【作者】 孙磊；

【导师】 侯公羽；

【作者基本信息】 中国矿业大学(北京) ， 工程管理， 2015， 博士

【摘要】 近年来随着社会经济的发展,越来越多的国内外城市拥有城市轨道交通系统。一方面,由于“准时”、“快捷”等特性,越来越多人在出行时选择轨道交通;另一方面,发生在轨道交通网络上的拥堵现象越来越严重。拥堵现象主要发生在轨道交通系统的站点和换乘通道,而车厢内乘客数量过多称为拥挤。站点拥堵意味着站点的交通需求过大,超过了该站点的交通容量,此处交通需求是指单位时间内想要通过轨道交通系统离开站点的乘客数,而交通容量表示单位时间内该站点能够输运离开的最大乘客数。北京市交通委将对站点拥堵发布绿、黄、红三级预警机制,可见拥堵现象已成为一种常态,越来越引起大家的关注。不论是站点拥堵还是换乘通道拥堵,拥堵现象的发生意味着乘客到达目的地越来越困难,严重影响轨道交通系统的正常运行,甚至会造成严重的踩踏事件,需引起交管部门的高度重视。因此研究发生在轨道交通网络上的动力学过程,对拥堵现象进行深入分析,并提出适合轨道交通网络的优化策略,从而改善系统运行的通畅效率,无疑是摆在交管部门面前的一个非常重要的现实课题。这方面研究对防止拥堵可能导致的灾难性后果,具有重要的理论意义、社会价值和经济价值。具体来说,本文主要研究工作如下:1.基于复杂网络理论研究了轨道交通网络上的拥堵现象。(1)站点拥堵和换乘通道拥堵的原因分析。在轨道交通系统运行过程中,发生拥堵的地方主要有两个,一个是站点,一个是换乘通道。站点出现拥堵,表明站点的交通需求超过站点的交通容量,导致候车乘客滞留站点。从交通需求方面看,站点拥堵的原因有以下两点:一是等待上车离开站点的交通需求过多;一是机车内等待通过该站点的交通需求过多,譬如通过该站点的交通需求过大,车厢过度拥挤,导致候车乘客无法上车。第一种原因独立于轨道交通系统,而第二种原因体现了轨道交通系统交通流量配置的不均衡。换乘通道拥堵,表明换乘流量超过换乘通道的容量,体现为轨道交通系统换乘流量配置的不均衡。(2)两者的联系。不论是站点拥堵还是换乘通道拥堵,都体现为乘客平均出行时间的延长,意味着乘客到达目的地越来越困难,轨道交通系统运行效率的下降,而这属于通畅性的研究范畴;站点拥堵的第二种原因和换乘通道拥堵,都可以理解为发生在轨道交通网络上的动力学过程;此外,站点拥堵的第一个原因,即等待上车的乘客较多,表明进行轨道交通系统的乘客较多,可能会导致换乘通道出现拥堵,而大量的换乘客流涌入换乘站点,进而又会导致换乘站点出现拥堵。站点拥堵,管理部门可以通过限流、增开机车等措施来合理疏导,而换乘通道由于其特殊的地理位置,一旦出现拥堵,往往很难有效疏导。鉴于换乘通道拥堵的危害性,本文将重点围绕换乘通道展开研究。2.提出了添加换乘行为的轨道交通网络模型,并基于该模型评估北京轨道交通网络重要换乘站点。(1)对轨道交通网络的发展状况进行了介绍。分别从线路里程及站点个数、客运量情况进行了数据说明。回顾了轨道交通网络的spacel模型,对该模型的主要拓扑结构指标进行研究,并指出该模型的不足之处在于没能考虑到乘客的换乘行为,而换乘行为在轨道交通系统运行中是普遍存在的。(2)提出添加换乘行为的轨道交通网络模型。①首先,给出模型构造方法。以两条轨道交通线路交汇的换乘站点为例,将该站点一分为二,每条线路分别通过其中一个站点,且两站点之间有换乘通道相连。②其次,对边进行解释,新模型的边分别表示相邻站点连边和换乘通道,而边长的物理意义分别是相邻站点间的平均运行时间和换乘时间,而边长是可以调节的,加开机车频次会相应减小相邻站点连边的长度,而拥堵程度提高时换乘时间会延长,该模型实际是加权路网模型。③再次,指出新模型的优势。可以计算出任意起终点之间的乘客最短出行时间,而这恰恰是理性乘客所追求的;换乘通道的存在使得换乘流量的确定成为可能,而换乘流量是接下去通畅性研究的重要环节。④最后,对新旧模型主要参数进行对比。(3)对北京轨道交通网络的重要换乘站点进行评估。基于复杂网络理论,分别从交通流量、换乘流量和网络效率角度评估了非高峰时段换乘站点的重要程度,其中交通流量和换乘流量分别代表两种不同的客流变动趋势,而通过关闭换乘通道从网络效率损失程度来判断换乘站点的重要程度也具有现实合理性。结果表明,添加换乘行为的轨道交通网络模型融入了轨道交通网络的两大特性,换乘行为和拥堵效应;重要换乘站点评估结果较spacel模型更加真实且符合预期。3.系统分析了基于输运模型的网络动力学过程,从新的视角提出通畅性概念。(1)系统分析网络上的动力学过程。为了加深对动力学过程的理解,对网络上两种常见动力学模型的发生过程进行研究。对相继故障模型的算法实现过程进行分析,并结合具体实例进行仿真试验,分解相继故障发生的过程。对基于输运模型的网络相变发生过程进行重点研究。①首先,关注承载能力的确定。在最短路径路由策略和负荷均匀分布策略下,分别采用基于概率统计的数值计算方法和仿真试验方法来研究网络的承载能力,两种方法的计算结果是相同的。②其次,关注网络相变过程中其它两种负荷数的变化规律,分别是消失负荷数和节点排队负荷数。通过仿真实验分别研究p指标(排队负荷数序列随时间的变化斜率)和x指标(消失负荷数序列随时间的变化斜率)的变化规律。结果表明:在节点处理能力相等策略下,最大介数节点最先产生拥堵,导致网络的进入和消失负荷数出现不平衡,进而导致网络进入拥堵状态;当进入负荷数r小于网络承载能力时,网络上的消失负荷数随r同步增长;当r超过承载能力时,消失负荷数与r的比值持续下降,负荷到达目的地越来越困难。(2)提出基于输运模型的通畅性概念。针对以往通畅性研究的不足,结合交通网络的拥堵效应,从系统运行的通畅效率角度,即负荷到达目的地的难易程度来衡量网络通畅性。既考虑到网络拓扑结构,又考虑到网络上的负荷运动,为探讨网络通畅性提供了新的视角。一旦网络进入拥堵状态,网络的通畅性就会变差,因此网络是否出现拥堵可以作为评价网络通畅性好坏的重要依据之一,而承载能力是网络从稳定状态到拥堵状态的相变点,在一定程度上能代表网络的通畅性能。(3)实证分析了影响网络通畅性的主要因素。通过仿真实验研究了拓扑结构策略(小世界网络、无标度网络和随机网络等)、路由策略(全局信息路由策略和局部信息路由策略)、分布策略(不同负荷起终点分布策略)和处理能力设计策略(相等策略和与节点边介数正相关策略)等对网络上动力学过程的影响。结果表明,核心节点的存在使得无标度网络的通畅性最差,而随机网络的通畅性最好,这对于规划部门进行路网设计具有重要指导意义;基于全局信息路由策略的无标度网络通畅性远优于局部信息路由策略。在全局信息路由策略下,虽然??1时负荷的平均出行时间稍长于??0(即最短路径路由策略),但网络的通畅性大幅提高,这对于交管部门从配置客流角度提高轨道交通的通畅性具有借鉴意义;负荷均匀分布策略下的网络通畅性最好,而当??0时,即负荷起点倾向于在大度数节点时,不论终点倾向于哪种情况,网络承载能力均出现大幅下降态势,这对于我们研究早晚高峰对轨道交通系统的冲击具有重要参考价值。上述策略均对网络上的动力学过程产生重大影响,这也是提出适合轨道交通网络通畅性优化策略的基础工作。4.提出基于轨道交通网络的输运模型,并利用该模型进行轨道交通网络上的动力学过程分析。(1)将输运模型融入到轨道交通网络中,需要对涉及到的关键概念进行细致说明,对输运模型中的一些参数赋予现实意义。以单位时间进入网络的乘客数r为例,表示单位时间内乘轨道交通离开站点的乘客,而不是刷卡进站乘客。(2)提出两种衡量轨道交通网络通畅性的指标,即网络的承载能力和乘客的平均出行时间。对于交管部门来说,站在宏观管理角度,关注的是轨道交通网络何时出现拥堵现象。对于乘客而言,关注的是平均出行时间。(3)提出适合轨道交通网络的通畅性优化策略。将提高网络通畅性的策略运用到轨道交通网络上,需考虑网络的实际情况。提出三种优化策略,分别是路由策略、拓扑结构改变策略和处理能力设计策略。路由策略是交管部门时时发布路况信息,合理引导乘客选择最优出行路径。拓扑结构改变策略是关闭换乘通道和调整机车开行频次。而处理能力设计策略是规划部门在设计规划阶段对不同的换乘站点合理分配处理能力。在路由策略下,轨道交通网络是加权时变网络,乘客会根据即时的网络模型选择最优出行路径,使得进入轨道交通系统的乘客数在高于适宜承载能力时最大介数节点(边)出现拥堵-畅通-拥堵的稳定循环状态,进而提高网络的通畅性。在拓扑结构改变策略下,合理的换乘通道关闭策略使得换乘客流配置的更加合理、均匀,网络的通畅性得到提高。而调整机车开行频次使得管理部门在成本和通畅性之间进行平衡,譬如在人流较少的低峰期,减少机车开行频率,轨道交通网络模型相邻站点连边的长度延长,导致网络的通畅性下降,但开行成本得到降低。在处理能力设计策略下,通过给边介数较大的换乘通道赋予较大的处理能力,网络的通畅性提升明显。(4)实证分析上述策略对北京轨道交通网络通畅性的影响。结果表明,三种策略均能改善网络的通畅性。其中处理能力设计策略对改善通畅性效果最明显,合理的设计策略使得网络的通畅性翻一番;路由策略的效果也很明显,但由于望京西站换乘通道的瓶颈效应,为乘客进出俸伯的必经换乘通道,路由策略没能发挥出应有的作用;而对于换乘通道关闭策略,换乘客流由更少的换乘通道分担,若换乘客流经重新配置后更加均匀,则可以提高网络的通畅性,否则可能会导致通畅性下降。5.实证分析线路故障和早晚高峰对轨道交通网络的影响。以北京轨道交通网络为例,实证分析4号线故障对轨道交通网络的影响。(1)线路故障对轨道交通网络的影响。①首先,分析了线路故障对网络拓扑结构的影响。结果表明,4号线故障后,拓扑结构指标(站点介数和换乘通道边介数)分布更加不均匀,某些线路的交通流量大增,会加剧该线路沿线车站的拥堵程度;网络通畅性大幅下降,从故障前的28下降到22。②其次,故障后的应对策略研究。通过仿真实验对比了路由策略和拓扑结构改变策略的运用效果,结果表明,路由策略对改善通畅性效果显著,网络通畅性从22回升到26;而不同的关闭策略会导致不同的效果,但总体来看关闭策略的效果不明显。(2)实证分析早晚高峰不均匀客流分布对轨道交通网络的影响。在乘客非均匀分布策略下,同一条换乘通道的正向换乘流量和逆向换乘流量是不同的。①首先,提出乘客分布策略的概念。针对早高峰的具体特点,提出三种不同的乘客分布策略。②其次,提出介数的近似计算方法。由于在负荷非均匀分布策略下,介数的计算只能通过仿真实验进行,根据广义介数的定义,通过大量重复实验找出通过节点(边)负荷数的统计规律性,作为计算节点(边)介数的依据。③再次,通过仿真实验对比三种不同的乘客分布策略对网络通畅性的影响。④最后,分析了加开机车频率和关闭换乘通道这两个拓扑结构改变策略对轨道交通网络的影响。结果表明,在乘客均匀分布策略下,介数近似计算值和实际值较吻合;在早高峰时期,乘客不均匀分布对轨道交通网络通畅性影响显著,且分布越不均匀,影响越显著;加开机车频率可以缓解站点拥堵,但可能会加剧换乘通道的拥堵程度。此外,还提出了未来优化策略研究的一些方向和思路。本文的研究思路如下:首先,提出需要解决的问题,即发生在轨道交通网络上的拥堵问题。涉及到两方面内容,轨道交通网络模型和拥堵问题。其次,分析问题。为了更好的模拟轨道交通网络,所建立的路网模型需考虑换乘行为和拥堵效应。而拥堵体现为网络上发生的一种动力学过程,有必要研究网络上的动力学过程模型,从新的视角提出网络通畅性概念,并对影响动力学过程的主要因素进行分析。最后,提出解决方案。建立基于轨道交通网络的输运模型,提出适合轨道交通网络的通畅性优化策略。总体来看,利用基于轨道交通网络的输运模型来研究发生在网络上的动力学过程是合适的;提出的三种通畅性优化策略均符合轨道交通网络的实际,实践中具有可操作性,且实验效果较理想;线路故障和早晚高峰客流冲击均对轨道交通网络影响显著,交管部门需提早做好应急预案。本文的分析及结论可以为交管部门日常的运营管理提供参考价值。更多还原

【Abstract】 In recent years with the development of social economy, more and more domestic and foreign cities have urban rail transit systems. On one hand, because of the characteristics of "time" and "fast", more and more people choose rail transit traffic in travel time; on the other hand, more and more serious congestion phenomenon occurred in rail transit system is. There are two areas in rail transit system which are easily to be congested, one is station, and the other is transfer channel, this state that too many passengers gather in locomotive car, is called cowed, not congested. Station congestion means that traffic demand is beyond traffic capacity of station, which can be understood as that process capacity of station is limited and can not delivery passengers in time. Level 3 early warning mechanisms which are respectively green, yellow and red for station congestion will be released by Beijing municipal committee. Congestion phenomenon has become normal, and more and more people pay attention to it. Whether station congestion or transfer channel congestion, congestion phenomenon means that the destination is more and more difficult to reach for passengers, decreases network connectivity efficiency, seriously affects normal operation of rail transit system, and even causes the worst stampede, so that traffic administrative department need to pay high attention to congestion phenomenon. So research on dynamic process occurred on rail transit system, in-depth analysis of congestion phenomenon, and strategy of optimization for rail transit system which are used to improve communication efficiency of rail transit network, is undoubtedly a very important realistic subject placed in front of traffic administrative department. These study which can help to prevent catastrophic consequences of congestion, has important social value and economic value. Specifically, main research work is as follows:1. Based on complex network theory congestion phenomenon occurred in rail transit system is studied.(1)Reasons of congestion about stations and transfer channels are analyzed. There are two areas in rail transit system which is easily to be congested, one is station, and the other is transfer channel,. Station congestion means that traffic demand is beyond traffic capacity of station, which results into phenomenon that passengers still stay at station. There are two reasons for station congestion: one is that number of passengers who enter into station is so much that stations can not maintain, the other is that station capacity to transport passengers is very inadequate. For example, traffic flow bypass station is too large which means that locomotive car is overcrowded, that waiting passengers can’t get on the car. The first reason is independent of rail transit system, and the second reason embodies unbalanced traffic flow regulated by rail transit system. Transfer channel congestion means that transfer flow is beyond capacity of transfer channel, and embodies unbalanced transfer flow regulated by rail transit system.(2)There is much relation between these two congestions. Both station congestion and transfer channel congestion all embody extension of average travel time for passengers, mean that it is more and more difficult for passengers to arrive at their destination, and operation efficiency of rail transit system declines, which belongs to the category of network connectivity; Station congestion from aspect of second reason and transfer channel congestion, can be understood as dynamics process happened on rail transit network; In addition, station congestion from aspect of first reason, that is to say that number of passengers which enter into rail transit network is so much, may result into transfer channel congestion, and these transfer passengers influx into transfer station, in turn can lead to transfer station congestion. Station congestion can be regulated by administrative department through reasonable measures such as increasing locomotive frequency and limiting number of passengers who enter into station. For transfer channel, because of its special geographical location, once congestion happens, it is often great difficult to conduct effective guidance. In view of dangers about transfer channel congestion, this paper will pay more attention to transfer channel.2. New model based on transfer behavior is proposed which is used to evaluate importance of transfer stations in Beijing rail transit system.(1)Development of Beijing rail transit system is introduced, including mile of all lines, number of stations, and passenger traffic situation. Space L model of rail transit network is reviewed, and deficiency of this model is pointed out, and deficiency lies in not considering transfer behavior, which is widespread in rail transit network.(2)New network model based on transfer behavior is proposed. ①Firstly, introduce method which is used to construct new model. According to the number of subway lines(N) which pass through transfer station, this method produces other N-1 transfer stations, there are transfer-channels between these transfer stations and each subway line passes through one transfer station. ②Secondly, the edge of new model is explained, which respectively means edge between neighbor stations and transfer channel, and physical meanings of length about these two kinds of edge are average running time between adjacent stations and transfer time, length of these two kinds of edge can be adjusted, for example, by increasing locomotive frequency length of edge between neighbor stations will be reduced accordingly, and when congestion occurs length of transfer channel will be extended. This model is actually a weighted network model. ③Thirdly, advantage of new model is pointed out. Shortest travel time path between any two stations can be calculated, which is precisely pursuit of rational passengers; Existence of transfer channel makes it possible to determine transfer flow, which is an important part in next studies about network connectivity. ④Finally basic parameters of Space L model and new model are compared through simulation.(3)Transfer behavior is of great importance in maintaining normal operation of rail transit network, it is of great significance to distinguish important transfer stations in network. Traffic flow and transfer flow at transfer station are used to assess the importance of every transfer station; a new method based on system analysis is proposed, and is applied to assess importance of transfer station in Beijing rail transit network. Results show that: in estimating traffic flow and transfer flow at transfer station, new model is more true and accurate; traffic flow and transfer flow at transfer station can better reflect importance of each transfer station, which is consistent with results obtained in method based on system analysis.3. Dynamics process on network is studied systematically based on traffic routing model which put forward new perspective of network connectivity.(1)In order to understand dynamic process on network deeply, process of two common used dynamic models were studied. This paper is concentrated on algorithm implementation process about two commonly used cascading failure models(load-capacity model and CML model). Simulations are performed in the software combined with actual cases, process of cascading failures is decomposed, and the running results are fully in line with expectations. More and more attentions are paid to research on phase transition based on traffic routing model. ①Firstly, determination of network capacity is paid more attention to. Under shortest path routing strategy and load uniform distribution strategy, two methods are used to study network capacity which respectively based on probability statistical and simulation, and results show that: Numerical calculation result and experimental result about the capacity of three different network are basically consistent; ②Secondly, in process of phase transition variations about three kinds of loads on network are focused on, which are total loads on network, loads that disappear from network and loads that are waiting for passing through some node. Through simulation, law of parameter P which means slope of time sequence for number of loads waiting to pass by node and parameter X which means slope of time sequence for number of loads disappearing from network are conducted. Results show that the node with maximum betweenness is easily to be congested, which results into unbalance between loads that enter into network and loads that disappear from network, and eventually results into network congestion; When R<Rc, the number of loads that disappear from network increases synchronously with R. When R>Rc, ratio of the number of loads that disappear from network and R decreases gradually, which means that it is more and more difficult for loads to reach their destination.(2)Concept of network connectivity based on traffic routing model is proposed. Aiming at shortcomings of previous studies about network connectivity, new concept is combined with congestion effect of transportation network, from connectivity efficiency perspective of system operation, namely that network connectivity is measured by degree of difficulty how loads arrive at destination. Considering both network topology and load transportation on network, new concept offers new perspective to explore network connectivity. Once there is congestion phenomenon on network, connectivity becomes poor, so that whether there is congestion occurred on network can be one of the most important basis to evaluate network connectivity, and network capacity is phase transition point of network from stable state to congestion state, to a certain extent, can represent network connectivity.(3)Main factors which can influence network connectivity are analyzed empirically. Through simulation influence of three topology strategies on network connectivity are studied which are respectively small-world network, scale-free network and random network; influence of two routing strategies on network connectivity are studied which are global routing strategy and local routing strategy; influence of different load distribution strategies on network connectivity are studied; two design strategies of processing capacity(equal strategy or strategy of capacity positively related to edge betweenness) are studied. All these strategies can influence dynamic process on network. Results show that: with the existence of core node, scale-free network has minimum capacity of network, and random network has the biggest network capacity. That is of great guiding significance for planning department to design road network; scale-free network connectivity under global information routing strategy is far better than local routing strategy. Under global information routing strategy, although load average travel time with β=1 is slightly longer than β = 0(shortest path routing strategy), network connectivity is greatly increased, which is of great significance to promote rail transit network connectivity from view point of routing strategy. Simulations show that the maximal capacity corresponds to for LRS, while the maximal capacity corresponds to for GRS; Under even load distribution strategy network connectivity is best, and when, namely that starting point of load tends to be in large degree node, regardless of destination point tend to be any case, network capacity all appears to drop sharply, which is of great guiding significance for us to conduct research about impact of morning and evening rush on rail transit system. The above strategies are to have a significant impact on dynamic process on network, and it is also foundation work which is basis to propose suitable connectivity optimization strategy of rail transit network.4. Traffic routing model based on rail transit network is proposed, and was used to analyze dynamic process on rail transit network.(1)Firstly, traffic routing model should be involved into rail transit network, which includes detailed instructions of key related concepts, and practical meaning of related parameters. For example in unit time loads which enter into network, mean loads which leave off station rather than loads which enter into station.(2)Secondly, two kinds of measures to evaluate rail transit network connectivity were proposed which are respectively network capacity and average travel time for passengers. Standing on macro management point of view, what traffic administrator pays more attention to is when congestion occurs on rail traffic network. While for passengers, it is average travel time.(3)Thirdly, suitable optimization strategy about rail transit network connectivity is proposed. When optimization strategy of network connectivity is applied to rail transit network, features of actual network should be considered. Three optimization strategies were proposed, namely routing strategy, topology-change strategy and capacity design strategy, for example routing strategy is that traffic administrator releases traffic information all the time, which can be a reasonable guide for passengers to select their optimal travel path. Topology-change strategy is to close transfer channel and adjust frequency of locomotive operation. Capacity design strategy is that planning department reasonably allocates different process capacity to different transfer channel in the planning stage. Under routing strategy, rail transit network is weighted network, and passengers will select their optimal path according to newly network model, which can improve network connectivity. Under topology-change strategy, reasonable shut down of transfer channel can result into uniform passenger flow, which can improve network connectivity. And adjusting frequency of locomotive operation can make administrator for a balance between cost and connectivity, for example at slack time, reducing frequency of locomotive operation, which will extend length of neighbor stations on newt network model, results into decline of network connectivity, but cost will have been reduced. Under capacity design strategy, transfer channels with larger edge betweenness are given larger capacity, and network connectivity is increased obviously.(4)Finally, empirical analysis shows that all above strategies can improve rail transit network connectivity. It is capacity design strategy which has the most obvious effect on network connectivity, and reasonable design strategy can double network capacity compared with equal strategy; routing strategy is followed by capacity design strategy, which failed to play a proper role due to bottleneck effect of transfer channel(60, 246), that is to say, for passengers in and out of FengBo this transfer channel is necessary path; For closure strategy of transfer channel, transfer flow are shared by fewer transfer channels, and if transfer flow is reconfigured more even, then it can improve network connectivity, otherwise likely to cause a decline of network connectivity.5. Empirical analysis about impact on rail transit network due to line fault and morning or evening rush is conducted.(1)Beijing rail transit network is used as underlying topology, and empirical analysis is conducted about impact on rail transit network due to 4th line fault. ①Firstly, impact on network topology caused by 4th line fault is analyzed. Results show that: after 4th line fault, betweenness distribution of all stations is deteriorated, and phenomenon of traffic surges in some lines happens which will aggravate degree of some station congestion; ②Secondly, due to reassignment of loads, congestion will happen in some transfer channels, average load transporting time will be extended, and connectivity of rail transit network will be degraded form 28 to 22. Secondly, strategies about how to face with failure are analyzed. Through simulation effects about topology-change strategy and routing strategy are compared, and results show that routing strategy is very useful to improve connectivity, which is promoted from 22 to 26; closing different transfer channels can lead to different results, but overall effects are not obvious.(2)Empirical analysis about impact on rail transit network due to uneven distribution of passenger flow is conducted. ①Firstly, concept of load distribution strategy is proposed. In accordance with specific features of morning rush, three different kinds of load distribution strategies are put forwarded. ② Secondly, approximate calculation method of betweenness is proposed. If load distribution is not even, betweenness can only be calculated through simulation experiment, which is based on definition of generalized betweenness, that is to say, through a large number of repeated experiments to find out statistical regularity of number of loads which pass through node(edge), as the basis of calculating node betweenness(edge). ③Thirdly, through simulation impact of three kinds of different load distributions on network connectivity are compared with. Under uneven load distribution strategy, transfer flows which pass through transfer channel have direction, and transfer flows from a-b is different from b-a. ④Finally, impacts on rail transit network by two topology-change strategies are conducted. Results show that based on even load distribution strategy, by comparing betweenness calculation results of two methods, approximate calculation of betweenness fits well with actual value; During morning rush, uneven load distribution has significant effects on connectivity of rail transit network, and more uneven load distribution is, more significant effect is; to increase locomotive operation frequency can relieve station congestion, but will worse degree of transfer channel congestion. In addition, direction and thinking in the future research about optimization strategy is put forward.Research idea of this paper is as follows: firstly, problem of network congestion occurred on rail transit is put forward, which involves two aspects, rail transit network model and congestion problem. Secondly, the problem is analyzed. Rail transit network model to be established should consider two important features, transfer behavior and effect of congestion. While congestion has two meanings, on one hand congestion means that it is more and more difficult for loads to arrive at their destination, which embodies lower connectivity efficiency of network, and belongs to category of connectivity study; on the other hand congestion embodies dynamic process occurred on rail transit network. Dynamic process based on traffic routing model is studied, from new perspective new concept of connectivity is put forward, and main factors influencing dynamic process is analyzed. Finally, solutions are put forward. Suitable optimization strategy for rail transit network connectivity is put forward.Overall, it is appropriate to study dynamic process on network through traffic routing model based on rail transit network; Three kinds of connectivity optimization strategies proposed in paper are consistent with characteristics of rail transit network, are operable in practice, and experimental effect is ideal; Line fault and morning or evening rush have a great impact on rail transit network, and traffic administrator should make contingency plans early. Analysis and conclusions of this paper can provide reference for traffic administrator at daily operation management.更多还原