Research On Short-term Prediction Of Passenger Flow Into An Urban Rail Transit Station

The passenger flow volume of urban rail transit station always has short-term random increasing phenomenon in some operation peak hours. This will have a significant impact on the safety operation of urban rail transit station. Therefore, it is necessary to effectively predict the random increase of the passenger flow into an urban rail transit station to make necessary prevention measures for the safety operation management of the station, which has become one of the important research subjects.The existing researches have made some achievements in some short-term traffic flow prediction studies. However, they rarely considered abnormal random increasing passenger flows. Regarding the parameter of each individual as the random parameter fluctuating around the average value, the random coefficient regression model is able to more rationally describe the influence difference of each explaining variable. In consideration of the attributes of the random passenger flows into an urban rail transit station, this research develops a random coefficient model to predict short-term passenger flows into an urban rail transit station. The random coefficient model established in this paper will be calibrated by Bayesian statistical inference based on Markov Montecarlo, because the probability convergence speed and error term of Markov Montecarlo method are independent of the dimension of a studied problem.Using the proposed random coefficient model, this research analyzes the changes of the station passenger flows into a typical rail transit station under different conditions. Research results show that the density of the passengers on the platform of the urban rail transit station is very easy to exceed or become close to the safe upper limit for the safety operation of the platform under the impact of short-term random increased passenger flow overlying to normal large passenger flow. This would cause a great threat to the safety operation of an urban rail transit station. The random coefficient model developed in this research is able to effectively predict the superposed large passenger flow into an urban rail transit station and accordingly, forewarn the station to take effective measures in time to reduce the harms which will be caused by the large passenger flow into the station. In addition, the passenger flow forecast results under different conditions show that using the forecast results and warning time provided by the proposed random coefficient model, improving the guidance of the passenger flows on station platform, in the station hall and outside the station will effectively reduce the maximum density of the passengers on the platform. As a result, the safety of the passengers on the platform can be ensured and the possibilities of accidents are decreased.