly connected, the edge distance between adjacent sites is the distance a between the sites. ij , otherwise, it is +∞;

[0063] The logical relationship between stations and routes in the road network and the directed graph of the road network together constitute the road network spatial topology model:

[0064] O={φ(L,V),G(V,A)}

[0065] Wherein, O represents the spatial topological model of the road network, in the network directed graph G(V,A), v∈V represents the nodes of the graph, and a∈A represents the edges of the graph. There are edges in the uplink and downlink directions between adjacent stations, and virtual platforms are set up at the transfer station to describe the transfer behavior of passengers in the transfer station.

[0066] S2.2, construct the network train operation time information model T (L, V, TR), which includes the arrival and departure time information of trains running on different lines. The network time dimension information is extracted from the train timetable. The train arrival and departure information in the train operation timetable is expressed as:

[0067]

[0068] In the formula, Respectively represent the arrival time of trains in the up and down directions, Respectively represent the departure time of trains in the up and down directions,

[0069] Respectively represent line l j The i-th site The kth vehicle The arrival time of up and down trains,

[0070] Respectively represent line l j The i-th site The kth vehicle The departure time of up and down trains, m, n, q are the number of lines in the network, the number of stations in the line, and the number of train departures on the line respectively;

[0071] S2.3, load the road network train operation time information model T(L,V,TR) into the road network space directed graph constructed in S2.1, and obtain the road network spatiotemporal topology model as follows:

[0072] H={O={φ(L,V),G(V,A)},T(L,V,TR)}

[0073] Where H represents the road network spatiotemporal topology model, O = {φ(L,V), G(V,A)} represents the spatial topology model constructed in S2.1, and T(L,V,TR) represents the road network train operation time information model loaded into the spatiotemporal topology model;

[0074] T(L,V,TR) includes the arrival and departure time of the train on the line, which can be expressed as:

[0075]

[0076]

[0077] In the formula, the constraint condition is equivalent to the logical relationship between stations and lines in the road network N = {φ(L,V)} in S2.1;

[0078] In a spatial directed graph, a constrained breadth-first algorithm is used to search for a set of feasible paths between OD pairs in the road network that satisfy the maximum number of transfers.

[0079] In the multi-level passenger flow state estimation system for urban rail transit network described in the present invention, the initial probability of path selection is calculated by using the mobile phone user travel trajectory reconstructed by mobile phone signaling data in S3, and the specific steps are as follows:

[0080] S3.1. According to the base station location and time information in the passenger's mobile phone signaling data, query the longitude and latitude of the passenger's mobile phone signaling data, match it with the longitude and latitude of the subway station, and reconstruct the passenger's travel trajectory in the subway network. The steps are as follows:

[0081] S3.1A, read passenger x mobile phone signaling data MS x The mobile base station number BSID corresponding to each signaling event;

[0082] S3.1B, read the latitude and longitude data of the base station BSID from the basic information data of the mobile base station, determine whether it belongs to the subway network dedicated base station, obtain the passenger's location and determine whether it is in the subway network, and reconstruct the passenger's travel trajectory in the subway network,

[0083] S3.1C, the travel trajectory of passenger x in the subway network based on signaling reconstruction is expressed as:

[0084] J x :{(v o ,t1),…,(v i ,t i ),…,(v d ,t n )}

[0085] Among them, v i represents the number of the ith station in the travel trajectory of passenger x, t i Indicates that after v i time;

[0086] S3.2. Matching the travel trajectory of mobile phone users with feasible paths between OD pairs of urban rail transit network, specifically including:

[0087] The spatial similarity between the passenger's mobile phone signaling reconstruction trajectory and the feasible path on the corresponding OD pair is defined as follows:

[0088]

[0089] In the formula, represents the trajectory p reconstructed by the mobile phone signaling of passenger x x The feasible path p with OD w The spatial similarity of For p x With p w The number of common sites, Indicates p x The number of sites, Indicates p w The number of sites, γ is a positive constant;

[0090] Calculate the spatial similarity between the passenger travel trajectory reconstructed based on mobile phone signaling and all feasible paths between OD pairs, and take the feasible path with the highest similarity to the reconstructed trajectory as the passenger's travel trajectory in the subway network;

[0091] S3.3. Based on the travel trajectories of the above passengers in the subway network, the number of specific mobile phone users on the feasible paths of the OD pairs is counted, and the initial probability of feasible path selection between the OD pairs is calculated as follows:

[0092]

[0093] Among them, r j represents the feasible path p for OD to w j The initial probability of path selection, n j Select a feasible path p for a specific mobile phone user passenger on OD pair w j The number of people, n w is the number of passengers with specific mobile phone users on OD pair w;

[0094] In S3, the passenger's travel chain is represented as:

[0095]

[0096] In the formula, λ x represents the passenger travel chain, x represents a passenger entering the urban rail transit system, ta represents the transaction time when the passenger passes through the entry gate, and te represents the transaction time when the passenger passes through the exit gate; x v (tr) represents the time when passenger x boarded train tr at station v, tr(x v ) represents the time when passenger x gets off the train tr at station v;

[0097] Based on the defined passenger travel chain, passengers are matched to trains through the acquired AFC data and train schedule data, specifically:

[0098] S3.4, according to the obtained train schedule, define the train operation chain of path p:

[0099]

[0100] In the formula, λ p represents the train operation chain of path p, are the jth travel subchain of the train operation chain, and the travel subchain includes the starting point of the train on section j Departure time on and end point Arrival time on The train operation chain without transfer path contains only one train operation sub-chain, while the train operation chain with transfer path contains multiple train operation sub-chains;

[0101] S3.5, define the similarity between the passenger travel time interval and the train operation time interval on the path, expressed as follows:

[0102]

[0103] In the formula, represents the time similarity between the travel chain of passenger x and the travel chain of train tr on path p, is the length of the intersection of the passenger travel time interval and the train operation time interval on segment j of path p, represents the travel time interval length of passenger x on segment j of path p, represents the length of the running time interval of train tr on section j of path p, and γ is a positive constant;

[0104] S3.6, based on the passenger transaction data recorded by the automatic ticket vending and checking system x , extract the passenger's entry and exit OD information, and then search the feasible short path set P of the road network w , passengers are classified according to the feasible path set and the number of transfers based on OD, and passenger travel is divided into single path without transfer, single path with single transfer, single path with multiple transfers, and multi-path travel according to the number of transfers on the path;

[0105] S3.7, for passengers on a single route without transfer, the trains taken by the passengers are matched using time similarity;

[0106] The passenger's travel time interval is determined by the transaction data Trade x The transaction time of the passenger entering and exiting the station is obtained from the train timetable information T(L,TR). The similarity between the running time interval of the train tr on path p after the passenger enters the station and the travel time interval of passenger x is calculated, and the passenger is matched to the time interval similarity. The highest train goes up;

[0107] S3.8, based on the non-transfer single-path passenger matching results obtained in S3.7, calculate the time from the passenger entering the station through the gate to boarding the train and the time from the passenger getting off the train to exiting the station through the gate, expressed as:

[0108]

[0109] In the formula, ACT represents the time from when the passenger enters the station through the gate to when he / she boards the train tr, and x v (tr) is the time when the passenger boards the train tr at station v, ta is the transaction time when the passenger passes through the entrance gate; EGT is the time when the passenger gets off the train and passes through the gate to exit the station, tr(x v) is the time when the passenger gets off the train tr at station v, and te is the transaction time when the passenger passes through the exit gate;

[0110] Assume that the passenger entry and exit time distribution of a single path without transfer at station v represents the passenger entry and exit time distribution at station v, and calculate the Gaussian distribution of passenger entry and exit time;

[0111] S3.9, to match single-transfer single-path passengers to trains, the following relationship needs to be satisfied between the passenger travel chain and the train operation chain:

[0112]

[0113] In the formula, t d (l,tr,v) represents the time when train tr on line l leaves station v, t a (l,tr,v) represents the time when train tr running on line l arrives at station v, ACT min (l,v) represents the minimum entry time of a passenger at station v on line l, EGT min (l,v) represents the minimum exit time of passengers at station v on line l, t d (l ′ ,tr ′ ,v) indicates the transfer line l ′ Transfer train running on tr ′ Time to leave site v, TFT min (l,l ′ ,v) indicates that at station v, passengers transfer from line l to line l ′ Minimum transfer time;

[0114] Match single-transfer single-path passengers to trains, and analyze the transfer time between transfer stations and lines in different transfer directions; obtain the ACT by calculating the entry and exit time distribution of station passengers through S3.8 min (l,v) and EGT min (l,v);

[0115] By analyzing the passenger's AFC transaction data and train schedule, and substituting them into constraint (2) in S3.9, the train taken by the passenger in the last segment of the single-transfer single-path travel sub-chain is found, and the boarding time of the passenger in the last segment of the travel sub-chain is obtained;

[0116] Taking the passenger boarding time of the last travel subchain as the estimated end time of the first travel subchain, taking ta as the estimated start time of the first travel subchain, calculate the time similarity between the estimated time interval of the passenger's first travel subchain and the train operation chain of the section where the passenger's first travel subchain is located;

[0117] Match the passenger to the train with the highest time similarity, and obtain the passenger's first travel subchain, thus obtaining the complete travel chain of the single-transfer passenger;

[0118] The transfer time of passengers is calculated as follows:

[0119]

[0120] Where TFT represents the passenger transfer time, Indicates that the passenger boards line l at transfer station v ′ The nth train time, Indicates that the passenger transfers from the i-th train on line l at the transfer station v The time to get off the bus; assuming that the transfer time distribution of single-transfer passengers at the station represents the transfer time distribution of transfer passengers, the transfer time distribution of transfer passengers at the station is obtained by calculating the transfer time of single-transfer passengers in each transfer direction at the station.

[0121] The multi-level passenger flow state estimation system for urban rail transit network of the present invention estimates the walking time of passengers entering and exiting the station and the walking time of transfer in S3, and calculates the mean time headway of trains at the station:

[0122]

[0123] in, For line l station v direction The average headway of trains, Respectively represent the running direction of line l station v The departure time of the k+1th and kth trains, K is the number of train departures; the average waiting time of passengers on the platform is half of the headway time, that is,

[0124]

[0125] in, Indicates line l station v direction The average waiting time of passengers is Indicates line l station v direction Average headway of trains;

[0126] Assume that the standard deviation of passenger waiting time is one-fourth of the standard deviation of train headway time, that is,

[0127]

[0128] in, For line l station v direction The standard deviation of the waiting time of passengers, For line l station v direction The headway of the i-th train, For line l station v direction Average headway of trains on

[0129] By estimating the distribution of passengers' entry and exit time, the distribution of passengers' walking time is obtained as follows:

[0130]

[0131] in, represents the distribution of passengers’ walking time into the station, N(μ ACT ,σ ACT ) represents the distribution of passengers’ arrival time, It represents the waiting time distribution of passengers at the platform after entering the station. The relationship between the mean and standard deviation is:

[0132]

[0133] Calculate the distribution of passengers' transfer walking time, specifically:

[0134]

[0135] in, represents the distribution of passengers' transfer walking time, represents the distribution of passengers’ transfer time,

[0136] represents the waiting time distribution of passengers at the transfer station, and the relationship between the mean and standard deviation is:

[0137]

[0138] Assuming that passengers get off at their destination and immediately walk to the exit gate, the passenger's exit walking time distribution is the exit time distribution.

[0139]

[0140] So far, we have obtained the distribution of passengers' walking time in and out of the urban rail transit system, the distribution of transfer walking time in the transfer direction of each transfer station, and the distribution of passengers' waiting time at each station.

[0141] In the multi-level passenger flow state estimation system for urban rail transit network described in the present invention, the feasible path travel time is estimated in S3, specifically,

[0142] Through the additivity of independent Gaussian distribution, the travel time distribution of the feasible path is obtained as the sum of the passenger's walking time in and out of the station, transfer walking time, waiting time, and time in the car:

[0143]

[0144] in, represents the travel time distribution of feasible paths, represents the walking time distribution of passengers entering the station. represents the transfer time distribution of path passengers, represents the waiting time distribution of the route, represents the time distribution within the vehicle along the path, It represents the walking time distribution of the route out of the station;

[0145] The mean and variance of the feasible path travel time distribution are,

[0146]

[0147] The passenger's in-car time is the time from the train leaving the platform after the passenger gets on the train to the train arriving at the platform where the passenger gets off. The calculation formula is as follows:

[0148]

[0149] in, represents the i-th interval [v i ,v i+1 ] train running time, Indicates that the train is at the i-th station v i parking time;

[0150] The train schedule data contains the arrival and departure times of trains at each station, and the running time of trains at adjacent stations is expressed as:

[0151] t r (l,[v i ,v i+1 ],tr)=t a (l,v i+1 ,tr)-t d (l,v i ,tr)

[0152] Among them, t r (l,[v i ,v i+1 ],tr) represents the line l from station v i Run to site v i+1 The running time of train tr;

[0153] Train at station v i The parking time is calculated as follows:

[0154] t l (l,v i ,tr)=t d (l,vi ,tr)-t a (l,v i ,tr)

[0155] Among them, t l (l,v i ,tr) means that the train tr running on l is at station v i parking time.

[0156] In the multi-level passenger flow state estimation system for urban rail transit network described in the present invention, the OD to travel time distribution is estimated according to the mixed Gaussian distribution in S3, specifically,

[0157] Define the definition of the correction ratio coefficient of the initial probability of path selection, which refers to the parameter set when the path selection ratio of all subway passengers is derived using the initial probability obtained from the mobile phone user signaling data, and is expressed as α;

[0158] Calculate the OD vs. travel time distribution based on the Gaussian mixture model:

[0159]

[0160] Among them, α j and r j are the modified proportional coefficient of OD for the jth feasible path and the initial probability of path selection, is the travel time distribution of OD for the jth feasible path.

[0161] Calibrate the path selection initial probability correction proportional coefficient, the specific method is as follows:

[0162] 1). Calculate the KL divergence of the estimated and observed OD travel time distributions,

[0163] Discretize the travel time distribution into K intervals according to a fixed time interval, and the KL divergence calculation formula is as follows:

[0164]

[0165] Among them, t k is the kth travel time length interval, K is the number of discretized travel time length intervals, and β(t k ) are the k-th discretization interval probabilities of the estimated and observed OD for travel time distribution respectively;

[0166] 2). Correct the proportional coefficient to calibrate the nonlinear optimization model construction,

[0167] Nonlinear programming is used to search for the optimal correction proportional coefficient. The objective function is the sum of the KL divergence of the estimated and observed travel time distributions of all OD pairs in the network. The formula is as follows:

[0168]

[0169] Among them, α is the correction coefficient matrix of OD to the path selection probability, and OD vs w i The kth discretized interval probability of the estimated and observed travel time distribution, Z is the number of network OD pairs, α i,j and r i,j Distribution is OD versus w i The modified proportional coefficient and initial probability of the jth feasible path in, Indicates OD to w i The travel time distribution of the jth path;

[0170] OD for w i The expected value of the logarithmic function of the observed travel time distribution is minimized by minimizing

[0171] Based on the path selection probability constraints, a nonlinear optimization model is constructed as follows:

[0172]

[0173] st0≤α i,j ·r i,j ≤1,j∈J,w∈W

[0174]

[0175] The constraints in the model ensure that the corrected path selection probability is between 0 and 1, and the sum of its probabilities is 1;

[0176] 3). Solving the optimal correction proportional coefficient

[0177] The BFGS algorithm is used to solve the nonlinear optimization model and obtain the optimal correction proportional coefficient α * , the formula is:

[0178]

[0179] Estimate the path selection probability between all pairs of network ODs, the formula is as follows:

[0180] p * =α * ·r

[0181] Among them, p * represents the path selection probability matrix between all OD pairs in the network, and r represents the initial probability matrix of feasible path selection between OD pairs.

[0182] The multi-level passenger flow state estimation system for an urban rail transit network described in the present invention uses a BP neural network in S4 to estimate the OD passenger flow of the urban rail transit network, uses monthly historical AFC transaction data as a training set, trains a BP neural network model, and uses the trained BP neural network to estimate the real-time OD passenger flow of the network; at the same time, the passenger exit transaction time is calculated based on the mean value of the feasible path travel time distribution between OD pairs estimated in S3 and the passenger entry transaction time in the AFC transaction record, and obtains the real-time passenger entry and exit transaction time set between OD pairs

[0183] The present invention discloses a multi-level passenger flow state estimation system for an urban rail transit network. S4 randomly and stratified samples the estimated real-time OD passenger flow in proportion according to the feasible path selection probability obtained in S3 to obtain the OD path passenger flow distribution result, and estimates the passenger exit time according to the passenger entry time and the path travel time distribution estimated in S3; the specific steps are: OD feasible path selection probability, calculate the passenger flow demand on the OD feasible path, as shown in the following formula:

[0184] θ u (p) = θ u ×p *

[0185] Among them, θ u (p) is the estimated real-time OD passenger flow demand within the analysis period Δt of the feasible path p, p * is the selection probability of the feasible path p;

[0186] from Random sampling θ u (p) passengers to a feasible path p, It is the estimated real-time passenger flow, which contains the temporal and spatial information of passengers entering and leaving the station. The data samples are: passenger ID, entry station, entry time, exit station, exit time;

[0187] The passenger's exit time is calculated based on the passenger's entry and gate time and the feasible path travel time estimated in step S3, and then The sampled passengers are removed from the calculation and the passenger flow demand on the OD feasible path is returned to be calculated until all passengers are assigned to the feasible path.

[0188] The multi-level passenger flow state estimation system for urban rail transit network described in the present invention, the passenger travel chain reconstruction algorithm in S5, includes reconstructing the travel chains of passengers without transfer and passengers with transfer, specifically,

[0189] S5.1, Initialization: Load the network directed graph G(V,A) constructed in S2.1, the feasible path set of network OD pairs searched in S2.3, and the network OD passenger flow demand within the input analysis period Δt

[0190] S5.2, allocating the network OD passenger flow according to the OD passenger flow allocation method for feasible paths in step S4;

[0191] S5.3, perform the operations in steps S5.4-S5.6 for each OD pair ω in the network OD pair set;

[0192] S5.4. Perform the operation in step S5.5 or S5.6 for the passengers on the feasible path p in the OD pair;

[0193] S5.5. If the feasible path p is a path without transfer, calculate the time similarity between the passenger travel chain and the train operation chain Match passengers to those with the highest time similarity train, thus we can get the travel chain of complete OD passengers on the OD pair ω non-transfer path and store it in the set middle, represents the set of passenger travel chains on the OD pair ω without transfer path p;

[0194] S5.6. If the feasible path p is a transfer path, and the number of transfers in the feasible path is Then the number of travel sub-chains in the passenger travel chain is Match each passenger's travel subchain with the train. Assume that The passengers in the trip sub-chain board the first train that arrives when they are waiting at the platform, and assign the passengers to that train. The time similarity between the passenger travel subchain and the train operation chain is calculated Match passengers to those with the highest time similarity The train of OD is obtained, thus obtaining the travel chain of complete OD passengers on the OD to ω transfer path and storing it in the set middle, represents the set of passenger travel chains on the OD pair ω transfer path p;

[0195] S5.7. So far, the set of travel chains of all passengers in the urban rail transit network within the analysis period Δt is obtained.

[0196] in They represent the travel chain set of passengers without transfers in the network within the analysis period Δt and the travel chain set of passengers with complete OD transfers respectively. The travel chains of all passengers in the network within the analysis period Δt are output, and the algorithm ends.

[0197] In the multi-level passenger flow state estimation system for an urban rail transit network described in the present invention, the number of passengers getting on and off the train is counted in S6, and the real-time passenger capacity of the trains in the line network is counted, and a train passenger capacity calculation model is constructed as follows:

[0198]

[0199] In the formula, Indicates time state I j The passenger capacity of the following train tr at station v is, express The passenger volume of the train at the previous state, Indicates that in time state I j Number of passengers boarding the train tr at the next station v, Indicates that in time state I j Number of passengers getting off at the next station v, train tr, The aforementioned time state I j Represents the jth prediction time interval, which contains the start and end time of the time state;

[0200] In S7, the real-time network section passenger flow at different time granularities is estimated, and the network section passenger flow calculation model is constructed as follows:

[0201] Calculate the section passenger flow based on the estimated real-time passenger capacity of the line network trains in S6 and define the granularity Then, within the analysis period Δt, it is necessary to estimate The cross-sectional passenger flow of time intervals, the i-th time interval in the analysis period Internal line l station v n-1 To site v n-1 Travel direction The passenger flow of the upper section is determined by the time state I j exist Site v n-1 The sum of the train passenger capacity is obtained,

[0202]

[0203] in, Indicates time granularity Next, analyze the time interval Inside, line l station v n-1 To site v n-1 Section Driving Directions of passenger flow, Indicates line l station v n-1 Time Status I j The train under tr k passenger capacity;

[0204] In S8, the number of passengers boarding the train, the number of passengers arriving at the platform, and the number of passengers waiting on the platform are counted, and the calculation model for the number of passengers waiting on the platform is constructed as follows:

[0205]

[0206] In the formula, Indicates time state I j Next station v line l direction The number of people waiting on the platform, express The number of people waiting on the platform at the previous time, Indicates that in time state I j Number of passengers boarding the train tr at the next station v, Indicates that in time state I j Enter from station v and arrive at line l The number of passengers on the platform, The aforementioned time state I j Represents the jth prediction time interval, which contains the start and end time of the time state;

[0207] In S9, the number of passengers in the transfer channel is inferred based on the spatial and temporal positions of getting on and off the bus in the transfer passenger travel chain and the passenger transfer walking time, specifically:

[0208] Read the passenger travel chain from the complete OD passenger travel chain set, and extract the transfer and alighting time tr(x v ), and at the same time extract the transfer walking time of passengers in the driving direction of the corresponding station from the estimated passenger travel time distribution in step S3 Assuming that the time it takes for a passenger to walk from the platform to the transfer passage is half of the time it takes to transfer, then passenger x walks to transfer station v and transfers from line l to line l ′ direction The time for the upper transfer channel is

[0209]

[0210] Given the time granularity Under the condition, the i-th time interval in the analysis period Δt Transfer station v transfers from line l to line l ′ direction The number of passengers in the upper transfer channel is

[0211]

[0212] in, Indicates a given time granularity Under the condition, the i-th time interval in the analysis period Δt Transfer station v transfers from line l to line l ′ direction Number of passengers in the upper transfer channel, Indicates the time interval Passengers gather in the inner transfer channel, Indicates time interval The cardinality of the set of passengers in the internal transfer channel, that is, the number of passengers in the transfer channel.

[0213] The present invention discloses a multi-level passenger flow state estimation system for an urban rail transit network. In S10, a multi-level passenger flow state indicator system for an urban rail transit system is constructed, including micro, meso and macro passenger flow indicators, wherein the micro passenger flow state indicators are passenger boarding and alighting time, location and arrival time at the platform, and walking time or waiting time of a single passenger; the meso passenger flow indicators include the real-time passenger capacity of the train, the real-time number of passengers on the platform, and the number of passengers in the transfer channel; the macro passenger flow indicators include the real-time path flow. On the basis of implementing steps S1-S9, the system is able to output the micro, meso and macro passenger flow indicators of the urban rail transit system in real time.

[0214] Thus, a multi-level passenger flow state estimation system for the urban rail transit network can be obtained. The system can output the real-time spatial and temporal position of passengers in the urban rail transit network during the analysis period, passenger travel characteristic parameters, real-time path passenger flow, train passenger capacity, network section passenger flow, number of people waiting on the platform, and number of people in the transfer channel. It can provide a basis for rail transit network simulation, timetable optimization, ticket clearing, real-time train passenger capacity and section passenger flow calculation, and emergency response to emergencies.

[0215] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects: it can accurately estimate the travel chain of passengers in the urban rail transit network, obtain the boarding and alighting time of all passengers in the network, thereby estimating the passenger capacity of trains running in the network, and further accurately estimate the real-time path passenger flow, train passenger capacity, network section passenger flow, number of people waiting on the platform, and number of people in the transfer channel under a given time and space state, thereby improving the accuracy of the estimation of the passenger flow state of the urban rail transit system, and at the same time facilitating the realization of refined simulation of macro, medium and micro passenger flows of the urban rail transit system. BRIEF DESCRIPTION OF THE DRAWINGS

[0216] Figure 1 The figure is a flow chart of the method of the present invention.

[0217] Figure 2 It is the directed graph of the urban rail transit network of the present invention.

[0218] Figure 3 The feasible path set search algorithm between OD pairs of the present invention is OD pair w ij (v i ,v j ) is a feasible path decision tree between . DETAILED DESCRIPTION

[0219] In order to make the purpose and technical solution of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0220] The present invention provides a method for post-estimation of train passenger capacity in an urban rail transit system based on spatiotemporal network construction and multi-source data driving, which can provide a basis for rail transit line network simulation, timetable optimization, ticket clearing, and emergency response to sudden events.

[0221] like Figure 1 As shown, the present invention comprises ten modules, including the following:

[0222] The road network data loading module loads the basic information of subway network stations and lines, AFC transaction data and train operation schedule data;

[0223] The road network spatiotemporal topology model is constructed and the constrained breadth-first path search algorithm module is used to realize the path search function;

[0224] Based on the passenger route selection probability estimation module of mobile phone signaling data, a nonlinear optimization model is constructed to estimate the selection probability of all passengers on the feasible route by expanding the sample of mobile phone users' travel selection probability;

[0225] The urban rail transit network passenger flow real-time allocation module builds a real-time OD passenger flow estimation model based on BP neural network according to the network OD path selection probability output by the passenger path selection probability estimation module, and allocates the estimated real-time OD passenger flow according to the path selection probability;

[0226] The real-time reconstruction algorithm module of passenger travel chain matches the passenger flow on the OD feasible path with the train, and locates the spatiotemporal information of passengers getting on and off the train during their travel. It uses the inferred minimum walking time for entering the station and the walking time for transfer as constraints, calculates the time similarity between the passenger travel sub-chain and the train operation chain on different paths, matches the passengers to the train, and thus reconstructs the complete travel chain of the passengers.

[0227] The module for calculating the real-time passenger capacity of trains is used to count and calculate the passenger capacity of trains at the stations according to the spatial and temporal coordinates of passengers getting on and off the trains at the stations in the passenger travel chain;

[0228] The real-time section passenger flow calculation module uses the real-time train passenger capacity calculation module to infer the real-time passenger flow of the line section at a given time granularity within the time period;

[0229] The platform waiting passenger number estimation module estimates the number of passengers in the transfer channel based on the nodes where passengers get on and off the train in the passenger travel chain and the passenger walking time inferred in S3, thereby obtaining the real-time number of passengers in the transfer channel of the urban rail transit network;

[0230] The real-time estimation model module for the number of passengers in the transfer channel estimates the number of passengers in the transfer channel based on the nodes where passengers get on and off the train in the passenger travel chain and the passenger walking time inferred in S3, thereby obtaining the real-time number of passengers in the transfer channel of the urban rail transit network;

[0231] The multi-level passenger flow status indicator output module of the urban rail transit system outputs the micro, meso and macro passenger flow status results calculated in the above nine modules.

[0232] The network data loading module reads and analyzes the basic information of subway network sites and lines, the passenger entry and exit transaction information recorded by the AFC system, and the train operation schedule information, and cleans the input data to obtain the data format required by the model; wherein:

[0233] The basic information of subway network stations and lines described in step S1:

[0234] The urban rail transit network station data mainly includes four attribute values: station number, station longitude, station latitude and the line to which it belongs, which can be expressed as:

[0235] VS={[v1,Lon1,Lat 1, l1],…,[v i ,Lon i ,Lat i ,l k ]}

[0236] Among them, v represents the site number, Lon represents the longitude of the site, Lat represents the latitude of the site, and l represents the line number to which the site belongs.

[0237] The urban rail transit network is divided into three levels in space, from high to low: "network, line, station". The network contains different operating lines. The urban rail transit network spatial model is represented by N = {φ(L,V)}, where N represents the road network, L represents the line set, and V represents the station set. The lines and stations form a network through a nonlinear relationship φ.

[0238] The basic properties of the network can be deduced from the logical relationship between lines, stations, and network in the network N = {φ(L,V)}:

[0239] At the line level, multiple stations constitute a line. j The site collection is represented as in, Indicates line l j The i-th station in, i=1,2,…,n, n is the line l j The number of sites in

[0240] At the network level, all lines constitute a road network, and the line set is represented as

[0241] L={l1,l2,…,l m}

[0242] In the formula, l j ,j=1,2,…,m represents a certain line in the road network, and m is the total number of lines in the road network;

[0243] The road network contains all sites, and the set of all sites in the road network is represented as

[0244]

[0245] Using φ(L,V) to represent the spatial relationship among the above stations, lines, and road networks, we have the road network

[0246] N={φ(L,V)}

[0247] In the above network spatial model N = {φ(L,V)}, line l j Contains two attribute values: line name and line rail transit clearing system (hereinafter referred to as ACC) number;

[0248] station It contains five attribute values: station name, line number, station ACC number, station line number, and whether it is a transfer station. The station ACC number uniquely encodes all stations in the network, while the station line number is encoded according to the order of the uplink and downlink directions of the line. The station code is not unique, and the codes of transfer stations in different lines are different. According to the attributes of the line and station, the query function of station and line information in the network can be realized.

[0249] The AFC transaction data described in step S1,

[0250] The transaction data recorded by the urban rail transit automatic ticket vending and checking system includes the passenger's travel starting and ending points and entry and exit times. The AFC transaction data of a complete trip of passenger x can be expressed as:

[0251] Trade x ={ID x ,v o ,ta,v d ,te}

[0252] Where, ID xrepresents the transaction number of passenger x recorded in the automatic ticket vending and checking system, v o ,v d They represent the starting and ending stations of passengers in the urban rail transit system, and ta and te represent the time when passengers pass through the entry and exit gates respectively.

[0253] The train timetable data described in step S1,

[0254] The train operation timetable data includes the arrival and departure time information of the trains on the line, which is expressed as

[0255] T(L,TR)

[0256] In the formula, L is the set of lines in the subway network, TR is the set of trains in the subway network, which can be expressed as

[0257] TR={tr1,tr2,…,tr k ,…,tr Q}, k = 1, 2, ..., K, K is the total number of vehicles dispatched in the network; each line in the network will dispatch different operating vehicles, line l j The set of trains dispatched from k is line l j The number of departures on.

[0258] The signaling base station basic information data in step S1,

[0259] The basic information data of the signaling base station includes the base station number, base station longitude, base station latitude and whether it belongs to the subway system, which is expressed as:

[0260] BS={[BSID1,Lon1,Lat 1, 1],…,[BSID i ,Lon i ,Lat i ,0]}

[0261] Among them, BSID i Indicates the mobile base station code, Lon i Indicates the longitude of the base station, Lat i Indicates the latitude of the base station, i = 1, 2, ..., n, n is the number of base stations, 0 means the base station does not belong to the subway network, 1 means the base station belongs to the subway network;

[0262] The mobile phone signaling data in step S1,

[0263] Mobile phone signaling data includes mobile station identification code, timestamp, mobile base station code and signaling event code. According to mobile phone signaling data, the continuous spatiotemporal position of passengers in the network is obtained. The mobile phone signaling data collected by passenger x in a trip in the subway network is expressed as:

[0264] MS x ={[MSID x ,t1,BSID1,EVID1],…,[MSID x ,t i ,BSID i ,EVID i ]}

[0265] Among them, MSID x Indicates the mobile station identification code MSID, t i Indicates timestamp, EVID i Indicates the signaling event code;

[0266] The main invention contents of the road network spatiotemporal topology model construction module include constructing a road network spatial topology structure according to the characteristics of the urban rail transit network, using a constrained breadth-first algorithm to search for a feasible path set between OD pairs that meets the maximum number of transfer constraints, and selecting the shortest path between OD pairs; then constructing a rail transit road network spatiotemporal topology model according to loading train operation schedule information in the road network topology model; specifically:

[0267] 1. Construct a spatial topological model of the road network to search for the shortest path between OD pairs of the road network;

[0268] According to the logical relationship between stations and lines in the road network N = {φ(L,V)}, the subway network is represented as a directed graph G(V,A), as follows: Figure 2 As shown, V is the set of stations in the subway network, and A is the set of directed edges (train track areas between stations) in the subway network; construct the adjacency matrix M of G(V,A) c and the distance matrix M d ,

[0269]

[0270] In the formula, c ij =1 indicates that sites i and j are directly connected, otherwise, they are not directly connected, which is indicated by 0; d ij Represents the distance of the directed edge from site i to site j. If two sites are directly connected, the edge distance between adjacent sites is the distance a between the sites. ij , otherwise, it is +∞;

[0271] The logical relationship between stations and routes in the road network and the directed graph of the road network together constitute the road network spatial topology model:

[0272] O={φ(L,V),G(V,A)}

[0273] In the network directed graph G(V,A), v∈V represents the nodes of the graph, and a∈A represents the edges of the graph. There are edges in both the upstream and downstream directions between adjacent stations. Virtual platforms are set up at transfer stations to describe the transfer behavior of passengers in the transfer station. For example, Line A and Line B are connected at transfer station v. T connected, then the transfer station v T Describes that there are stations on Line A and Line B and And with the virtual platform v T For a two-way connection, the walking time for passengers to transfer between the two lines is the time it takes for passengers to get off the platform of the previous service train and arrive at the virtual platform v. T The walking time plus the time from the virtual platform v T Walking time to the platform where the subsequent service train arrives.

[0274] 2. Construct the network train operation time information model T (L, V, TR). The information model mainly includes the arrival and departure time information of trains running on different lines. The network time dimension information needs to be extracted from the train timetable. The train arrival and departure information in the train operation timetable can be expressed as:

[0275]

[0276] In the formula, Respectively represent the arrival time of trains in the up and down directions, Respectively represent the departure time of trains in the up and down directions, Respectively represent line l j The i-th site The kth vehicle The arrival time of up and down trains, Respectively represent line l j The i-th site The kth vehicle The departure time of up and down trains, m, n, q are the number of lines in the network, the number of stations in the line, and the number of train departures on the line respectively;

[0277] 3. Load the road network train operation time information model T(L,V,TR) into the road network spatial topology model, and obtain the road network spatiotemporal topology model as follows:

[0278] H={O={φ(L,V),G(V,A)},T(L,V,TR)}

[0279] Wherein, O = {φ(L,V), G(V,A)}, represents the spatial topology model constructed in step S22, and T(L,V,TR) represents the network train operation time information model loaded into the spatiotemporal topology model;

[0280] T(L,V,TR) includes the arrival time and departure time of the train on the line, which can be expressed as:

[0281]

[0282] stl j ∈L

[0283]

[0284]

[0285] Wherein, the constraint condition is equivalent to the logical relationship between stations and lines in the road network N = {φ(L,V)} in S21.

[0286] The constrained breadth-first algorithm is used to search for the shortest path between all OD pairs in the network. The specific algorithm is:

[0287] A1. Initialization: Load the road network directed graph G(V,A) constructed in step S2.1, set the maximum number of transfers for road network passengers X, and input the road network OD pair set R w ;

[0288] A2. Cycle R w OD in w ij (v i ,v j ), and perform the following operations to generate a feasible path decision tree for OD, such as Figure 3 As shown:

[0289] B1. Create an empty queue Q, with OD as the starting point v i As the root node, put it into Q, define array C to represent the number of path transfers, and set C = 0;

[0290] B2. When the queue Q is not empty, loop and perform the following operations:

[0291] C1. Take out the node in Q and determine whether it is the destination v j If yes, go to step C3; otherwise, go to step C2;

[0292] C2. Take the node taken from Q as the parent node v p , search for sites connected to the parent node in G(V,A) As its direct child node, and put it into queue Q; determine whether the child node is a transfer station, if so, the number of transfers of the node is increased by one, otherwise, no operation is performed; the distance from the parent node station to the child node station The number of transfers and sub-node station numbers are recorded together Stored in the corresponding child node;

[0293] C3. End the search; determine the number of transfers of the tree subnodes of the feasible path decision tree c r , r=1,2,…,m (m is the number of leaf nodes) is less than or equal to the maximum number of transfers, if so, keep the branch, otherwise perform pruning;

[0294] A3. Read the feasible path decision tree generated in step A2. Each leaf node branch corresponds to a feasible path. Calculate the cost of each feasible path and select the path with the least path cost as the OD pair w. ij The shortest path between Store OD in the shortest path set P w middle.

[0295] The passenger path selection probability estimation module based on mobile phone signaling data mainly includes calculating the initial path selection probability based on the mobile phone user travel trajectory reconstructed by mobile phone signaling data, analyzing the passenger entry and exit time and transfer time distribution (including walking and waiting time) based on the travel data of a single path without transfer and a single path with transfer, thereby inferring the travel time distribution of a feasible path, and then estimating the OD pair travel data distribution through the mixed Gaussian distribution theory, and then constructing a nonlinear optimization model with the KL divergence of the estimated and observed OD pair travel time distribution as the optimization target, and using the mobile phone user's path selection probability to expand the sample to estimate the path selection probability of all subway passengers; specifically:

[0296] 1. Calculate the initial probability of path selection using the mobile phone user travel trajectory reconstructed using mobile phone signaling data, specifically,

[0297] According to the base station location and time information in the passenger's mobile phone signaling data, the longitude and latitude of the passenger's mobile phone signaling data are queried and matched with the longitude and latitude of the subway station to reconstruct the passenger's travel trajectory in the subway network. Specifically,

[0298] Read passenger x mobile phone signaling data MS x The mobile base station number BSID corresponding to each signaling event;

[0299] Read the latitude and longitude data of the base station BSID from the basic information data of the mobile base station to determine whether it belongs to a dedicated base station for the subway network, obtain the passenger's location and determine whether it is in the subway network, and reconstruct the passenger's travel trajectory in the subway network, including passenger entry judgment, travel trajectory completion and passenger exit judgment;

[0300] The travel trajectory of passenger x in the subway network based on signaling reconstruction is expressed as:

[0301] J x :{(v o ,t1),…,(v i ,t i ),…,(v d ,t n )}

[0302] Among them, v i represents the number of the ith station in the travel trajectory of passenger x, t i Indicates that after v i time.

[0303] Match the travel trajectory of mobile phone users with feasible paths between OD pairs of urban rail transit network, including:

[0304] The spatial similarity between the passenger's mobile phone signaling reconstruction trajectory and the feasible path on the corresponding OD pair is defined as follows:

[0305]

[0306] In the formula, represents the trajectory p reconstructed by the mobile phone signaling of passenger x x The feasible path p with OD w The spatial similarity of For p x With p w The number of common sites, Indicates p x The number of sites, Indicates p w The number of sites, γ is a positive constant;

[0307] The spatial similarity between the passenger travel trajectory reconstructed based on mobile phone signaling and all feasible paths between OD pairs is calculated, and the feasible path with the highest similarity to the reconstructed trajectory is taken as the passenger's travel trajectory in the subway network.

[0308] Based on the travel trajectories of the above passengers in the subway network, the number of specific mobile phone users on the feasible paths of the OD pairs is counted, and the initial probability of feasible path selection between the OD pairs is calculated. The formula is as follows:

[0309]

[0310] Among them, r j represents the feasible path p for OD to w j The initial probability of path selection, n j Select a feasible path p for a specific mobile phone user passenger on OD pair w j The number of people, nw is the number of passengers with specific mobile phone users on OD pair w.

[0311] 2. Define the passenger's travel chain, specifically:

[0312]

[0313] In the formula, λ x represents the passenger travel chain, x represents a passenger entering the urban rail transit system, ta represents the transaction time when the passenger passes through the entry gate, and te represents the transaction time when the passenger passes through the exit gate; x v (tr) represents the time when passenger x boarded train tr at station v, tr(x v ) represents the time when passenger x gets off the train tr at station v;

[0314] A passenger trip chain contains one or more trip subchains. It constitutes the travel sub-chain of a passenger taking a train in the urban rail transit system. The travel chain of passengers without transfers contains only one travel sub-chain, while the travel chain of passengers with transfers contains multiple travel sub-chains. In the above formula, the travel chain of the passenger contains k travel sub-chains.

[0315] 3. Passengers are matched to trains using the acquired AFC data and train schedule data, thereby inferring the passenger's travel chain, specifically:

[0316] According to the obtained train schedule, define the train operation chain of path p:

[0317]

[0318] In the formula, λ p represents the train operation chain of path p, are the jth travel subchain of the train operation chain, and the travel subchain includes the starting point of the train at section j Departure time on and end point Arrival time on The train operation chain without transfer path contains only one train operation sub-chain, while the train operation chain with transfer path contains multiple train operation sub-chains;

[0319] Define the similarity between the passenger travel time interval and the train operation time interval on the path:

[0320]

[0321] In the formula, represents the time similarity between the travel chain of passenger x and the travel chain of train tr on path p, is the length of the intersection of the passenger travel time interval and the train operation time interval on segment j of path p, represents the travel time interval length of passenger x on segment j of path p, represents the length of the running time interval of train tr on section j of path p, and γ is a positive constant;

[0322] According to the passenger transaction data recorded by the automatic ticket vending and checking system x , extract the passenger's entry and exit OD information, and then search the feasible short path set P of the road network w , passengers are classified according to the feasible path set and the number of transfers based on OD, and passenger travel is divided into single path without transfer, single path with single transfer, single path with multiple transfers, and multi-path travel according to the number of transfers on the path;

[0323] For passengers with a single route without transfers, the trains taken by the passengers are matched using time similarity; the travel time interval of the passengers is determined by the transaction data Trade x The transaction time of the passenger entering and exiting the station is obtained from the train timetable information T(L,TR). The similarity between the running time interval of the train tr on path p after the passenger enters the station and the travel time interval of passenger x is calculated, and the passenger is matched to the time interval similarity. The highest train goes up;

[0324] According to the non-transfer single-path passenger matching results obtained in S3.7, the time from the passenger entering the station through the gate to boarding the train and the time from the passenger getting off the train to exiting the station through the gate are calculated, which can be expressed as

[0325]

[0326] In the formula, ACT represents the time from when the passenger enters the station through the gate to when he / she boards the train tr, and x v (tr) is the time when the passenger boards the train tr at station v, ta is the transaction time when the passenger passes through the entrance gate; EGT is the time when the passenger gets off the train and passes through the gate to exit the station, tr(x v ) is the time when the passenger gets off the train tr at station v, te is the transaction time when the passenger passes through the exit gate; let the entry and exit time distribution of passengers on a single path without transfer at station v represent the entry and exit time distribution of passengers at station v, and calculate the Gaussian distribution of passenger entry and exit time;

[0327] To match single-transfer single-path passengers to trains, the following relationship needs to be satisfied between the passenger travel chain and the train operation chain:

[0328]

[0329] In the formula, t d(l,tr,v) represents the time when train tr on line l leaves station v, t a (l,tr,v) represents the time when train tr running on line l arrives at station v, ACT min (l,v) represents the minimum entry time of a passenger at station v on line l, EGT min (l,v) represents the minimum exit time of passengers at station v on line l, t d (l ′ ,tr ′ ,v) indicates the transfer line l ′ Transfer train running on tr ′ Time to leave site v, TFT min (l,l ′ ,v) indicates that at station v, passengers transfer from line l to line l ′ Minimum transfer time;

[0330] Match single-transfer single-path passengers to trains, and analyze the transfer time between transfer stations and lines in different transfer directions; obtain the ACT by calculating the entry and exit time distribution of station passengers through S3.8 min (l,v) and EGT min (l,v);

[0331] By analyzing the passenger's AFC transaction data and train schedule, and substituting them into constraint (2) in S3.9, the train that the passenger takes for the last segment of the travel sub-chain of the single-transfer single-path passenger is found, and the boarding time of the passenger's last segment of the travel sub-chain is obtained;

[0332] Taking the passenger boarding time of the last travel subchain as the estimated end time of the first travel subchain, taking ta as the estimated start time of the first travel subchain, calculate the time similarity between the estimated time interval of the passenger's first travel subchain and the train operation chain of the section where the passenger's first travel subchain is located;

[0333] Match the passenger to the train with the highest time similarity, and obtain the passenger's first travel subchain, thus obtaining the complete travel chain of the single-transfer passenger;

[0334] Then the transfer time of passengers is calculated. The transfer time calculation method is as follows:

[0335]

[0336] Where TFT represents the passenger transfer time, Indicates that the passenger boards line l at transfer station v ′ The nth train time, Indicates that the passenger transfers from the i-th train on line l at the transfer station v The time of getting off the bus; assuming that the transfer time distribution of single-transfer passengers at the station represents the transfer time distribution of transfer passengers, the transfer time distribution of transfer passengers at the station is obtained by calculating the transfer time of single-transfer passengers in each transfer direction at the station;

[0337] 4. Estimate the walking time for passengers entering and exiting the station and the walking time for transfers, and calculate the average headway time of trains at the station:

[0338]

[0339] in, For line l station v direction The average headway of trains, Respectively represent the running direction of line l station v The departure time of the k+1th and kth trains, K is the number of train departures; the average waiting time of passengers on the platform is half of the headway time, that is,

[0340]

[0341] in, Indicates line l station v direction The average waiting time of passengers is Indicates line l station v direction Average headway of trains;

[0342] Assume that the standard deviation of passenger waiting time is one-fourth of the standard deviation of train headway time, that is,

[0343]

[0344] in, For line l station v direction The standard deviation of the waiting time of passengers, For line l station v direction The headway of the i-th train, For line l station v direction Average headway of trains on

[0345] Calculate the distribution of passengers' walking time when entering the station. According to the additivity of independent Gaussian distribution, using the estimated distribution of passengers' entering and leaving the station, we can get the distribution of passengers' walking time when entering the station as:

[0346]

[0347] in, represents the distribution of passengers’ walking time into the station, N(μ ACT ,σ ACT ) represents the distribution of passengers’ arrival time, It represents the waiting time distribution of passengers at the platform after entering the station. The relationship between the mean and standard deviation is:

[0348]

[0349] Calculate the distribution of passengers' transfer walking time, specifically:

[0350]

[0351] in, represents the distribution of passengers' transfer walking time, represents the distribution of passengers’ transfer time,

[0352] represents the waiting time distribution of passengers at the transfer station, and the relationship between the mean and standard deviation is:

[0353]

[0354] Assuming that passengers get off at their destination and immediately walk to the exit gate, the passenger's exit walking time distribution is the exit time distribution.

[0355]

[0356] So far, we have obtained the distribution of passengers' walking time in and out of the urban rail transit system, the distribution of transfer walking time in the transfer direction of each transfer station, and the distribution of passengers' waiting time at each station.

[0357] 5. Estimate the travel time distribution of feasible paths, specifically,

[0358] According to the additivity of independent Gaussian distribution, the travel time distribution of the feasible path can be obtained as the sum of the passenger's walking time in and out of the station, transfer walking time, waiting time, and time in the car.

[0359]

[0360] in, represents the travel time distribution of feasible paths, represents the walking time distribution of passengers entering the station. represents the transfer time distribution of path passengers, represents the waiting time distribution of the route, represents the time distribution within the vehicle along the path, It represents the walking time distribution of the route out of the station;

[0361] The mean and variance of the feasible path travel time distribution are,

[0362]

[0363] The passenger's in-car time is the time from the train leaving the platform after the passenger gets on the train to the train arriving at the platform where the passenger gets off. The calculation formula is as follows:

[0364]

[0365] in, represents the i-th interval [v i ,v i+1 ] train running time, Indicates that the train is at the i-th station v i parking time;

[0366] The train schedule data contains the arrival and departure times of trains at each station, and the running time of trains at adjacent stations is expressed as:

[0367] t r (l,[v i ,v i+1 ],tr)=t a (l,v i+1 ,tr)-t d (l,v i ,tr)

[0368] Among them, t r (l,[v i ,v i+1 ],tr) represents the line l from station v i Run to site v i+1 The running time of train tr;

[0369] Train at station v i The parking time is calculated as follows:

[0370] t l (l,v i ,tr)=t d (l,v i ,tr)-t a (l,v i ,tr)

[0371] Among them, t l (l,v i ,tr) means that the train tr running on l is at station v i parking time;

[0372] 6. Estimate the distribution of OD versus travel time based on the mixed Gaussian distribution, specifically,

[0373] Define the definition of the correction ratio coefficient of the initial probability of path selection, which refers to the parameter set when the path selection ratio of all subway passengers is derived using the initial probability obtained from the mobile phone user signaling data, and is expressed as α;

[0374] Calculate the OD vs. travel time distribution based on the Gaussian mixture model:

[0375]

[0376] Among them, α j and r j are the modified proportional coefficient of OD for the jth feasible path and the initial probability of path selection, is the travel time distribution of OD for the jth feasible path.

[0377] Calibrate the path selection initial probability correction proportional coefficient, the specific method is as follows:

[0378] 1). Calculate the KL divergence of the estimated and observed OD travel time distributions,

[0379] Discretize the travel time distribution into K intervals according to a fixed time interval, and the KL divergence calculation formula is as follows:

[0380]

[0381] Among them, t k is the kth travel time length interval, K is the number of discretized travel time length intervals, and β(t k ) are the k-th discretization interval probabilities of the estimated and observed OD for travel time distribution respectively;

[0382] 2). Correct the proportional coefficient to calibrate the nonlinear optimization model construction,

[0383] Nonlinear programming is used to search for the optimal correction proportional coefficient. The objective function is the sum of the KL divergence of the estimated and observed travel time distributions of all OD pairs in the network. The formula is as follows:

[0384]

[0385] Among them, α is the correction coefficient matrix of OD to the path selection probability, and OD vs w i The kth discretized interval probability of the estimated and observed travel time distribution, Z is the number of network OD pairs, α i,j and r i,j Distribution is OD versus w i The modified proportional coefficient and initial probability of the jth feasible path in, Indicates OD to wi The travel time distribution of the jth path;

[0386] OD for w i The expected value of the logarithmic function of the observed travel time distribution is minimized by minimizing

[0387] Based on the path selection probability constraints, a nonlinear optimization model is constructed as follows:

[0388]

[0389] st0≤α i,j ·r i,j ≤1,j∈J,w∈W

[0390]

[0391] The constraints in the model ensure that the corrected path selection probability is between 0 and 1, and the sum of its probabilities is 1;

[0392] 3). Solving the optimal correction proportional coefficient

[0393] The BFGS algorithm is used to solve the nonlinear optimization model and obtain the optimal correction proportional coefficient α * , the formula is:

[0394]

[0395] Estimate the path selection probability between all pairs of network ODs, the formula is as follows:

[0396] p * =α * ·r

[0397] Among them, p * represents the path selection probability matrix between all OD pairs in the network, and r represents the initial probability matrix of feasible path selection between OD pairs.

[0398] The real-time passenger flow distribution module of the urban rail transit network mainly includes estimating the OD passenger flow of the urban rail transit network by using a BP neural network, training a BP neural network model with monthly historical AFC transaction data as a training set, and estimating the real-time OD passenger flow of the network by using the trained BP neural network; according to the feasible path selection probability obtained in S3, the estimated real-time OD passenger flow is randomly stratified sampled in proportion to obtain the OD path passenger flow distribution result, and the passenger exit time is estimated according to the passenger entry time and the path travel time distribution estimated in S3; wherein:

[0399] S4A.1. Collect historical AFC passenger flow data, clean the missing values, outliers and noise in the data, and standardize the data;

[0400] S4A.2. Extract features from historical AFC data, extract time features such as hours, dates, holidays, working days or non-working days; extract the distance between the starting and ending points, and the number of path transfers;

[0401] S4A.3, training set and test set division, the preprocessed data set is divided into training set and test set, with a ratio of 7:3. The training set is used to train the model, and the test set is used to verify the prediction performance of the model;

[0402] S4A.4, build a BP neural network model, take the features extracted in S4.2 and the OD passenger flow in the analysis period in the historical data of the previous month as input, set the hidden layer to two layers, the number of hidden layer nodes is twice that of the input layer, and the output layer is the estimated period OD passenger flow;

[0403] S4A.5. Use the training set to train the BP neural network, and use the test set to verify and tune the model;

[0404] S4A.6, real-time OD passenger flow prediction, input the AFC entry transaction data of the analysis period in the actual scenario, and estimate the real-time OD passenger flow data;

[0405] S4B.1. Calculate the passenger flow demand on the OD feasible path based on the obtained OD feasible path selection probability.

[0406] θ u (p) = θ u ×p *

[0407] Among them, θ u (p) is the estimated real-time OD passenger flow demand within the analysis period Δt of the feasible path p, p * is the selection probability of the feasible path p;

[0408] S4B.2, from Random sampling θ u (p) passengers to the feasible path p, calculate the passenger's exit time based on the passenger's entry and gate time and the feasible path travel time estimated in step S3, and then The sampled passengers are removed and the process returns to step S4B.1 until all passengers are assigned to feasible paths;

[0409] The passenger travel chain real-time reconstruction algorithm module mainly includes matching passenger flow on OD feasible paths with trains, locating the spatiotemporal information of passengers getting on and off the train during travel; using the inferred minimum walking time for entering the station and the transfer walking time as constraints, calculating the time similarity between passenger travel sub-chains and train operation chains on different paths, matching passengers to trains, and thus reconstructing the complete travel chain of passengers; wherein:

[0410] S5.1, Initialization: Load the network directed graph G(V,A) constructed in S2.1, the feasible path set of network OD pairs searched in S2.3, and the network OD passenger flow demand within the input analysis period Δt

[0411] S5.2, allocating the network OD passenger flow according to the OD passenger flow allocation method for feasible paths in step S4;

[0412] S5.3, perform the operations in steps S5.4-S5.6 for each OD pair ω in the network OD pair set;

[0413] S5.4. Perform the operation in step S5.5 or S5.6 for the passengers on the feasible path p in the OD pair;

[0414] S5.5. If the feasible path p is a transfer path, execute the operation in step S5.5A or S5.5B;

[0415] S5.5A. If the passenger trip is a complete OD trip, calculate the time similarity between the passenger trip chain and the train operation chain. Match passengers to those with the highest time similarity train, thus we can get the travel chain of complete OD passengers on the OD pair ω non-transfer path and store it in the set middle, represents the set of complete OD passenger travel chains on the OD pair ω without transfer path p;

[0416] S5.5B, if the passenger's trip is a non-complete OD trip, the passenger is assigned to the first bus that arrives when he is waiting at the platform. This will get the trip chain of non-complete OD passengers on the OD pair ω non-transfer path and store it in the set middle, represents the set of incomplete OD passenger travel chains on the OD pair ω without transfer path p;

[0417] S5.6. If the feasible path p is a transfer path, and the number of transfers in the feasible path is Then the number of travel sub-chains in the passenger travel chain is Match each passenger's travel subchain with a train, specifically executing the operation in step S5.6A or S5.6B;

[0418] S5.6A, if the passenger trip is a complete OD trip, assuming that The passengers in the trip sub-chain board the first train that arrives when they are waiting at the platform, and assign the passengers to that train. The time similarity between the passenger travel subchain and the train operation chain is calculated Match passengers to those with the highest time similarity train, thus we can get the travel chain of complete OD passengers on the OD to ω transfer path and store it in the set middle, represents the set of complete OD passenger travel chains on the OD pair ω transfer path p;

[0419] S5.6B, if the passenger's trip is a non-complete OD trip, then in each trip sub-chain, the passenger is assigned to the first train waiting at the platform. This way, the trip chain of the non-complete OD passenger on the OD-ω transfer path can be obtained and stored in the set middle, represents the set of incomplete OD passenger travel chains on the OD pair ω transfer path p;

[0420] S5.7. So far, the set of travel chains of all passengers in the urban rail transit network within the analysis period Δt is obtained. in They respectively represent the travel chain set of complete OD passengers without transfers in the network within the analysis period Δt, the travel chain set of incomplete OD passengers without transfers, the travel chain set of OD passengers with complete transfers, and the travel chain set of OD passengers with incomplete transfers. The travel chains of all passengers in the network within the analysis period Δt are output, and the algorithm ends.

[0421] The main invention content of the train real-time passenger capacity calculation module includes matching the nodes of passengers getting on and off the train in the passenger travel chain with the train timetable, counting the number of passengers getting on and off the station train, and then using the train passenger capacity statistical model to obtain the train passenger capacity, thereby inferring the passenger capacity of the network operation train; specifically:

[0422] Construct a train passenger capacity calculation model:

[0423]

[0424] In the formula, Indicates time state I j The passenger capacity of the following train tr at station v is, express The passenger volume of the train at the previous state, Indicates that in time state I j Number of passengers boarding the train tr at the next station v, Indicates that in time state Ij Number of passengers getting off at the next station v, train tr, The aforementioned time state I j Represents the jth prediction time interval, which contains the start and end time of the time state;

[0425] Count the number of people boarding the platform:

[0426] Generate time state I based on passenger travel chain j Next stop v Passengers who have successfully boarded the bus will gather Statistics time status I j Number of passengers boarding the train arriving at the next station v:

[0427]

[0428] In the formula, Indicates time state I j Next station v train tr k Number of passengers boarding the vehicle.

[0429] Count the number of people getting off the platform:

[0430] Generate time state I based on passenger travel chain j Get-off stop v Get-off passengers gathering Statistics time status I j Number of passengers getting off the train arriving at the next station v:

[0431]

[0432] In the formula, Indicates time state I j Next station v train tr k Number of passengers alighting.

[0433] Substitute the number of passengers boarding and getting off the train at the platform into the train passenger capacity calculation model to obtain Thus, the passenger capacity of all trains running in the entire urban rail transit network can be obtained.

[0434]

[0435] In the formula, Indicates time state I j The passenger capacity of trains running in the urban rail transit network is as follows: Indicates time state I j Next site v m Train tr k of passenger capacity.

[0436] The real-time section passenger flow calculation module mainly includes the real-time passenger flow of the line section at a given time granularity within the time period obtained by the real-time train passenger capacity inference calculation and analysis module; specifically:

[0437] Calculate the section passenger flow based on the estimated real-time passenger capacity of the line network trains in S6 and define the granularity Then, within the analysis period Δt, it is necessary to estimate The cross-sectional passenger flow of time intervals, the i-th time interval in the analysis period Internal line l station v n-1 To site v n-1 Travel direction The passenger flow of the upper section is determined by the time state I j exist Site v n-1 The sum of the train passenger capacity is obtained,

[0438]

[0439] in, Indicates time granularity Next, analyze the time interval Inside, line l station v n-1 To site v n-1 Section Driving Directions of passenger flow, Indicates line l station v n-1 Time Status I j The train under tr k of passenger capacity.

[0440] In this way, the real-time passenger flow in the up and down directions of each line and section of the urban rail transit network at different time granularities within the analysis period Δt can be obtained, which can provide a basis for rail transit network simulation, timetable optimization, ticket clearing, real-time train passenger capacity and section passenger flow calculation, and emergency response to emergencies.

[0441] The platform waiting passenger number estimation module mainly includes counting the number of passengers boarding the train, the number of passengers arriving at the platform, and calculating the number of passengers waiting at the platform. The platform waiting passenger number calculation model is constructed as follows:

[0442]

[0443] In the formula, Indicates time state I j Next station v line l direction The number of people waiting on the platform, express The number of people waiting on the platform at the previous time, Indicates that in time state I j Number of passengers boarding the train tr at the next station v, Indicates that in time state I j Enter from station v and arrive at line l The number of passengers on the platform, The aforementioned time state I j represents the jth predicted time interval, and includes information on the start and end time of the time state; the platform waiting passenger number calculation model further includes:

[0444] Generate time state I based on passenger travel chain j Next stop v Passengers who have successfully boarded the bus will gather Indicates time state I j Passenger alighting x q Get on the train at station v k+n , Statistical time status I j Number of passengers boarding the train arriving at the next station v:

[0445]

[0446] In the formula, Indicates time state I j Next station v train tr k Number of passengers on board;

[0447] Generate time state I based on passenger travel chain j Enter from station v and arrive at line l Passengers gathering on the platform Indicates time state I j Enter station v and arrive at line l Passengers on the platform x q , the time it takes for passengers to arrive at the platform is calculated by the following formula

[0448]

[0449] In the formula, Indicates that passenger x arrives at station v, direction of route l The time to get on the platform, if Then passenger x is divided into the set In

[0450]

[0451] In the formula, Indicates that in time state I j Enter from station v and arrive at line l The number of passengers on the platform, Representing a collection The cardinality of .

[0452] Substitute the calculated number of people boarding the train and arriving at the platform into the platform waiting passenger number calculation model to calculate the time state I j The number of waiting passengers at different platforms of all stations can be obtained, thus obtaining the number of waiting passengers at all platforms of the entire urban rail transit network.

[0453]

[0454] In the formula, Indicates time state I j The number of passengers waiting for trains at all platforms in the next urban rail transit network, Indicates time state I j Next site v m Line n direction The number of people waiting on the platform.

[0455] The real-time estimation model module for the number of passengers in the transfer channel mainly includes inferring the number of passengers in the transfer channel according to the spatial and temporal positions of getting on and off the bus in the transfer passenger travel chain and the passenger transfer walking time, specifically:

[0456] Read the passenger travel chain from the complete OD passenger travel chain set, and extract the transfer and alighting time tr(x v ), and at the same time extract the transfer walking time of passengers in the driving direction of the corresponding station from the estimated passenger travel time distribution in step S3 Assuming that the time it takes for a passenger to walk from the platform to the transfer passage is half of the time it takes to transfer, then passenger x walks to transfer station v and transfers from line l to line l ′ direction The time for the upper transfer channel is

[0457]

[0458] Given the time granularity Under the condition, the i-th time interval in the analysis period Δt Transfer station v transfers from line l to line l ′ direction The number of passengers in the upper transfer channel is

[0459]

[0460] in, Indicates a given time granularity Under the condition, the i-th time interval in the analysis period Δt Transfer station v transfers from line l to line l ′ direction Number of passengers in the upper transfer channel, Indicates the time interval Passengers gather in the inner transfer channel, Indicates time interval The cardinality of the set of passengers in the internal transfer channel, that is, the number of passengers in the transfer channel.

[0461] The multi-level passenger flow status indicator output module of the urban rail transit system mainly includes the construction of a multi-level passenger flow status indicator system for the urban rail transit system, including micro, meso and macro passenger flow indicators, wherein the micro passenger flow status indicators are the passenger boarding and alighting time, location and arrival time at the platform, and the walking time or waiting time of a single passenger; the meso passenger flow indicators include the real-time passenger capacity of the train, the real-time number of people waiting for the train at the platform, and the number of people in the transfer channel; the macro passenger flow indicators include the real-time path flow. On the basis of implementing steps S1-S9, the system can output the micro, meso and macro passenger flow indicators of the urban rail transit system in real time.

[0462] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above examples. The above examples and descriptions are only for explaining the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A multi-level passenger flow state estimation system for an urban rail transit network, characterized in that: The steps include: S1, obtain and analyze the basic data of urban rail transit network stations and lines, passenger entry and exit transaction data recorded by the AFC system, train operation timetable data, obtain basic information data of signaling base stations, and obtain mobile phone signaling data; S2, constructing a spatiotemporal topological model of the road network and a constrained breadth-first algorithm to search for a feasible path set between OD pairs. According to the train operation schedule information loaded in the road network topological model, a spatiotemporal topological model of the rail transit road network is constructed, and a constrained breadth-first algorithm is used to search for a feasible path set between OD pairs that meets the maximum number of transfer constraints; S3, construct a passenger route selection probability estimation model based on mobile phone signaling data, calculate the initial probability of route selection with the mobile phone user travel trajectory reconstructed from the mobile phone signaling data, and define the passenger's travel chain; The passenger entry and exit time and transfer time distribution are analyzed by using the travel data of single-path without transfer and single-path with transfer, so as to obtain the travel time distribution of feasible paths. Then, the travel data distribution between OD pairs is estimated by using the mixed Gaussian distribution theory, and then a nonlinear optimization model is constructed with the KL divergence of the estimated and observed travel time distribution between OD pairs as the optimization target. The path selection probability of all passengers in urban rail transit is estimated by expanding the sample using the path selection probability of mobile phone users. S4, combining the feasible path selection probability obtained in step S3, the real-time OD passenger flow estimated by the BP neural network is randomly stratified sampled in proportion to obtain the OD path passenger flow distribution result, and the passenger exit time is estimated according to the passenger entry time and the path travel time distribution estimated in S3, and the real-time estimated passenger exit time between OD pairs is obtained; S5, build a passenger travel chain reconstruction algorithm model, match the passenger flow on the OD feasible path with the train, and locate the spatiotemporal information of passengers getting on and off the train during the trip; The minimum passenger entry and transfer walking time inferred from the passenger entry and exit time of single-path travel without transfer and single-path travel with transfer in S3 is used as a constraint. The passenger exit time between OD pairs estimated in real time in S4 is used to calculate the time similarity between the passenger travel subchain and the train operation subchain on the feasible paths between different OD pairs, match the passengers to the trains, and thus reconstruct the complete travel chain of the passengers. S6, according to the matching of the nodes where passengers get on and off the train in the complete travel chain of the passengers with the train timetable, the number of passengers getting on and off the train at the station is counted, and the real-time train passenger capacity is obtained by using the train passenger capacity statistical model, thereby obtaining the real-time train passenger capacity of the urban rail transit line network; S7, based on the real-time passenger capacity of the train inferred in S6, estimate the real-time passenger flow of the line section at different time granularities within the analysis time period to obtain the real-time passenger flow of the urban rail transit line network; According to the real-time passenger volume of trains under different time states during the analysis period inferred in S6, S8. According to the matching of the nodes where passengers get on and off the train in the complete travel chain of the passengers with the train timetable, the number of passengers getting on the train at the station is calculated, and the number of passengers waiting on the platform is obtained by using the estimation model of the number of passengers waiting on the platform, and the number of passengers waiting on the platform at the urban rail transit line is obtained; S9, construct a real-time estimation model for the number of passengers in the transfer channel, and infer the number of passengers in the transfer channel according to the nodes where passengers get on and off the bus in the complete travel chain of the passengers and the average walking time of the passengers inferred in S3, so as to obtain the real-time number of passengers in the transfer channel of the urban rail transit network; S10. Construct a multi-level passenger flow status indicator system for urban rail transit systems, including micro, meso and macro passenger flow indicators. Micro passenger flow status indicators include passenger boarding and alighting time, location and arrival time at the platform, and individual passenger walking time or waiting time; Meso-level passenger flow indicators include the real-time passenger capacity of trains, the real-time number of people waiting for trains on platforms, and the number of people in transfer passages; Macro passenger flow indicators include real-time route flow; Based on steps S1-S9, the system can output micro, meso and macro passenger flow indicators of the urban rail transit system in real time.

2. A multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: The basic data of rail transit network stations and lines are obtained in S1, as follows: S1.1, the urban rail transit network station data includes four attribute values: station number, station longitude, station latitude and the line to which it belongs, expressed as: VS={[v1,Lon1,Lat 1, l1],…,[v i ,Lon i ,Lat i ,l j ]} VS represents the basic data of network sites and lines, i represents the site number, and v i represents the site number, Lon represents the longitude of the site, Lat represents the latitude of the site, k represents the line number, l j Indicates the line number to which the station belongs; N = {φ(L,V)} represents the spatial model of urban rail transit network. N represents the road network, L represents the set of lines, V represents the set of stations, and the lines and stations form a network through a nonlinear relationship φ; The basic properties of the network are derived from the logical relationship between lines, stations, and network in the network N = {φ(L,V)}: At the line level, line l j The site collection is represented as in, Indicates line l j The i-th station in, i=1,2,…,n, n is the line l j The number of sites in At the network level, the line set is represented by L = {l1,l2,…,l m }, where l j ,j=1,2,…,m represents a certain line in the road network, and m is the total number of lines in the road network; The set of all sites in the road network is represented as Using φ(L,V) to represent the spatial relationship between stations, routes, and road networks, the road network spatial model is expressed as follows: N={φ(L,V)} In the model, line l j Contains two attribute values: line name and line rail transit clearing system ACC number; station Contains five attribute values: station name, line number, station ACC number, station line number, and whether it is a transfer station; S1.2, obtain AFC transaction data. The transaction data recorded by the urban rail transit automatic ticket vending and checking system includes the passenger's travel starting and ending points and entry and exit times. The AFC transaction data of a complete trip of passenger x is expressed as: Trade x ={ID x ,v o ,t.a,v d ,t.e} Where, ID x represents the transaction number of passenger x recorded in the automatic ticketing system, v o ,v d They represent the starting and ending stations of passengers in the urban rail transit system, and ta and te represent the time when passengers pass through the entry and exit gates respectively; S1.3, obtain train timetable data. The train operation timetable data includes the arrival and departure time information of the operating trains on the line, which is expressed as: T(L,TR) In the formula, L is the set of lines in the subway network, TR is the set of trains in the subway network; TR={tr1,tr2,…,tr k ,…,tr K }, k = 1, 2, ..., K, K is the total number of departures in the road network; Line j The set of trains dispatched from Indicates line l j The g-th train departs from the line l. j The number of departures on S1.4, obtain the basic information data of the signaling base station, the steps are as follows: The basic information data of the signaling base station includes the base station number, base station longitude, base station latitude and whether it belongs to the subway system, which is expressed as: BS={[BSID1,Lon1,Lat 1, 1],…,[BSID i ,Lon i ,Years i ,0]} Among them, BSID i Indicates the mobile base station code, Lon i Indicates the longitude of the base station, Lat i Indicates the latitude of the base station, i = 1, 2, ..., n, n is the number of base stations, 0 means the base station does not belong to the subway network, 1 means the base station belongs to the subway network; S1.5, obtain mobile phone signaling data, the steps are as follows: Mobile phone signaling data includes mobile station identification code, timestamp, mobile base station code and signaling event code. According to mobile phone signaling data, the continuous spatiotemporal position of passengers in the network is obtained. The mobile phone signaling data collected by passenger x in a trip in the subway network is expressed as: MS x ={[MSID x ,t1,BSID1,EVID1],…,[MSID x ,t i ,BSID i ,EVID i ]} Among them, MSID x Indicates the mobile station identification code MSID, t i Indicates timestamp, EVID i Indicates the signaling event code.

3. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: The construction of the spatiotemporal topological model of the rail transit network in S2 specifically includes: S2.1, construct a road network spatial topology model to search for the shortest path between OD pairs in the road network; According to the logical relationship between stations and lines in the road network N = {φ(L,V)}, the rail transit network is represented as a directed graph G(V,A). Where V is the set of stations in the rail transit network, A is the set of directed edges in the rail transit network, and the directed edges represent the train track areas between stations; Construct the adjacency matrix M of G(V,A) c and the distance matrix M d , In the formula, the adjacency matrix M c The matrix element c ij =1 indicates that sites i and j are directly connected, otherwise, they are not directly connected, which is represented by 0; the distance matrix M d The matrix element d ij Represents the distance of the directed edge from site i to site j. If two sites are directly connected, the edge distance between adjacent sites is the distance a between the sites. ij , otherwise, it is +∞; The logical relationship between stations and routes in the road network and the directed graph of the road network together constitute the road network spatial topology model: O={φ(L,V),G(V,A)} Wherein, O represents the spatial topological model of the road network, in the network directed graph G(V,A), v∈V represents the nodes of the graph, and a∈A represents the edges of the graph. There are edges in the uplink and downlink directions between adjacent stations, and virtual platforms are set up at the transfer station to describe the transfer behavior of passengers in the transfer station. S2.2, construct the network train operation time information model T (L, V, TR), which includes the arrival and departure time information of trains running on different lines. The network time dimension information is extracted from the train timetable. The train arrival and departure information in the train operation timetable is expressed as: In the formula, Respectively represent the arrival time of trains in the up and down directions, Respectively represent the departure time of trains in the up and down directions, Respectively represent line l j The i-th site The kth vehicle The arrival time of up and down trains, Respectively represent line l j The i-th site The kth vehicle The departure time of up and down trains, m, n, q are the number of lines in the network, the number of stations in the line, and the number of train departures on the line respectively; S2.3, load the road network train operation time information model T(L,V,TR) into the road network space directed graph constructed in S2.1, and obtain the road network spatiotemporal topology model as follows: H={O={φ(L,V),G(V,A)},T(L,V,TR)} Where H represents the road network spatiotemporal topology model, O = {φ(L,V), G(V,A)} represents the spatial topology model constructed in S2.1, and T(L,V,TR) represents the road network train operation time information model loaded into the spatiotemporal topology model; T(L,V,TR) includes the arrival and departure time of the train on the line, which can be expressed as: In the formula, the constraint condition is equivalent to the logical relationship between stations and lines in the road network N = {φ(L,V)} in S2.1; In a spatial directed graph, a constrained breadth-first algorithm is used to search for a set of feasible paths between OD pairs in the road network that satisfy the maximum number of transfers.

4. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: In S3, the mobile phone user travel trajectory reconstructed by mobile phone signaling data is used to calculate the initial probability of path selection. The specific steps are as follows: S3.

1. According to the base station location and time information in the passenger's mobile phone signaling data, query the longitude and latitude of the passenger's mobile phone signaling data, match it with the longitude and latitude of the subway station, and reconstruct the passenger's travel trajectory in the subway network. The steps are as follows: S3.1A, read passenger x mobile phone signaling data MS x The mobile base station number BSID corresponding to each signaling event; S3.1B, read the latitude and longitude data of the base station BSID from the basic information data of the mobile base station, determine whether it belongs to the subway network dedicated base station, obtain the passenger's location and determine whether it is in the subway network, and reconstruct the passenger's travel trajectory in the subway network, S3.1C, the travel trajectory of passenger x in the subway network based on signaling reconstruction is expressed as: J x :{(v o ,t1),…,(v i ,t i ),…,(v d ,t n )} Among them, v i represents the number of the ith station in the travel trajectory of passenger x, t i Indicates that after v i time; S3.

2. Matching the travel trajectory of mobile phone users with feasible paths between OD pairs of urban rail transit network, specifically including: The spatial similarity between the passenger's mobile phone signaling reconstruction trajectory and the feasible path on the corresponding OD pair is defined as follows: In the formula, represents the trajectory p reconstructed by the mobile phone signaling of passenger x x The feasible path p with OD w The spatial similarity of For p x With p w The number of common sites, Indicates p x The number of sites, Indicates p w The number of sites, γ is a positive constant; Calculate the spatial similarity between the passenger travel trajectory reconstructed based on mobile phone signaling and all feasible paths between OD pairs, and take the feasible path with the highest similarity to the reconstructed trajectory as the passenger's travel trajectory in the subway network; S3.

3. Based on the travel trajectories of the above passengers in the subway network, the number of specific mobile phone users on the feasible paths of the OD pairs is counted, and the initial probability of feasible path selection between the OD pairs is calculated as follows: Among them, r j represents the feasible path p for OD to w j The initial probability of path selection, n j Select a feasible path p for a specific mobile phone user passenger on OD pair w j The number of people, n w is the number of passengers with specific mobile phone users on OD pair w; In S3, the passenger's travel chain is represented as: In the formula, λ x represents the passenger travel chain, x represents a passenger entering the urban rail transit system, ta represents the transaction time when the passenger passes through the entry gate, and te represents the transaction time when the passenger passes through the exit gate; x v (tr) represents the time when passenger x boarded train tr at station v, tr(x v ) represents the time when passenger x gets off the train tr at station v; Based on the defined passenger travel chain, passengers are matched to trains through the acquired AFC data and train schedule data, as follows: S3.4, according to the obtained train schedule, define the train operation chain of path p: In the formula, λ p represents the train operation chain of path p, are the jth travel subchain of the train operation chain, and the travel subchain includes the starting point of the train at section j Departure time on and end point Arrival time on The train operation chain without transfer path contains only one train operation sub-chain, while the train operation chain with transfer path contains multiple train operation sub-chains; S3.5, define the similarity between the passenger travel time interval and the train operation time interval on the path, expressed as follows: In the formula, represents the time similarity between the travel chain of passenger x and the travel chain of train tr on path p, is the length of the intersection of the passenger travel time interval and the train operation time interval on segment j of path p, represents the travel time interval length of passenger x on segment j of path p, represents the length of the running time interval of train tr on section j of path p, and γ is a positive constant; S3.6, based on the passenger transaction data recorded by the automatic ticket vending and checking system x , extract the passenger's entry and exit OD information, and then search the feasible short path set P of the road network w , passengers are classified according to the feasible path set and the number of transfers based on OD, and passenger travel is divided into single path without transfer, single path with single transfer, single path with multiple transfers, and multi-path travel according to the number of transfers on the path; S3.7, for passengers on a single route without transfer, the trains taken by the passengers are matched using time similarity; S3.8, based on the non-transfer single-path passenger matching results obtained in S3.7, calculate the time from the passenger entering the station through the gate to boarding the train and the time from the passenger getting off the train to exiting the station through the gate, expressed as: In the formula, ACT represents the time from when the passenger enters the station through the gate to when he / she boards the train tr, and x v (tr) is the time when the passenger boards the train tr at station v, ta is the transaction time when the passenger passes through the entrance gate; EGT is the time when the passenger gets off the train and passes through the gate to exit the station, tr(x v ) is the time when the passenger gets off the train tr at station v, and te is the transaction time when the passenger passes through the exit gate; S3.9, to match single-transfer single-path passengers to trains, the following relationship needs to be satisfied between the passenger travel chain and the train operation chain: Where, t d (l,tr,v) represents the time when train tr on line l leaves station v, t a (l,tr,v) represents the time when train tr running on line l arrives at station v, ACT min (l,v) represents the minimum entry time of a passenger at station v on line l, EGT min (l,v) represents the minimum exit time of passengers at station v on line l, t d (l ′ ,tr ′ ,v) indicates the transfer line l ′ Transfer train running on tr ′ Time to leave site v, TFT min (l,l ′ ,v) indicates that at station v, passengers transfer from line l to line l ′ Minimum transfer time; Match single-transfer single-path passengers to trains, and analyze the transfer time between transfer stations and lines in different transfer directions; obtain the ACT by calculating the entry and exit time distribution of station passengers through S3.8 min (l,v) and EGT min (l,v); By analyzing the passenger's AFC transaction data and train schedule, and substituting them into constraint (2) in S3.9, the train that the passenger takes for the last segment of the travel sub-chain of the single-transfer single-path passenger is found, and the boarding time of the passenger's last segment of the travel sub-chain is obtained; Match the passenger to the train with the highest time similarity, and obtain the passenger's first travel subchain, thus obtaining the complete travel chain of the single-transfer passenger; The transfer time of passengers is calculated as follows: Where TFT represents the passenger transfer time, Indicates that the passenger boards line l at transfer station v ′ The nth train time, Indicates that the passenger transfers from the i-th train on line l at the transfer station v The time to get off the bus; assuming that the transfer time distribution of single-transfer passengers at the station represents the transfer time distribution of transfer passengers, the transfer time distribution of transfer passengers at the station is obtained by calculating the transfer time of single-transfer passengers in each transfer direction at the station.

5. A multi-level passenger flow state estimation system for urban rail transit network according to claim 1 or 4, characterized in that: In S3, the walking time for passengers entering and exiting the station and the walking time for transfers are estimated, and the mean headway time of trains at the station is calculated: in, For line l station v direction The average headway of trains, Respectively represent the running direction of line l station v The departure time of the k+1th and kth trains, K is the number of train departures; the average waiting time of passengers on the platform is half of the headway time, that is, in, Indicates line l station v direction The average waiting time of passengers is Indicates line l station v direction Average headway of trains; Assume that the standard deviation of passenger waiting time is one-fourth of the standard deviation of train headway time, that is, in, For line l station v direction The standard deviation of the waiting time of passengers, For line l station v direction The headway of the i-th train, For line l station v direction Average headway of trains on By estimating the distribution of passengers' entry and exit time, the distribution of passengers' walking time is obtained as follows: in, represents the distribution of passengers’ walking time into the station, N(μ ACT ,σ ACT ) represents the passenger arrival time distribution, It represents the waiting time distribution of passengers at the platform after entering the station. The relationship between the mean and standard deviation is: Calculate the distribution of passengers' transfer walking time, specifically: in, represents the distribution of passengers' transfer walking time, represents the distribution of passengers’ transfer time, represents the waiting time distribution of passengers at the transfer station, and the relationship between the mean and standard deviation is: Assuming that passengers get off at their destination and immediately walk to the exit gate, the passenger's exit walking time distribution is the exit time distribution. So far, we have obtained the distribution of passengers' walking time in and out of the urban rail transit system, the distribution of transfer walking time in the transfer direction of each transfer station, and the distribution of passengers' waiting time at each station.

6. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: The feasible path travel time is estimated in S3, specifically, Through the additivity of independent Gaussian distribution, the travel time distribution of the feasible path is obtained as the sum of the passenger's walking time in and out of the station, transfer walking time, waiting time, and time in the car: in, represents the travel time distribution of feasible paths, represents the walking time distribution of passengers entering the station. represents the transfer time distribution of path passengers, represents the waiting time distribution of the path, represents the time distribution within the vehicle along the path, It represents the walking time distribution of the route out of the station; The mean and variance of the feasible path travel time distribution are, The passenger's in-car time is the time from the train leaving the platform after the passenger gets on the train to the train arriving at the platform where the passenger gets off. The calculation formula is as follows: in, represents the i-th interval [v i ,v i+1 ] train running time, Indicates that the train is at the i-th station v i parking time; The train schedule data contains the arrival and departure times of trains at each station, and the running time of trains at adjacent stations is expressed as: t r (l,[v i ,v i+1 ],tr)=t a (l,v i+1 ,tr)-t d (l,v i ,tr) Among them, t r (l,[v i ,v i+1 ],tr) represents the line l from station v i Run to site v i+1 The running time of train tr; Train at station v i The parking time is calculated as follows: t l (l,v i ,tr)=t d (l,v i ,tr)-t a (l,v i ,tr) Among them, t l (l,v i ,tr) means that the train tr running on l is at station v i parking time.

7. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: In S3, the distribution of OD versus travel time is estimated based on the mixed Gaussian distribution, specifically, Define the definition of the correction ratio coefficient of the initial probability of path selection, which refers to the parameter set when the path selection ratio of all subway passengers is derived using the initial probability obtained from the mobile phone user signaling data, and is expressed as α; Calculate the OD vs. travel time distribution based on the Gaussian mixture model: Among them, α j and r j are the modified proportional coefficient of OD for the jth feasible path and the initial probability of path selection, is the travel time distribution of OD for the jth feasible path. Calibrate the path selection initial probability correction proportional coefficient, the specific method is as follows: 1). Calculate the KL divergence of the estimated and observed OD travel time distributions, Discretize the travel time distribution into K intervals according to a fixed time interval, and the KL divergence calculation formula is as follows: Among them, t k is the kth travel time length interval, K is the number of discretized travel time length intervals, and β(t k ) are the k-th discretization interval probabilities of the estimated and observed OD for travel time distribution respectively; 2). Correct the proportional coefficient to calibrate the nonlinear optimization model construction, Nonlinear programming is used to search for the optimal correction proportional coefficient. The objective function is the sum of the KL divergence of the estimated and observed travel time distributions of all OD pairs in the network. The formula is as follows: Among them, α is the correction coefficient matrix of OD to the path selection probability, and OD vs w i The kth discretized interval probability of the estimated and observed travel time distribution, Z is the number of network OD pairs, α i,j and r i,j Distribution is OD versus w i The modified proportional coefficient and initial probability of the jth feasible path in, Indicates OD to w i The travel time distribution of the jth path; OD for w i The expected value of the logarithmic function of the observed travel time distribution is minimized by minimizing Based on the path selection probability constraints, a nonlinear optimization model is constructed as follows: s.t.0≤α i,j ·r i,j ≤1,j∈J,w∈W The constraints in the model ensure that the corrected path selection probability is between 0 and 1, and the sum of its probabilities is 1; 3). Solving the optimal correction proportional coefficient The BFGS algorithm is used to solve the nonlinear optimization model and obtain the optimal correction proportional coefficient α * , the formula is: Estimate the path selection probability between all pairs of network ODs, the formula is as follows: p * =a * ·r Among them, p * represents the path selection probability matrix between all OD pairs in the network, and r represents the initial probability matrix of feasible path selection between OD pairs.

8. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: In S4, BP neural network is used to estimate the OD passenger flow of urban rail transit network. The monthly historical AFC transaction data is used as the training set to train the BP neural network model, and the trained BP neural network is used to estimate the real-time OD passenger flow of the network. At the same time, the passenger exit transaction time is calculated based on the mean of the feasible path travel time distribution between OD pairs estimated in S3 and the passenger entry transaction time in the AFC transaction record, and the real-time passenger entry and exit transaction time set between OD pairs is obtained. The probability of selecting the OD feasible path is used to calculate the passenger flow demand on the OD feasible path as follows: i u (p)=θ u ×p * Among them, θ u (p) is the estimated real-time OD passenger flow demand within the analysis period Δt of the feasible path p, p * is the selection probability of the feasible path p; from Random sampling θ u (p) passengers to the feasible path p, calculate the passenger's exit time based on the passenger's entry and gate time and the feasible path travel time estimated in step S3, and then The sampled passengers are removed from the calculation and the passenger flow demand on the OD feasible path is returned to be calculated until all passengers are assigned to the feasible path.

9. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: The passenger travel chain reconstruction algorithm in S5 includes the reconstruction of the travel chains of passengers without transfer and passengers with transfer, specifically, S5.1, Initialization: Load the network directed graph G(V,A) constructed in S2.1, the feasible path set of network OD pairs searched in S2.3, and the network OD passenger flow demand within the input analysis period Δt S5.2, allocating the network OD passenger flow according to the OD passenger flow allocation method for feasible paths in step S4; S5.3, perform the operations in steps S5.4-S5.6 for each OD pair ω in the network OD pair set; S5.

4. Perform the operation in step S5.5 or S5.6 for the passengers on the feasible path p in the OD pair; S5.

5. If the feasible path p is a path without transfer, calculate the time similarity between the passenger travel chain and the train operation chain Match passengers to those with the highest time similarity train, thus we can get the travel chain of complete OD passengers on the OD pair ω non-transfer path and store it in the set middle, represents the set of passenger travel chains on the OD pair ω without transfer path p; S5.

6. If the feasible path p is a transfer path, and the number of transfers in the feasible path is Then the number of travel sub-chains in the passenger travel chain is Match each passenger's travel subchain with the train. Assume that The passengers in the trip sub-chain board the first train that arrives when they are waiting at the platform, and assign the passengers to that train. The time similarity between the passenger travel subchain and the train operation chain is calculated Match passengers to those with the highest time similarity The train of OD is obtained, thus obtaining the travel chain of complete OD passengers on the OD to ω transfer path and storing it in the set middle, represents the set of passenger travel chains on the OD pair ω transfer path p; S5.

7. So far, the set of travel chains of all passengers in the urban rail transit network within the analysis period Δt is obtained. in They represent the travel chain set of passengers without transfers in the network within the analysis period Δt and the travel chain set of passengers with complete OD transfers respectively. The travel chains of all passengers in the network within the analysis period Δt are output, and the algorithm ends.

10. The multi-level passenger flow state estimation system for urban rail transit network according to claim 1, characterized in that: In S6, the number of passengers getting on and off the train is counted, and the real-time passenger capacity of the trains in the network is counted. The train passenger capacity calculation model is constructed as follows: In the formula, Indicates time state I j The passenger capacity of the following train tr at station v is, express The passenger volume of the train at the previous state, Indicates that in time state I j Number of passengers boarding the train tr at the next station v, Indicates that in time state I j Number of passengers getting off at the next station v, train tr, The aforementioned time state I j Represents the jth prediction time interval, which contains the start and end time of the time state; In S7, the real-time network section passenger flow at different time granularities is estimated, and the network section passenger flow calculation model is constructed as follows: Calculate the section passenger flow based on the estimated real-time passenger capacity of the line network trains in S6 and define the granularity Then, within the analysis period Δt, it is necessary to estimate The cross-sectional passenger flow of time intervals, the i-th time interval in the analysis period Internal line l station v n-1 To site v n-1 Travel direction The passenger flow of the upper section is determined by the time state I j exist Site v n-1 The sum of the train passenger capacity is obtained, in, Indicates time granularity Next, analyze the time interval Inside, line l station v n-1 To site v n-1 Section Passenger flow in driving direction dr, Indicates line l station v n-1 Time Status I j The train under tr k passenger capacity; In S8, the number of passengers boarding the train, the number of passengers arriving at the platform, and the number of passengers waiting on the platform are counted, and the calculation model for the number of passengers waiting on the platform is constructed as follows: In the formula, Indicates time state I j The number of people waiting on the platform at the next stop v, line l, direction r, express The number of people waiting on the platform at the previous time, Indicates that in time state I j Number of passengers boarding the train tr at the next station v, Indicates that in time state I j Enter from station v and arrive at line l The number of passengers on the platform, The aforementioned time state I j Represents the jth prediction time interval, which contains the start and end time of the time state; In S9, the number of passengers in the transfer channel is inferred based on the spatial and temporal positions of getting on and off the bus in the transfer passenger travel chain and the passenger transfer walking time, specifically: Read the passenger travel chain from the complete OD passenger travel chain set, and extract the transfer and alighting time tr(x v ), and at the same time extract the transfer walking time of passengers in the driving direction of the corresponding station from the estimated passenger travel time distribution in step S3 Assuming that the time it takes for a passenger to walk from the platform to the transfer passage is half of the time it takes to transfer, then passenger x walks to transfer station v and transfers from line l to line l ′ direction The time for the upper transfer channel is Given the time granularity Under the condition, the i-th time interval in the analysis period Δt Transfer station v transfers from line l to line l ′ direction The number of passengers in the upper transfer channel is: in, Indicates a given time granularity Under the condition, the i-th time interval in the analysis period Δt Transfer station v transfers from line l to line l ′ direction Number of passengers in the upper transfer channel, Indicates the time interval Passengers gather in the inner transfer channel, Indicates time interval The cardinality of the set of passengers in the internal transfer channel, that is, the number of passengers in the transfer channel.

Images