Subway Traffic Calculation Model For Research

Operation scheduling of subway refers to providing trains with stable departureintervals and effective operations to transport and evacuate people in time, so as toimprove the passengers’ feelings of comfort to the utmost extent within the subwaydepartments’ scope of service ability, and to keep the gains of subway departments atthe meantime. The research object of this thesis is short-term subway passenger flowforecast and the operation scheduling optimizing computation on this basis. It aims atprobing into the forecast method of short-term subway passenger flow, breaking throughthe accuracy of the existing research result of short-term passenger flow. Based on this,the author tries to establish a operation scheduling optimizing plan which can cope withthe change of passenger flow in time.Real-time passenger flow is one of the main factors in operation scheduling ofsubway. It requires a high accuracy of real-time passenger flow forecast. Generallyspeaking, rail passenger flow forecast is made on the basis of inhabitants’ tripdistribution or passenger flow distribution of other means of traffic choices. In thisthesis, subway short-term passenger flow forecast is based on historical passenger flowdata. First, the characteristics of subway traffic and common change trend of subwaypassenger flow will be analyzed, then through a comparison of the advantages anddisadvantages of different forecast methods based on the time sequence method,regression method, and BP nerve net method, etc. and with the Support Vector Machinemethod being adopted, a subway short-term passenger flow forecast model will be setup at last. Ant Colony Optimization, which can solve excessively complicated problems,is adopted as the way of fixing the model parameters to search the optimal modelparameter. At last, taking the station passenger flow, section passenger flow and theactual pull-in station passenger flow of No.1subway in Chengdu as an example, theauthor testifies that the forecast of this model is more accurate than other computationmethods such as BP nerve net method, with relative error and mean-square error as theevaluation principle. The operation scheduling optimizing model has combined two aspects, that is, thepassengers’ feelings and the gains of subway companies. Passengers’ feeling can bedivided into time cost and the extent of congestion, in which time cost contains the timeof boarding, waiting, and transferring, and the congestion extent has also been countedas extra time cost. The influence of the gains of subway companies covers companyinvestment and ticket income. The investment part includes the quantity of trains, thedeparture interval, and staffing, etc. Ticket income is decided by the number ofpassengers and ticket price. Considering the passengers’ feelings and the gains ofsubway companies, the author in this thesis has set up a rational operation schedulingoptimizing model. The computation result of this model is to help subway companiesreduce the transferring time and improve boarding comforts of passengers on subway.Meanwhile，to keep a certain balance between the company’s investments and gains isanother goal. As the model is a restricted non-linear optimizing problem with manyvariables, the author adopts the Genetic Algorithm computation model which is veryadaptable, efficient, and of strong robustness. At last, combined with the operationsituation of No.1subway in Chengdu, the author computes the departure time of thetrain after dispatch optimization.