Dynamic Robust Analysis Of Rail Transit And Short-term Passenger Flow Forecasting

With the continuous development of the economy and the acceleration of urbanization,means of transportation have become diversified.Urban rail transit is a kind of important basic transportation facility in the city and the preferred travel mode for residents due to its environmental,fast,and low-cost advantages.However,the contradiction between the supply and demand of rail transit is increasingly prominent.Rail transit congestion and failure become urgent problems that restrict the development of urban smart transportation.Therefore,formulating reasonable and efficient rail transit service scheduling strategies and emergency management strategies is crucial to improve the robustness of the rail transit and ensure the efficient operation of the rail transit.Whether these strategies are efficient and feasible to a large extent depends on the understanding of the characteristics and laws of passenger travel behaviors for rail transit managers.Moreover,the validity of strategies can also be affected by the relationship between passenger travel decisions and rail traffic congestion or failures,and the conditions of real-time passenger travel.Therefore,this paper carries out two main researches from the perspective of passenger travel behaviors based on complex network theory and rail transit data.One is to dynamically analyze the robustness of the rail transit network(RTN)based on the dynamic behaviors of passengers.Based on this,rail transit managers can formulate reasonable emergency management plans in rail transit failures and provide reliable services for passengers.The other is to predict the short-term passenger flows of stations.Therefore,rail transit managers can grasp the travel of passengers in advance,so as to formulate efficient service scheduling plans,optimize the operation mode of the rail transit system,and realize the smart travel of passengers.(1)Aiming at the impact of passenger flow redistribution on the robustness of the RTN when a station fails,an RTN based on the dynamic relationship between the rail structure and the passenger flow distribution is constructed.Then we improve the ability of the cascade failure model in characterizing the temporal and spatial characteristics of passenger flow.Finally,we describe the cascading failure process and analyze the dynamic robustness at different periods.To begin with,to grasp the real-time passenger flows of stations and the dynamic changes of passenger flows,this paper analyzes the temporal-spatial characteristics of passenger travel behaviors.Then,according to two situations of whether passengers are already on the subway train when the rail transit station fails,a two-layer rail transit network(TL-RTN)and a spatiotemporal rail transit network(TS-RTN)are constructed,respectively.Following that we merge the strategy of passenger flow redistribution and the capacity of each station into the coupled map lattice model.So we can simulate the cascading failure processes of TL-RTNs in different periods and analyze the dynamic robustness of networks according to the cascading failures size.By incorporating station in-strength into the influence function of the linear threshold model,we can simulate the cascading failure processes of TS-RTNs at different times.Moreover,we analyze the dynamic robustness of networks according to the new robustness metric.Based on the data of rail transit in Shanghai,experiments show that the RTN robustness is related to both external perturbations and failure modes(i.e.,random and target failure modes).More significantly,the large volume of passenger flow can increase the impact of external perturbations and failure modes on the RTN robustness.Moreover,during the peak hours on weekdays,due to the large passenger flow,a small perturbation will trigger a 20% cascading failure of a network.Having ranked the cascade size caused by the stations,we find that this phenomenon is determined by both the hub nodes and their neighbors.(2)Aiming at the problem that the graph convolutional network(GCN)cannot accurately fit the peak passenger flow,a method based on the maximum difference of aggregate information is proposed.It optimizes the ability of GCN extracting the spatial characteristics of the rail transit network,and realizes the accurate prediction of the short-term passenger flow of stations under multiple time granularity in the future.First,this paper decomposes GCN formula.After normalizing the adjacency matrix of GCN,we increase the weight of nodes on the diagonal of the adjacency matrix to offset the smoothing effect of the smoothing filter in the GCN Fourier domain on the local peaks,and maximize the difference in aggregated information.Then temporal features are extracted by the above-mentioned temporal sequence with spatial features as the input of the gated recurrent unit(GRU).Thus,a combination model of optimized GCN combined with GRU is constructed,that is,a spatiotemporal graph convolutional network based on the maximum difference of aggregated information(E-GCN).Finally,based on the data of rail transit in Shanghai,we use the E-GCN model to conduct vast experimental simulations.By comparing with other mainstream prediction algorithms,we find E-GCN can effectively optimize the aggregated information of nodes and solve the problem that the peak passenger flow cannot be accurately fitted.The final results perform better and achieve a higherprecision forecast of passenger flow.In summary,this paper carries out two researches(i.e.,dynamic robustness analysis of rail transit network and short-term passenger flow forecast of stations).It helps rail transit managers 1)understand passenger travel behaviors;2)comprehend effects of cascading failures caused by stations at different times on network robustness;3)master future short-term passenger travel trends,thus improving the service scheme and emergency management strategy in time.