Research On Passenger Flow Forecasting Of Railway Station

Under the environment of rapid development of railway undertakings,in-depth study on the issue of railway passenger flow has been the basis for improving the allocation of railway resources and enhancing the efficiency and profitability of railway passenger transportation.However,the changes in railway passenger flow are the result of many factors.During non-holidays,the decrease in the number of passenger leads to the low occupancy rate of trips in some routes and the high cost of railway operations,and during the holidays,sudden increase in the number of passenger causes the fact that railway transport capacity can not meet the basic needs of travelers travel issues.This thesis has studied the prediction of railway passenger flow to solve the above problems,aiming to achieve the accurate prediction of railway passenger flow,and it will provide a reasonable and effective vehicle scheduling program to provide a scientific basis.The main work and research results of this thesis are as follows:(1)Getting the timing characteristics of railway passenger flow from the perspective of statistical analysis.By plotting the timing diagram,this thesis analyzes the passenger flow during weekdays and summer holidays and makes research on the passenger flow during holidays which obtains the timing characteristics of passenger flow during weekdays and holidays.The analysises provide the basis for the subsequent formulation of reasonable forecasting method.At the same time,this thesis studies the related factors of railway passenger flow and puts forward the method of grading,and it analyzes the influence of weather and temperature on railway passenger flow.(2)Obtaining the fluctuation feature of railway passenger flow from the angle of modal identification.For the first time,this thesis uses adaptive waveform matching endpoint extension improved EMD algorithm to analyze the characteristics of railway passenger flow.Through the Hilbert spectrum analysis,this thesis calculates the continuous frequency and period of each component.And the time series features of each superimposed component of the railway passenger flow and the trend of passenger flow are obtained.(3)Proposing an improved EMD-SVR passenger flow forecasting model based on adaptive waveform matching endpoint extension.The model decomposes the passenger flow by the improved EMD algorithm and uses the support vector regression model based on genetic algorithm optimization to predict.Based on the correlation and the power contribution rate between each component and the original time series,this thesis builds the EMD-SVR1 model based on total components and the EMD-SVR2 model based on the merged components.The optimal model is selected through experiments to get the accurate forecast of daily passenger flow.(4)Stablishing a holiday passenger flow forecasting model based on passenger flow fluctuation coefficient.Because the passenger flow during holidays have such problems as large changes and lacking of historical data,this model uses the hierarchical clustering method to cluster the fluctuating time series of passenger flow.And according to the result of clustering,the similarity between stations is defined.Then,With the base of passenger flow as an indicator,this thesis proposes a method of defining the volatility coefficient based on the similar degree of passenger flow.Finally,this thesis gets the holiday passenger flow forecasting result using the forecasting value of passenger flow on weekdays and the volatility coefficient.Experimental results show that the model has high prediction accuracy.