The Optimization Of Timetable For Urban Rail Transit Based On Hadoop Platform

With the population growth in our country,the traffic jam is getting worse.Urban rail transit has become the main tool for alleviating traffic jams and satisfying public travel needs with its convenience,punctuality and efficiency.The train timetable is an important part of the operation and organization of the urban rail transit system,which clearly defines the train departure time,running time and dwell time.The practical urban rail transit timetable can effectively improve the efficiency of the train transportation,reduce passenger waiting time and improve passenger travel satisfaction.The traffic passenger flow of urban rail transit varies with time and space,and its distribution plays a decisive role in the preparation and optimization of train timetable.Based on the dynamic changes of passenger flows,compiling and optimizing train timetable for urban rail transit can effectively balance the benefits of passenger travel and operation management,and improve passenger satisfaction and train transport efficiency.In this thesis,based on the passenger travel data in automatic fare collection system of Beijing urban rail transit,the Hadoop platform for data processing was applied to analyze the characteristics of passenger flow distribution and predict the future short-term passenger flow.Then,the urban rail transit timetable optimization model was established and an adaptive genetic algorithm was designed based on the dynamic passenger flow.Finally,the model verification and implementation of the Beijing urban rail transit line 13 in the Hadoop platform was carried out.The main research contents were as flows.(1)Based on the actual passenger flow data of automatic fare collection system,a Haoop distributed computing platform was set up to process the data and statistics of the passenger flow.According to the statistical results,the characteristics of time,station,period and direction of line passenger flow were analyzed,which could be used for subsequent passenger flow forecasting and timetable optimization.(2)According to the statistical results of passenger travel data and the distribution characteristics of passenger flow,an extreme learning machine prediction scheme based on time series was studied.According to historical passenger flow,the passenger flow in the next day and any five minutes could be predicted,and the ability of extreme learning machines to predict future short-term passenger flows could be evaluated.(3)Considering the train operation constraints,the multi-objective optimization of timeable aiming at decreasing the waiting time of passenger,the travel time of passenger and the difference of passenger on load and the fixed number as well as improving the train running balance was built,which was based on dynamic passenger flow and takes the train departure time and dwell time as decision variables.Due to the large number of decision variables in the model,the traditional mathematical solution was hard to solve the problem.In this thesis,the genetic algorithm was used to solve the model.In order to improve the global search ability of the algorithm,this thesis designed an adaptive crossover operator and mutation operator according to the characteristics of the model.(4)Based on the passenger flow data of Beijing urban rail transit line 13,this thesis completed the model solution and verification under the Hadoop platform.After inputting passenger flow data,train related data,timetable data and genetic algorithm related data of the line,different optimization timetable with different time periods were obtained through calculation of Hadoop platform.Compared with unadjusted timetable results,passenger waiting time and travel time have been greatly optimized,and the experimental results verified the effectiveness of the proposed model and algorithm.