Research On Passenger-to-train Assignment Algorithm And Operation Status Features Analysis Using AFC Data

With the acceleration of urbanization,urban road traffic congestion is becoming more and more serious.Many cities develop subway system to reduce various problems in urban ground transportation.With more and more people choose subway as the main travel tool,the operation status of subway is not only directly related to the travel experience of passengers,but also an important information to operators.Different subway operation features and passengers' subjective travel experience are often used as an important information for operators to guide and improve operation status.At present,the research on metro operation status focuses on the operation index data and the stated preference of passengers.However,the operation index data focuses on the status of train operation system,while the passenger subjective feeling survey results have limited guidance and reference for future operation management.With the development and application of big data technology,through the statistical analysis of existing AFC data,it provides new ideas and technical methods for improving metro operation management,creating better service and operation conditions.In this dissertation,through the analysis of massive metro AFC data,mining the AFC data of metro operation status with pertinence and systematicness,which can provide operators with more precise,more scientific and more economic technical method to improve the operation status.It provides valuable reference for improving the evaluation index system of metro operation status in the future.In this dissertation,we proposed a passenger-to-train algorithm based on AFC data and the extraction method of operation status features.Passengers are divided into nontransfer passengers and transfer passengers,and different passenger-to-train algorithm are established based on their travel process.According to the results of passenger train selection,the study puts forward the extraction method on operation status features.At the same time,special passenger groups(such as the elderly and the disabled)are also concerned in this study.The main contents of this dissertation include:(1)Basic analysis of AFC data.Based on the AFC data of Nanjing metro system,this paper analyzes the travel time,travel location,passenger type,and total volume comparison between transfer and non-transfer passengers.The results reveal the differences on travel pattern,travel time and location of different types of passengers.Its main purpose is to classify,filter and analyze the basic data for passenger-to-train algorithm.(2)A passenger-to-train assignenent algorithm is developed for non-transfer passengers.Firstly,base on the time recorded in AFC data of passenger and the operation scheduling plan,an algorithm for calculating the set of feasible trains satisfied the time constrain is proposed.According to the results of the algorithm,passengers are classified according to the number of feasible trains,and the maximum likelihood method is used to estimate the walking time distribution based on the passengers with single feasible trains.Moreover,the estimation take personal attributes,travel time and other parameters into consideration.Using the estimated walking time,an assignment algorithm is established for passengers with multiple feasible trains.In the algorithm,the travel time difference of different types of passengers in the station is considered to make the assignment result more accurate.Its main purpose is to provide data and technical support for the passenger-to-train assignment algorithm.(3)A passenger-to-train assignenent algorithm is developed for transfer passengers.Firstly,breadth first search algorithm and shortest path constraints are used to determine the shortest path of each travel path.Combined with the AFC data and operation scheduling plan,the algorithm calculating the set of feasible trains based on the travel information.According to the results of the algorithm,passengers are classified according to the number of feasible train in the set,and the maximum likelihood method is used to estimate the travel time distribution of transfer channels in different directions by using the passenger with single set of feasible trains.By integrating all the travel time parameters of non-transfer passengers and the travel time parameters of transfer channels,a data driven passenger-to-train algorithm for multiple sets of feasible trains is established.Through the case study and section passenger flow analysis,the accuracy of the algorithm is further demonstrated.Its main purpose is to provide technical support for the feature extraction method of metro operation status.(4)An approach for extracting operation status features is developed.According to the results of passenger-to-train algorithm,the operation status features of station,travel path and subway network are extracted and calculated from two dimensions of time and flow.At the same time,the influencing factors of metro network travel time delay are explored by using generalized linear model.It is found that the total number of passengers entering the station,train headway and travel period have a significant impact on the total delay of network travel time.The research results can provide operators with more targeted extraction methods and status analysis of station,travel path and subway network operation characteristics,and directly reflect the change trend of operation status at different times and locationsThis dissertation provides a new perspective driven by big data analysis for the existing passenger-to-train assignment theory,and proposes a targeted extraction method for the features of metro operation status by using the passenger-to-train assginment results,which has important theoretical and methodological significance for improving the subway operation level in the future.