MetroTrainOperationPlanAnalysisBasedonStationTravel … · 2020. 7. 18. · of metro network...

Research ArticleMetro Train Operation Plan Analysis Based on Station TravelTime Reliability

Ruihua Xu12 Fangsheng Wang 12 and Feng Zhou 12

1Key Laboratory of Road and Traffic Engineering of the Ministry of Education Tongji University Shanghai 201804 China2Shanghai Key Laboratory of Rail Infrastructure Durability and System Safety Tongji University Shanghai 201804 China

Correspondence should be addressed to Fangsheng Wang wangfangshengtongjieducn and Feng Zhouzhoufeng24tongjieducn

Received 18 July 2020 Accepted 3 April 2021 Published 15 April 2021

Academic Editor Zhixiang Fang

Copyright copy 2021 Ruihua Xu et al -is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

-e train operation plan plays an essential role in metro systems and directly affects transportation organization efficiency andpassenger service level In metro systems passengers have paid more attention to the travel time reliability (TTR) reflecting thereliability of metro operation management -is article proposes an analysis method of train operation plan based on TTR in thestation dimension First an automated fare collection (AFC) data-driven framework is established to calculate the station traveltime reliability (STTR) and analyze the train operation plan at different periods -e framework structure consists of four stepsAFC data preprocessing STTR calculation and assignment clustering algorithm design based on SOM neural network and trainoperation plan analysis and optimization Second the proposed method is applied to the Beijing metro network as a case studySeveral promising results are analyzed that allow the optimization of the existing train operation plan Our research shows thatSTTR is a good supplement for the existing metro operation assignment studies which can help analyze and optimize the trainoperation plan effectively -is study is also applicable to other metro networks with AFC systems

1 Introduction

With the ongoing socioeconomic development urban trafficcongestion has become increasingly severe especially in largecities like Beijing and ShanghaiMetro is playing an increasinglyimportant role in urban public transportation owing to theoutstanding advantages of faster velocity higher reliability andlarger capacity With the continuous expansion of the scale ofmetro networks passenger demand shows a high-speed growthwhile the distribution of passenger demand presents the un-balance characteristics in the time-space dimension-ere is anincreasingly prominent contradiction between the trans-portation capacity supply and the passenger flow demand andit puts forward higher requirements for train operation orga-nizations under networked operating conditions

As a significant part of the metro operation and manage-ment the train operation plan directly affects transportationorganization efficiency and passenger service level In manylarge cities train running intervals are continuously shrinking

during the morning rush hours while some stations are stillhighly congested Passengers have particular travel character-istic which generally concentrates in individual stations orperiods Restricted by objective conditions of metro networkstructure metro transport capacity cannot meet the passengerflow demand at specific locations and periods resulting insevere partial congestion in the network -e fundamentalreason is that the configuration of network transport capacitydoes not match the distribution of passenger travel needs in thespace-time dimension

Recent research focused on extracting relevant indexesto reflect the train operation plan quality such as train full-load rate [1ndash4] and platform congestion degree [5ndash7] Onone hand the existing research methods screen the topbottom ranking sectionsstations according to the operationindicators including section full-load rate and stationpassenger volume and essentially sort operation indicatorsand get the concerned sections or stations but cannot obtainthe potential causes On the other hand passengers have

HindawiJournal of Advanced TransportationVolume 2021 Article ID 8813461 13 pageshttpsdoiorg10115520218813461

paid more attention to the travel time reliability (TTR) inpublic transportation and TTR has been one of the mostsignificant factors affecting transportation services level[8 9] In general passengers will always ride the first train toarrive after they reach the platform unless the train is toocrowded When the transportation capacity cannot meet thepassenger demand in some stations and sections there willbe a backlog of passengers waiting in station platforms andthus there will be a direct impact on TTR As for an OD pairin the metro network TTR typically has two definitions (1)the probability that passengers can complete a trip within aspecified time (2) the fluctuation degree of the average traveltime of passengers

As the node for passengers to start and finish the journey inthe train operation plan stations are the core of transportationorganizations in the metro system Based on the characteristicsof metro network structure and operation management wepropose the definition of station travel time reliability (STTR) asthe fluctuation degree between the actual time and standardtravel time of eachOD from this station as the starting station toother stations Based on the support of big data STTR analyzesand evaluates the TTR of inbound passenger flow totally toreflect the passenger service level at different stations and pe-riods Combined with operation experience and travel inves-tigations the factors that affect the fluctuation of STTR valueconsist of the following three aspects

(i) Passenger flow of the station is excessive(ii) Train running interval of the line is large that is the

transportation capacity is insufficient(iii) Station location trains are too crowded when ar-

riving at the station because their capacity has beenused in front of this station

To sum up the train operation plan analysis should not belimited to the ranking of indicators but also should pay at-tention to the analysis of potential causes -is study aims todevelop a data-driven approach to analyze train operationplans based on the STTR of all stations in the network -econtributions of this article are as follows

(1) Based on the AFC data an STTRmeasurement modelis built to calculate the value of passenger TTR fromstation dimension and principal component analysis(PCA) is used to process clustering elements

(2) Combining the Self Organizing Maps (SOM)neural network a station clustering framework isestablished with the STTR values and influencefactors to analyze the train operation plan moreobjectively and comprehensively and explore thespecific reasons for low STTR level

(3) Apply the proposed approach to the Beijing metro asa case study and several results are analyzed thatinspire the optimization of the existing train oper-ation plan

2 Literature Review

Numerous studies in the literature related to the train op-eration plan analysis consist of the following aspects

operation organization [1ndash4] and station service level [5ndash7]Li et al [1 2] constructed an interaction model of trains andpassengers and obtained evaluation indexes such as full-loadrate number of passengers and average waiting time andoptimized the train operation plan based on the matchingdegree of capacity supply and passenger demandWang et al[3] evaluated the adaptability of train operation schemes andpassenger demand from three aspects total adaptabilitystructural adaptability and quality adaptability Lu [4] di-vided transportation efficiency into three levels capacityoutput efficiency capacity utilization efficiency and trans-port demand satisfaction efficiency Tian [5] used the pas-senger flow aggregation and congestion as an indicator tomeasure the service level of the station and as one of thebases for the preparation of the train operation plan Liu andChen [6] used the minimization of factors such as thewaiting time of passengers at the station as the objectivefunction to establish a multiobjective nonlinear mixed-in-teger optimization model evaluates and optimizes the lineoperation plan Shafahi and Khani [7] considered theminimum transfer waiting time as the goal and combinedheuristic algorithms to optimize the transfer of the roadnetwork We find these analysis methods focused on re-storing the passenger travel process to extract relevant in-dexes such as train full-load rate platform waiting timeHowever in the process of path restoration the parameterssuch as the passenger walking time and the train maximumpassenger capacity will have a few differences and fluctua-tions in the space-time dimension -e pattern of empiricalvalues for these parameters will cause vast subjectivity andrandomness in evaluation results

-e theory of TTR was first proposed on urban roadtraffic and there are several types of research about the TTRanalysis in public transportation Considering travel be-havior analysis in the road network Asakura and Kashi-wadani [10] gave a concept of TTR the probability thatpassengers can complete the trip within the specified timeand measured TTR of an OD pair in a deteriorated roadnetwork [11 12] Lam and Xu [13] calculated TTRrsquos value byestablishing a traffic flow simulator model and access thereliability of metro systems organization management Belland Chirs [14] analyzed travel time change based on sen-sitivity analysis and described TTR by travel time varianceWhile some scholars [15 16] used the buffer time index(BTI) to describe the TTR BTI is the fluctuation degreebetween the actual and planned travel time at a specifiedperiod Besides Lomax et al [16] defined the unit distancetravel time and defined the BTI as the rate between theaverage travel time and the time of passengers having a 95chance of arriving at the destination on time

To our knowledge little attention has been paid to in-troducing TTR to the train operation plan analysis in thestation dimension which is of great significance to the metrooperation management Zhang et al [17] presented a newunit distance TTR evaluation index and method to assess theBeijing metro network Li et al [18] proposed a TTR cal-culation algorithm to analyze the reliability of transfer timequantitatively Chen [19] proposed the definition andevaluation method of metro network operation reliability

2 Journal of Advanced Transportation

and established a train operation delay propagation modelBased on the data-drivenmethod [20 21] this article focuseson calculating the STTR and analyzing and optimizing thetrain operation plan combined with the clusteringalgorithm

3 Data Description

31 AFC Data -e study addressed in this article requirespassenger travel time data extracted from the automated farecollection (AFC) data -e AFC system has become theprimary method of collecting metro fares in many citiesthroughout the world AFC system provides a large quantityof passenger flow information recording passengersrsquo ac-tivities with original station ID destination station ID tap-intime and tap-out time Necessary elements for the modelformula are summarized (Table 1)

32 Train Diagram -e train diagram illustrates the rela-tionship between space and time for train operation (Fig-ure 1) Necessary elements for the model formula aresummarized (Table 2) According to the train diagram datawe can extract each linersquos running interval at different pe-riods for the model formula

4 Methodology

As mentioned the existing analysis methods emphasizescreening the topbottom ranking sectionsstationsaccording to the operation indicators which are essentiallythe index ranking methods However an increasing numberof researchers and professionals have identified shortcom-ings in traditional analysis methods For example theseindicators may be subject to bias and error in evaluationresults Moreover the manual methods usually only focus ongetting the concerned sections or stations but cannot obtainthe potential causes For these reasons alternative conceptsand methods need to be developed -is article proposes acluster-driven method for analyzing the train operationplan consisting of four steps AFC data preprocessing STTRmeasurement model cluster-based analysis method andtrain operation plan optimization

Step 1 AFC data processingInput AFC data calculate the lower and upper bound ofeach OD pairrsquos travel time thresholds and removeabnormal records that are not between the lower andupper bound of thresholdsStep 2 STTR measurement modelBased on the Cumulative Chance Measurement Model(CCMM) calculate the values of STTR (NSTTR andLSTTR) by the actual and standard of travel time andPCA is used to process clustering elementsStep 3 Cluster-based analysis methodIntroduce SOM neural network to clustering algorithmfor station classification and explore the specific rea-sons for low STTR levelStep 4 Train operation plan optimization

By combining the level of STTR and influencing fac-tors including passenger flow train running intervalsand station location coefficient analyze stationscharacteristics of different clusters and design ap-propriate optimization measures in train operationplans for low-reliability stations and lines For theconvenience of model formulation relevant sets andparameters are listed in Table 3

41 AFCData Processing In general passengersrsquo travel timebetween the same OD will be within a reasonable sectionTypically the threshold of the route travel time is deter-mined by the results of travel surveys First obtain the actualtravel time set of each OD by extracting each passengertravel time from the network AFC ticket dataset Passengertravel time is the difference between the passengerrsquos tap-intime and tap-out time in the smart card Secondly sort theactual travel time data for each OD pair in ascending order-e lower and upper bound of the travel time threshold ofthe OD (Station i to Station j) are obtained from the fol-lowing formulae

tlowij tceil 5lowastMij( 1113857

tupij min t

lowij lowast (1 + a) t

lowij + U1113872 1113873

(1)

where tlowij is the lower bound of the travel time threshold tupij

is the upper bound of the travel time threshold tceil(5lowastMij) isthe actual travel times value for the fifth percent [22] Mij isthe number of passengers a is the relative threshold coef-ficient U is the absolute threshold

-e values of a and U are determined through travelsurveys normally a is 06 and U is 20 minutes [22] -enthe data with the actual travel time at [tlowij t

upij ] are retained

and the noise data are removed for each OD travel time set

42 STTRMeasurementModel -emeasure indicating TTRincludes two types probability and fluctuation -e formerindicates the probability that the passenger could completethe trip within the specified time and the latter reflects thefluctuation degree between the actual and planned traveltime -e study in this article focuses on the quantitativerelationship between passenger travel time and train op-eration plan so that we decide to use the fluctuation indicatoras the basis of the model

As distinct from manual methods the proposed methodintegrates multiple indicators (STTR passenger flow trainrunning intervals geographic location etc) for clusteranalysis and classification of stations -us through theanalysis of various categories we can evaluate the operationeffect of the train operation plan of stations and lines-erefore we propose a measurement model to calculate theSTTR value and analyze the correlation between STTRvalues with these factors and provide the basis for clusteranalysis in the next charter

Firstly the lower bound of the travel time threshold(tlowij ) is used as the standard travel time (tstandij ) of the ODand the passengersrsquo TTR of one OD pair (station i to station

Journal of Advanced Transportation 3

j) is measured by the average and standard of travel time asshown in equation (2)

TTRij taveij minus t

standij1113872 1113873

tstandij

(2)

where taveij is the average travel time between i and station jWe divide STTR into Network STTR (NSTTR) and Line

STTR (LSTTR) NSTTR is the relationship between thisstation and all other stations in the network whereas LSTTRis the relationship between this station and all other stationsin the same line Based on the CCMM presented in TTRstudies [23] we measure the STTR (NSTTR LSTTR) ofstation i as shown in equation (3)

NSTTRi 1113936jisinSjnei TTRij lowast t

standij lowastMij1113872 1113873

1113936jisinSMij

LSTTRi 1113936jisinSljnei TTRij lowast t


1113936jisinSiMij

(3)

where S is the stations set of the metro network and Si is thestations set of the line which station i belongs to

Secondly according to the previous analysis in thisarticle the influencing factors for the STTR level includepassenger flow train running intervals and station locationso that we use the three influencing factors and the values ofNSTTR and LSTTR as clustering elements

(i) Passenger flow the inbound passenger flow of thestation that is the total OD passenger flow with thestation as the departure station during this period

(ii) Train running intervals the train operation planrunning intervals of the line where the station lo-cates during this period

(iii) Station location coefficient analyze the geographiclocation of all stations in the network and extract

the central station and set its station location co-efficient as 0 and the station location coefficients ofother stations are determined by the OD standardtravel time from it to the central station

(iv) NSTTR(v) LSTTR

-ese five elements are the add-in values in this model-e first three are derived from the passenger flow statisticssystem train operation plans and geographical statisticsdata Moreover the calculations of NSTTR and LSTTR aredirectly related to AFC data Traditional evaluation methodsfocus on the calculation and simple ranking of indicatorsbut the specific causes of the station or linersquos poor indicatorsare not enough

-irdly we use PCA to reduce the clustering elementsrsquodimensional reduction and analyze the correlation betweenSTTR values with these factors As a multivariate statisticalmethod based on orthogonal transformation PCA indexesmultiple related variables of the research object into a fewunrelated variables and retains feature vectors with signif-icant contributions [24] -ese unrelated comprehensivevariables include most information provided by the originalvariables thereby achieving dimensionality reductionSpecific steps are as follows

Step 1 NormalizationScale clustering elements to a normal distribution witha mean of 0 and a variance of 1Step 2 Correlation coefficient matrixCompose the normalized clustering elements into a 5-dimensional random vector

X x1 x2 x3 x4 x5( 1113857 (4)

where the covariance of xm and xn is the correlationcoefficient of them namely

pmn Cov xm xn( 1113857 (5)

-e correlation between xm and xn is

(i) positive correlation when pmn gt 0(ii) negative correlation when pmn lt 0(iii) irrelevant when pmn 0

-e larger the absolute value of pmn the stronger thelinear correlation of xm and xn Finally obtain thecorrelation coefficient matrix D(X) of XStep 3 Principal components extractionExtract the feature root of D(X) and convert it to thecorresponding standard feature vector μk which is the

Table 1 Necessary elements of AFC data

Date Original station ID Tap-in time Destination station ID Tap-out time Smart card no20161018 000521 07 15 00 001313 09 19 15 20014199520161018 000551 10 04 56 000429 10 21 20 20014199620161018 000535 10 02 32 000559 10 45 23 200141997

730Station1

Station2

Station3

Station4

Station5

830 930900800

Figure 1 Example of actual train diagram


contribution rate of the main component Zk and se-quentially extract Z1 Z2 Zr Moreover the cu-mulative contribution rate of these principalcomponents reaches the specified threshold which is70 generally-erefore previous approaches focus on index calculationand ranking screens the topbottom ranking sectionsstations according to the operation indicators Comparedwith the traditionalmeasurementmethods about TTR themodel proposed in this article has the following differ-ences (1) the evaluation index (STTR) in this model isdirectly calculated by AFC data and does not need pathrestoration (2) because of the metro network complexityand OD quantity diversity this model calculates TTRvalues from station dimension and divides them into twolevels of network and line (3) by integrating analysis withother factors the model analyzes the correlation betweenSTTR values with these factors and provides data supportfor cluster analysis method and train operation planoptimization in the next charters

43 Cluster-Based Analysis Method In this section we willidentify the stations with low reliability and provide somesuggestions for improving the STTR level Clusteringanalysis is commonly used to categorize large amounts ofdata Considering different clusters tend to show distinctdifferences in the clustering analysis results and the ab-normal points can help distinguish the potential outliers inthe data In this article by analyzing the different partsdetermined by cluster analysis we can identify low-reli-ability metro stations and propose optimization suggestionsin the train operation plan

We use the SOM neural network to categorize andanalyze stations Based on the values of principal

components the stations with higher similarity are in thesame group and the attributes of these stations are con-sidered to be the same As an unsupervised learning neuralnetwork SOM [25] has strong self-organization charac-teristics and only an input layer-competitive layer (Figure 2)

Compared with the K-means clustering algorithm theadvantages of the SOM neural network include thefollowing

(i) Not affected by the initialization of the clustercentroid

(ii) Improving the processing ability of nonlinear data(iii) Reducing the influence of noise data

-e SOM neural network cluster algorithm includes thefollowing parts

431 Determine the Number of Clusters We use SilhouetteCoefficient (SC) method to determine the number of clustergroups that is the number of station categories -e SCmethod combines the clustering degree of Cohesion andSeparation -e Cohesion refers to the average distancebetween the sample point i and all other elements in thesame cluster denoted as a(i) -e Separation means theaverage distance between the sample point i and the points inthe other cluster traversing other clusters to obtain theminimum value denoted as b(i) the cluster is the neighborcluster of i -e sample point i contour coefficient is

s(i) b(i) minus a(i)

max a(i) b(i) (6)

-e larger the average of all stationsrsquo contour coefficientthe better the number of clusters

Table 2 Necessary elements of train diagram

Date Train ID Destination station ID Stop station ID Stop order Arrival time Departure time20161018 0429 000941 000933 5 07 12 15 07 12 4520161018 0431 000941 000933 5 17 15 20 17 15 5020161018 0103 000521 000545 13 10 45 23 10 45 53

Table 3 Relevant sets and parameters in model formulation

Setsparameters DefinitionMij Passengers number of OD (station i to station j)a -e relative threshold coefficientU -e absolute thresholdtlowij Travel time threshold lower bound of OD (station i to station j)

tupij Travel time threshold up bound of OD (station i to station j)

tstandij -e standard travel time value of OD (station i to station j)tceil(5lowastMij)

Actual travel times value for the fifth percenttaveij Travel time average value of OD (station i to station j)TTRij Travel time reliability value of OD (station i to station j)S Stations set of the metro networkSi Stations set of the line which station i belongs toNSTTRi -e weighted average TTR value between station i and all other stations in the networkLSTTRi -e weighted average TTR value between station i and all other stations in the same line


432 SOM Network Initialization Import the principalcomponent data of each station into the input layer of theSOM neural network -e data format is

Uk uk1 u

k2 u

kN1113960 1113961

T (7)

where k 1 2 r r is the number of principal componentsand N is the number of stations

We construct the initial neuron network of the com-petition layer and the weight vector expression of theneuron node j and the input layer data is

Wj wj1 wj2 wjs1113960 1113961T (8)

where j 1 2 M M is the number of neuron nodes in thecompetition layer

433 Competitive Learning in SOM Network SOM neuralnetwork adopts the method of competitive learning duringthe training process Each input data point finds a node thatmatches it best in the competitive layer called its activationneuron (WN) -en use the stochastic gradient descentmethod to update the parameters of the active node and thedata points it covers -e competitive learning process in-cludes the following steps

Step 1 -e initialization parameters of the competitionlayer nodes have the same parameter dimensions as theinput layer data dimensionsStep 2 According to the Euclidean distance matchpoint i of the input layer to the nearest node(WN) inthe competitive layer

ui minus WN

min ui minus wj

1113882 1113883 j 1 2 M (9)

Step 3 Set WN as the center the connection weightsbetween other neurons in the neighborhood of thecompetition layer and the input layer neurons aremodified

wj(t + 1) wj(t) + hcj(t) Xj(t) minus wj(t)1113960 1113961

j 1 2 M jne c(10)

where t is the number of iterations wj(t) is the con-nection weight of the node j and the input layer at the

moment t Xj(t) is the input sample vector of the nodej at the moment t and hcj(t) is the neighborhoodkernel function of WN at the moment t namely

hcj(t) exp minusd2cj(t)

2lowast δ2(t)⎛⎝ ⎞⎠ (11)

where d2cj(t) is the lateral distance between neuron j

and WN and δ(t) is the amount of network width atthe moment t that is

δ(t) δ(0)lowast exp minusnlowast log δ(0)

10001113888 1113889 (12)

where δ(0) is set as the radius of the initial gridStep 4 Update the node parameters until the featuremap gradually converges -e neurons in the SOMcompetition layer continually iterate and cluster si-multaneously to divide the stations covered by eachneuron into groups

44TrainOperationPlanOptimization After these steps weobtain several station clusters Combining the values ofNSTTR and LSTTR we can analyze the characteristics of allclusters and identify the stations with low reliability Bycombining passenger flow analysis train running intervaland station location we could put forward several sugges-tions for train operation plans from these three aspects

Take Line X as an example as shown in Figure 3 there isonly a long routing with the train running interval being 4minutes in the train operation plan of Line X

And the measures we could apply include the following

(1) Minify train running interval It is the most con-venient method to improve the STTR level of allstations by increasing the number of trains per houras shown in Figure 4

(2) Adopt the long-short routing operation mode Asshown in Figure 5 if the stations with low-reliabilityconcentrate in a certain section (Station C to StationB) a short routing can be introduced

(3) Fare incentives or congestion alerts

When the train operation planrsquos transport capacity isclose to saturation we can adopt some other measures toencourage passengers to choose other routes to the desti-nation station In metro systems fare incentives areemerging as a method to manage peak-hour congestionincluding two strategies a time-based fare incentive strategy(TBFIS) and a route-based fare incentive strategy (RBFIS)With the development of science and technology passengerscan be reminded of the congestion in some stations andsections in real-time through mobile apps or large screens inthe stations to switch paths in time

5 Case Study on Beijing Metro

51 9e Network and Existing Analysis Method In thissection the quality of methodology will be illustrated using a

X1 X2 X3

Input layer

Competitive layer

Figure 2 -e structure of SOM neural network


real case of the Beijing metro system In 2016 its networkconsisted of 18 lines and 326 stations (Figure 6) and therewere more than 6000000 daily trips on average

Over the past few years the Beijing metro system hasdeveloped rapidly and is now one of the worldrsquos largestAccording to the aggregation and dissipation of passengerflow we divide the area inside the red frame into urbandistricts Divide the area outside the red frame into suburbandistricts In this article we use a total of 39453138 AFCrecords on weekdays calculate all the results by C Net andPLSQL database programming According to tap-in time inAFC data we divide the study period into three parts (1)morning peak periods 6 00ndash10 00 (2) off-peak periods10 00ndash16 00 (3) evening peak periods 16 00ndash20 00

-e train operation plan analysis method currentlyemployed by the Beijing metro system is the index rankingmethod -is method screens the topbottom ranking sec-tionsstations according to the operation indicators in-cluding section full-load rate and station passenger volume-e existing method is essentially to sort operation indi-cators and get the concerned sections or stations

52 General Analysis First we calculate the values ofNSTTR and LSTTR of each station at different periods basedon the STTRmeasurement model -e values of NSTTR andLSTTR reflect the STTR level Figure 7 shows the visuali-zation of NSTTR and LSTTR values of all stations at differentperiods the darker the color the bigger the NSTTRLSTTRvalue that is the lower the STTR level

-e values of NSTTR and LSTTR present a significantdifference in the space-time dimension In the morningpeak the urban stationsrsquo NSTTR and LSTTR values aresmall while the suburban stationsrsquo values are larger butthere is the opposite in the evening peak Furthermore thereis a relative balance in the off-peak periods

-en we select six lines with huge passenger demandthe NSTTR and LSTTR values are shown in Figure 8 and the

stations at the red frame in the horizontal axis are urbanstations -e results show that the LSTTR value of moststations is slightly lower than the NSTTR value and theirchanging trend is nearly consistent in one line In themorning peak the values of NSTTR and LSTTR in urbanstations are smaller that is the STTR level of urban stationsis higher than suburban stations generally

53 Clustering Analysis Based on the general analysis wechoose the morning peak as the study period for clusteringanalysis First we obtain each stationrsquos passenger demandfrom the AFC data and calculate the running interval of eachline (Table 4) from the train diagram data According to theBeijing metro networkrsquos geographical location analysis weregard Tiananmen West Station as the central station of thenetwork as the red pentagram in Figure 6 -en we cal-culate the location coefficient of all stations

531 Clustering Elements Processing We use SPSS statisticalsoftware for the PCA on cluster elements and the corre-lation coefficient matrix of cluster elements is shown inTable 5 -ere are positive correlations between the NSTTRLSTTR and passenger flow train running intervals andstation location coefficient

-e principal components with the top two rankings areextracted and the cumulative variance is 768 As shown inthe principal component matrix (Table 6) the principalcomponent PC1 represents mainly passenger flow trainrunning intervals and location coefficient while the prin-cipal component PC2 represents mainly NSTTR andLSTTR

532 Clustering Results First we use the Contour Coeffi-cient Method [26] to determine the optimal clusteringnumber as 4 -en we use MATLAB software for clusteranalysis -e initial network of the SOM neural network is a6lowast 6 neuron network During the competitive learningprocess each neuron updates its position and stationsconnected to it -en the categories of stations in thenetwork are classified As shown in Figure 9 the final po-sitions of all neurons are in the red network and the blackpoints are the station points-e number of stations coveredby each neuron is shown in Figure 10

Finally the station clustering results and the distributionin the network are shown in Figures 11 and 12 where thesame color points are the stations of the same group and thesize of the station shape in Figure 11 is proportional to thepassenger volume

Based on the above analysis we analyze each stationrsquosgroup characteristics

Cluster 1 In Cluster 1 the values of PC1 and PC2 are lowthat is the levels of STTR and influence factors are high-ese stations are distributed mainly in the urban districtstheir passenger demand stress is weak and transport ca-pacity supply is high -ese stations do not need to improvethe train operation plan

Station A Station B

15 trainshour

Running interval 4mins

Figure 3 Original train operation plan of Line X

Station A Station B

20 trainshour


Figure 4 Minify train running interval


Cluster 2 In Cluster 2 the values of PC1 are low whilePC2 is high that is the STTR levels are high but in-fluence factors are weak -ere is a high matching oftransport capacity and passenger demand but a weaklevel of STTR in these stations -e representative linesand stations are Line YZ and Line FS (East) Regard thesestations as potential stations that need attention

Cluster 3 In Cluster 3 the values of PC1 are high while PC2is low that is the STTR levels are weak but influence factorsare high -e representative lines and stations are Line4(South) Line9 (South) nad Line BT

Cluster 4 In Cluster 4 the values of PC1 and PC2 are highthat is the levels of STTR and influence factors are weak-e representative lines and stations are Line CP(North)Line8(North) Line5(North) Line15 (Northeast) andLine14 (West)

54 Train Operation Plan Analysis Considering the distri-bution of each cluster station in the metro network in thischarter we focus on the lines and stations in Cluster 3 andCluster 4 and divide the potential causes of weak STTR intothe following three aspects

(1) Passenger flow stress is strong-ere is intense passenger flow stress in Line5 (North)Line8 (North) Line6 (East) and Line BT -ese linesand stations may require minifngyi train runningintervals Furthermore introduce the additional cus-tom buses to divert commuter passenger flow

(2) Train running interval is large-ere is a weak STTR level in Line4(North) becausethe train running intervals in these stations are240 seconds while those of other stations of Line4are 120 seconds thus Line4(North) requires

Station A Station C Station B

15 trainshour5 trainshour

Running interval 4minsRunning interval 3mins

Figure 5 Long-short routing operation mode

Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1

Line 14 (west)Line 9

Line FS

Line 4Line 8

Line YZ

Line 10

Line 16

Line 2Line 14 (east)

Line 7 Line BT

Line 6

Line 15

NORTH

Figure 6 Beijing metro network


Morning peakperiods

Off-peakperiods

Evening peakperiods

NSTTR value LSTTR value

High level

Low level

0200s

400s

600s

gt600s

Figure 7 -e visualization of NSTTR and LSTTR

Line 1 Line 4 Line 5

Line 6 Line 15Line 8

450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

Figure 8 -e values of NSTTR and LSTTR of six lines


minifying train running intervals Similar lines andstations need to minify train running intervals in-cluding Line CP Line14 (West) and Line15 and it isalso appropriate to adopt the long and short routingoperation mode for them

(3) Station location problemsAs shown in Figure 13 Line FS is located in thesouthwest of Beijing and has only a transfer station(GGZ) which connects with Line9 GGZ is a ter-minal station of the two lines which means thepassengers of Line FS who want to go to urbandistricts must pass through Line9

According to the AFC data in the morning peak theproportion of transfer passengers in the Line FSrsquos total

passenger flow is 827 -us the STTR level of Line9 isweak while the passenger flow stress is weak and thetransport capacity supply is adequate in Line9

Minify the running intervals of Line9 (South) appro-priately to increase the transport capacity and reduce theimpact from Line FS to Line 9 At the same time Line9 isaffected by the transfer passenger flow in the west section ofLine 14 at QLZ and intensify the passenger flow pressure ofLine9 Specific strategies including fare incentives or con-gestion alerts can be used to encourage more passengersfrom Line14 (West) to choose to transfer to Line 10 atStation XJ instead of Line9 at Station QLZ so that it canreduce the transportation pressure of Line9

55 Comparison and Analysis Take the existing analysismethod using station passenger volume to compare it withthe proposed approach in this article Figure 14 shows thetop 20 stations with inbound passenger volume -e X-axisrepresents the station name Y-axis represents the inboundpassenger volume the color represents the stationrsquos clusterin Figure 11 It can be seen that most of these stations belongto Cluster 4 (red) and Cluster 2 (green)

In the proposed approach the TTR and other influ-encing factors are integrated and analyzed by SOM neuralnetwork these stationsrsquo passenger service level reflected bySTTR values varies greatly and is divided into differentclusters Moreover according to its influencing factorsanalyze the optimization measures that need to be taken Inthe existing method all of the top 20 stations are consideredin the train operation plan so the comprehensive rating isinsufficient and cannot explain why the station service levelis low

Table 4 Train running interval of lines in the morning peak

Line name Running interval(s) Line name Running interval(s) Line name Running interval(s)Line1 120 Line6 257 Line14-East 327Line2 129 Line7 200 Line15 327Line3 129 Line8 212 Line CP 360Line4-main 120 Line9 133 Line FS 360Line4-DX 240 Line13 180 Line YZ 360Line5 180 Line14-west 514 Line BT 180

Table 5 Correlation coefficient matrix

Passenger flow Train running intervals Location coefficient NSTTR LSTTRPassenger flow 1 minus021 minus003 016 023Train running intervals minus021 1 048 018 022Location coefficient minus003 048 1 023 019NSTTR 016 018 023 1 063LSTTR 023 022 019 063 1

Table 6 Principal component matrix of clustering elements processing

Passenger flow Train running interval Location coefficient NSTTR LSTTRPC1 06221 053 04002 02749 03104PC2 01621 02746 0216 062 06835

2

15

1

05

0

ndash05

PC 2

ndash1

ndash15

ndash2ndash3 ndash2 ndash1 0 1 2 3

PC 1

Figure 9 SOM neurons final locations


5

4

3

2

1

0

ndash15 643210ndash1

9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8

Figure 10 -e number of stations covered by SOM neurons

1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4

Figure 11 Stations clustering renderings

Figure 12 Distribution of various cluster stations


6 Conclusion

-is article contributes a method for analyzing the trainoperation plan based on the STTR in the metro system

(1) STTR which is the fluctuation degree between theactual time and standard travel time of each OD fromthis station as the starting station to other stations iscalculated by AFC data

(2) -e clustering algorithm based on SOM neuralnetwork is efficient in classifying stations andidentifying the potential causes of weak STTRlevel SOM is an unsupervised learning neuralnetwork with strong self-organization and visu-alization characteristics

(3) Taking the Beijing metro network as an example theframework is applied and the results are given anddiscussed in detail Besides several suggestions areput forward to optimize the train operation plan

-e application case of the Beijing metro network showsthat the proposed method can be used to analyze the trainoperation plan effectively It is also applicable to other metronetworks withAFC systems A possible future research directionis to expand the methodology framework to the reliability oftransfer time in the time-space dimension More efforts are alsonecessary to adopt diversifiedmeasures for optimizing operationmanagement such as asymmetric operation plans

Data Availability

-e AFC data used to support the findings of this study weresupplied by Beijing Metro Co Ltd under license and socannot be made freely available Requests for access to thesedata should be sent to Mr Wang 15221628818qqcom

Conflicts of Interest

-e authors declare that they have no conflicts of interest

Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ

Figure 13 -e lines connected with Line9

20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name

Top 20 stations with inbound passenger demand

Figure 14 Top 20 stations with inbound passenger demand


Acknowledgments

-is work was supported by the National Key R amp DProgram of China (2018YFB1201402) -e project wasfunded by the Ministry of Science and Technology -eauthors are grateful for this support

References

[1] S Li R Xu and Z Jiang ldquoEvaluation method for matchingdegree between train diagram capacity and passenger demandfor urban rail transitrdquo China Railway Science vol 38 no 3pp 137ndash144 2017

[2] S Li R Xu and K Han ldquoDemand-oriented train servicesoptimization for a congested urban rail line integrating shortturning and heterogeneous headwaysrdquo Transportmetrica ATransport Science vol 15 no 2 pp 1459ndash1486 2019

[3] W Wang L Cheng T Chen and S Ni ldquoA research onadaptability evaluation of train operation plan and passengerflowrdquo Railway Transport and Economy vol 41 no 7pp 65ndash71 2019

[4] F Lu 9eory of Transport Efficiency of Urban Rail TransitNetwork Beijing Jiaotong University Beijing China 2016

[5] S Tian Study on Optimization of Train Routing Planning ofUrban Rail Transit Considering Congestion Degree of TransferStation Beijing Jiaotong University Beijing China 2019

[6] Y Liu and D Chen ldquoAn optimization of long and shortrouting train plan of urban rail transitrdquo Urban Rail Transitvol 41 no 02 pp 117ndash122 2019

[7] Y Shafahi and A Khani ldquoA practical model for transferoptimization in a transit network model formulations andsolutionsrdquo Transportation Research Part A Policy andPractice vol 44 no 6 pp 377ndash389 2010

[8] L Bos D Ettema and E Molin ldquoModeling effect of traveltime uncertainty and traffic information on use of park-and-ride facilitiesrdquo Transportation Research Record Journal of theTransportation Research Board vol 1898 no 1 pp 37ndash442004

[9] A J Pel N H Bel and M Pieters ldquoIncluding passengersrsquoresponse to crowding in the Dutch national train passengerassignment modelrdquo Transportation Research Part A Policyand Practice vol 66 pp 111ndash126 2014

[10] Y Asakura and M Kashiwadani ldquoRoad network reliabilitycaused by daily fluctuation of traffic flowrdquo in Proceedings ofthe 19th PTRC Summer Annual Meeting Proceeding SeminarG pp 73ndash84 Brighton England 1991

[11] Y Asakura ldquoReliability measure of an origin and destinationpair in a deteriorated road network with variable flowrdquo inProceeding of 4th Meeting of the EURO Working Group inTransportion pp 75ndash77 Mayaguez Puerto Rico April 1996

[12] Y Asakura M Kashiwadani and K I Fujiwara ldquoFunctionalhierarchy of a road network and its relations to time reli-abilityrdquo Proceedings of JSCE vol 583 no 583 pp 51ndash60 2010

[13] W H K Lam and G Xu ldquoA traffic flow simulator for networkreliability assessmentrdquo Journal of Advanced Transportationvol 33 no 2 pp 159ndash182 1999

[14] E Bell and C Chirs Reliability of Transport Networks Re-search Studies Press Ltd Biggleswade England 2000

[15] S Dai C Zhu and Y Chen ldquoResearch on time reliability ofurban public transitrdquo Journal of Wuhan University of Tech-nology (Transportation Science amp Engineering) vol 5pp 869ndash871 2008

[16] T Lomax D Schrank S Turner et al Selecting Travel Re-liability Measure Texas Transportation Institute CollegeStation TX USA 2003

[17] W Zhang P Zhao and X Yao ldquoReliability evaluation ofBeijing metro rate travel timerdquo Shandong Science vol 26no 6 pp 77ndash81 2013

[18] W Li J Zhou F Zhou and R Xu ldquoCalculation of travel timereliability of park-and-ride based on structural reliabilityalgorithmrdquo Journal of Southeast University(Natural ScienceEdition) vol 46 no 1 pp 226ndash230 2016

[19] J Chen Research on Operation Reliability of Urban RailTransit Network Tongji University Shanghai China 2007

[20] P V S Rao P K Sikdar K V K Rao and S L DhingraldquoAnother insight into artificial neural networks throughbehavioural analysis of access mode choicerdquo ComputersEnvironment amp Urban Systems vol 22 no 5 pp 485ndash4961998

[21] T-H Tsai C-K Lee and C-H Wei Neural Network BasedTemporal Feature Models for Short-Term Railway PassengerDemand Forecasting Pergamon Press Inc Oxford England2009

[22] W Zhu W L Fan A M Wahaballa and J Wei ldquoCalibratingtravel time thresholds with cluster analysis and Afc data forpassenger reasonable route generation on an urban rail transitnetworkrdquo Transportation vol 47 2020

[23] M Wachs and T G Kumagai ldquoPhysical accessibility as asocial indicatorrdquo Socio-Economic Planning Sciences vol 7no 5 pp 437ndash456 1973

[24] K Pearson ldquoOn lines and planes of closest fit to systems ofpoints in spacerdquo Philosophical Magazine vol 2 no 6 1901

[25] T Kohonen ldquoSelf-organizing maps in Springer Series inInformation Sciencesrdquo vol 30 Springer-Verlag BerlinGermany 3rd edition 2001

[26] R J Peter ldquoSilhouettes a graphical aid to the interpretationand validation of cluster analysisrdquo Journal of Computationalamp Applied Mathematics vol 20 1999


paid more attention to the travel time reliability (TTR) inpublic transportation and TTR has been one of the mostsignificant factors affecting transportation services level[8 9] In general passengers will always ride the first train toarrive after they reach the platform unless the train is toocrowded When the transportation capacity cannot meet thepassenger demand in some stations and sections there willbe a backlog of passengers waiting in station platforms andthus there will be a direct impact on TTR As for an OD pairin the metro network TTR typically has two definitions (1)the probability that passengers can complete a trip within aspecified time (2) the fluctuation degree of the average traveltime of passengers

As the node for passengers to start and finish the journey inthe train operation plan stations are the core of transportationorganizations in the metro system Based on the characteristicsof metro network structure and operation management wepropose the definition of station travel time reliability (STTR) asthe fluctuation degree between the actual time and standardtravel time of eachOD from this station as the starting station toother stations Based on the support of big data STTR analyzesand evaluates the TTR of inbound passenger flow totally toreflect the passenger service level at different stations and pe-riods Combined with operation experience and travel inves-tigations the factors that affect the fluctuation of STTR valueconsist of the following three aspects

(i) Passenger flow of the station is excessive(ii) Train running interval of the line is large that is the

transportation capacity is insufficient(iii) Station location trains are too crowded when ar-

riving at the station because their capacity has beenused in front of this station

To sum up the train operation plan analysis should not belimited to the ranking of indicators but also should pay at-tention to the analysis of potential causes -is study aims todevelop a data-driven approach to analyze train operationplans based on the STTR of all stations in the network -econtributions of this article are as follows

(1) Based on the AFC data an STTRmeasurement modelis built to calculate the value of passenger TTR fromstation dimension and principal component analysis(PCA) is used to process clustering elements

(2) Combining the Self Organizing Maps (SOM)neural network a station clustering framework isestablished with the STTR values and influencefactors to analyze the train operation plan moreobjectively and comprehensively and explore thespecific reasons for low STTR level

(3) Apply the proposed approach to the Beijing metro asa case study and several results are analyzed thatinspire the optimization of the existing train oper-ation plan

2 Literature Review

Numerous studies in the literature related to the train op-eration plan analysis consist of the following aspects

operation organization [1ndash4] and station service level [5ndash7]Li et al [1 2] constructed an interaction model of trains andpassengers and obtained evaluation indexes such as full-loadrate number of passengers and average waiting time andoptimized the train operation plan based on the matchingdegree of capacity supply and passenger demandWang et al[3] evaluated the adaptability of train operation schemes andpassenger demand from three aspects total adaptabilitystructural adaptability and quality adaptability Lu [4] di-vided transportation efficiency into three levels capacityoutput efficiency capacity utilization efficiency and trans-port demand satisfaction efficiency Tian [5] used the pas-senger flow aggregation and congestion as an indicator tomeasure the service level of the station and as one of thebases for the preparation of the train operation plan Liu andChen [6] used the minimization of factors such as thewaiting time of passengers at the station as the objectivefunction to establish a multiobjective nonlinear mixed-in-teger optimization model evaluates and optimizes the lineoperation plan Shafahi and Khani [7] considered theminimum transfer waiting time as the goal and combinedheuristic algorithms to optimize the transfer of the roadnetwork We find these analysis methods focused on re-storing the passenger travel process to extract relevant in-dexes such as train full-load rate platform waiting timeHowever in the process of path restoration the parameterssuch as the passenger walking time and the train maximumpassenger capacity will have a few differences and fluctua-tions in the space-time dimension -e pattern of empiricalvalues for these parameters will cause vast subjectivity andrandomness in evaluation results

-e theory of TTR was first proposed on urban roadtraffic and there are several types of research about the TTRanalysis in public transportation Considering travel be-havior analysis in the road network Asakura and Kashi-wadani [10] gave a concept of TTR the probability thatpassengers can complete the trip within the specified timeand measured TTR of an OD pair in a deteriorated roadnetwork [11 12] Lam and Xu [13] calculated TTRrsquos value byestablishing a traffic flow simulator model and access thereliability of metro systems organization management Belland Chirs [14] analyzed travel time change based on sen-sitivity analysis and described TTR by travel time varianceWhile some scholars [15 16] used the buffer time index(BTI) to describe the TTR BTI is the fluctuation degreebetween the actual and planned travel time at a specifiedperiod Besides Lomax et al [16] defined the unit distancetravel time and defined the BTI as the rate between theaverage travel time and the time of passengers having a 95chance of arriving at the destination on time

To our knowledge little attention has been paid to in-troducing TTR to the train operation plan analysis in thestation dimension which is of great significance to the metrooperation management Zhang et al [17] presented a newunit distance TTR evaluation index and method to assess theBeijing metro network Li et al [18] proposed a TTR cal-culation algorithm to analyze the reliability of transfer timequantitatively Chen [19] proposed the definition andevaluation method of metro network operation reliability



3 Data Description



4 Methodology






tupij min t


lowij + U1113872 1113873

(1)




upij ] are retained








standij1113872 1113873

tstandij

(2)





1113936jisinSMij



1113936jisinSiMij

(3)







(iv) NSTTR(v) LSTTR




X x1 x2 x3 x4 x5( 1113857 (4)


pmn Cov xm xn( 1113857 (5)






730Station1

Station2

Station3

Station4

Station5

830 930900800













max a(i) b(i) (6)











Uk uk1 u

k2 u

kN1113960 1113961

T (7)



Wj wj1 wj2 wjs1113960 1113961T (8)




ui minus WN

min ui minus wj

1113882 1113883 j 1 2 M (9)



j 1 2 M jne c(10)




2lowast δ2(t)⎛⎝ ⎞⎠ (11)




10001113888 1113889 (12)











X1 X2 X3

Input layer

Competitive layer

















Station A Station B

15 trainshour



Station A Station B

20 trainshour














Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References





























3 Data Description



4 Methodology






tupij min t


lowij + U1113872 1113873

(1)




upij ] are retained








standij1113872 1113873

tstandij

(2)





1113936jisinSMij



1113936jisinSiMij

(3)







(iv) NSTTR(v) LSTTR




X x1 x2 x3 x4 x5( 1113857 (4)


pmn Cov xm xn( 1113857 (5)






730Station1

Station2

Station3

Station4

Station5

830 930900800













max a(i) b(i) (6)











Uk uk1 u

k2 u

kN1113960 1113961

T (7)



Wj wj1 wj2 wjs1113960 1113961T (8)




ui minus WN

min ui minus wj

1113882 1113883 j 1 2 M (9)



j 1 2 M jne c(10)




2lowast δ2(t)⎛⎝ ⎞⎠ (11)




10001113888 1113889 (12)











X1 X2 X3

Input layer

Competitive layer

















Station A Station B

15 trainshour



Station A Station B

20 trainshour














Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References






























standij1113872 1113873

tstandij

(2)





1113936jisinSMij



1113936jisinSiMij

(3)







(iv) NSTTR(v) LSTTR




X x1 x2 x3 x4 x5( 1113857 (4)


pmn Cov xm xn( 1113857 (5)






730Station1

Station2

Station3

Station4

Station5

830 930900800













max a(i) b(i) (6)











Uk uk1 u

k2 u

kN1113960 1113961

T (7)



Wj wj1 wj2 wjs1113960 1113961T (8)




ui minus WN

min ui minus wj

1113882 1113883 j 1 2 M (9)



j 1 2 M jne c(10)




2lowast δ2(t)⎛⎝ ⎞⎠ (11)




10001113888 1113889 (12)











X1 X2 X3

Input layer

Competitive layer

















Station A Station B

15 trainshour



Station A Station B

20 trainshour














Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References






































max a(i) b(i) (6)











Uk uk1 u

k2 u

kN1113960 1113961

T (7)



Wj wj1 wj2 wjs1113960 1113961T (8)




ui minus WN

min ui minus wj

1113882 1113883 j 1 2 M (9)



j 1 2 M jne c(10)




2lowast δ2(t)⎛⎝ ⎞⎠ (11)




10001113888 1113889 (12)











X1 X2 X3

Input layer

Competitive layer

















Station A Station B

15 trainshour



Station A Station B

20 trainshour














Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References





























Uk uk1 u

k2 u

kN1113960 1113961

T (7)



Wj wj1 wj2 wjs1113960 1113961T (8)




ui minus WN

min ui minus wj

1113882 1113883 j 1 2 M (9)



j 1 2 M jne c(10)




2lowast δ2(t)⎛⎝ ⎞⎠ (11)




10001113888 1113889 (12)











X1 X2 X3

Input layer

Competitive layer

















Station A Station B

15 trainshour



Station A Station B

20 trainshour














Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References










































Station A Station B

15 trainshour



Station A Station B

20 trainshour














Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References






































Line CP

Line 5

Line

13

Line 4

Line XJ

Line 10Line 6

Line S1Line 1


Line FS

Line 4Line 8

Line YZ

Line 10

Line 16


Line 7 Line BT

Line 6

Line 15

NORTH



Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References




























Morning peakperiods

Off-peakperiods

Evening peakperiods


High level

Low level

0200s

400s

600s

gt600s




450

400

350

300

250

200

150

100

PGY

IDW

LJCS

SH

YQBS

QN

CGZX

CGZ

PAL

BHB

NLG

X DS

CYM

DD

QH

JL JTL

SLB

QN

LD

LP HQ CY CF

WZX

YLTZ

BGBY

HX

HJF

DXY LC ZX

Z

YZL

PXF

_GD

DJ

HY YX

XXK YT

Z

LCQ

GYN

M

PKG

Y

ATZX BT

C

AH

Q

AD

LBJ

GLD

J

SSH

NLG

X

QH

DLX

K

LDK

BST

ALP

KGY

ALL

DTL

D GZ

WJX W

J

CGZ

MQ

Y SH GZ

HLK HSY

NFX SM SY FB

GCL

BJYL

YBB

SYQ

LW

KSW

SLG

ZFJS

BWG

MXD

NLS

LFX

M XDTA

MX

TAM

DW

FJ DD

JGM

YAL

GM

DW

LSH

SHD

AH

QB

BGM XY

YMY

BJD

XDM

ZGC

HD

HZ

RMD

XW

GC

GJT

SGD

WY

XZM

XJK

PAL XS

LJH

TXD

XWM

CSK

TRT

BJN

ZM

JBJM

XG

YXQ XG

XHM

GM

DB

GM

DN ZY Q

YLH

CXD

JH

CHCZ

YHZ

SWYY

JDTG

Y

TTYB

TTYN LS

QLS

QN

BYLB

DTL

DH

XXJ

BKH

XXJ

NK

HPX

QH

PLBJ

YHG

BXQ

ZZZL D

SD

SK DD

CWM

CQK

TTD

MPH

YLJ

YSJ

Z

TTY

STTR

(s)

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

50

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

650

600

550

500

450

400

350

300

250

200

STTR

(s)

500

450

400

350

300

250

200

150

100

STTR

(s)

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR

NSTTR

LSTTR
















2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References









































2

15

1

05

0

ndash05

PC 2

ndash1

ndash15


PC 1



5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References




























5

4

3

2

1

0


9

12 8 15 4 10 7

9 8 5 5 4

11

5 14 8 12 12 8

11 9 7 15 19

3 6 10 12 10 15

12 12 7 9 7 8


1

3

2

PC 1

PC 2

2

15

1

05

0

ndash05

ndash1

ndash15


Cluster 1Cluster 2

Cluster 3Cluster 4

4




6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References




























6 Conclusion






Data Availability




Line 9

Line 10Line FS

Line 14 (west)

GGZ

XJ

LLQ

QLZ


20000

Pass

enge

r vol

ume

TTY

HY

TTYB

HLG LS

Q SJZ

HLG

DD

J

LZ SLH

PGY

HD

WLJ

CCJ

CF JST

LJY

WZX

YL

YQL

BBS

XG CSS

18000

16000

14000

12000

10000

8000

6000

4000

2000

0

Station name




Acknowledgments


References




























Acknowledgments


References




























Date post:	08-Sep-2021
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

MetroTrainOperationPlanAnalysisBasedonStationTravel … · 2020. 7. 18. · of metro network...

Documents