| The construction of urban rail transit in our country has ushered in rapid development,and the scale and complexity of the network have continued to increase.As cities continue to build new lines,they are also using big data and cloud platforms to improve transportation organizations,making my country’s urban rail transit smarter.Passenger flow forecasting is the basis of smart metro and supports the normal operation of various functions,especially in scenarios such as holidays and predictable large-scale events.Therefore,from the perspective of realizing the multi-scenario prediction of rail transit passenger flow,this article has done the following research:(1)Analyze the concept of various sub-functions of urban rail transit passenger flow prediction and the operating mechanism of the overall passenger flow prediction,untangle the data exchange between various sub-functions,highlight the importance of short-term passenger flow prediction,and determine that this article will focus on inbound passenger flow.Secondly,it conducts in-depth analysis based on the inbound data of a certain city’s rail transit after sorting,to explore the characteristics of passenger flow in different scenarios,and provide theoretical support for modeling later.Finally,the commonly used prediction algorithms are introduced,and some problems existing in prediction are analyzed to clarify the direction of modeling below.(2)Based on the analysis of historical passenger flow,an interpretable structural decomposition method is proposed for inbound passenger flow,that is,for each urban rail station’s short-term designated daily passenger flow forecast,the station inbound volume can be regarded as the background passenger flow and the impact of the scene.It also proposes a combined prediction method of multiple linear regression and long-short-term memory neural network,and designs an urban rail passenger flow prediction mechanism based on the MLR-LSTM combined model to complete the prediction of background passenger flow and scene superposition.Finally,compared with the prediction results of multiple linear regression and quadratic exponential smoothing,the quality of prediction is evaluated by indicators such as MAPE and RMSE,which verifies the effectiveness of the model.(3)A short-term passenger flow prediction system under multiple scenarios has been developed for the urban rail transit operation management department.First,the function of the system is designed in detail,and the interface is designed based on the function;second,the system structure is explained,the system operation mechanism and the provision and use of data are clarified through the logical view,and the hardware facilities required by the system are introduced through the deployment view.Finally,the system interface is introduced in detail,mainly with the AFC system,weather,plan operation diagram,scene information configuration and data transfer existing in other passenger flow forecasting subsystems.(4)Carry out case analysis on the basic data of rail transit in a certain city.By predicting randomly designated regular days,Qingming Festival,the food and drinks fair and other scenarios,and analyzing the forecast data at the network level,line level,and station level,the results show that the forecast error of the conventional daily network is within 1%,and the line forecast The error is within 6%,the forecast error of Century City Station is within 5%;the forecast error of the holiday network is within4%,the forecast error of the line is within 8%,and the forecast error of Century City Station is within 2%;During the fair,the forecast error of the network is within 2%,the forecast error of the line is within 15%,and the forecast error of the Xibo City Station is within 1%,which has a certain auxiliary decision-making effect for the operation management department.Finally,based on the forecast results,appropriate passenger flow control and operation adjustment measures are proposed.There are 64 pictures,20 tables and 62 references. |
PDF Full Text Request