Study On Short-term Forecasting Method For Boarding Passenger Flows In Urban Rail Transit

As one of the measures to solve the "big city diseases",urban rail transit plays an irreplaceable role in its public transportation system.With prosperity of the economy and the expansion of the city's area,China's urban rail transit construction has been accelerating,and developing by improvements in diversification,networked operation and technological optimization.The short-time passenger flow forecast of urban rail transit not only provides the urban rail transit operation managers with the information of changing of passenger flow,helping to make appropriate adjustments to the train operation plan,it can also provide reference for evacuating and guiding the passenger flow in emergency.Therefore,the short-term boarding passenger flow forecasting is of great significance in the operation and management of urban rail transit systems.Firstly,taking the boarding passenger flow of six stations of different types as examples,the spatial and temporal distribution of urban rail transit passenger flow is carefully analyzed.Considering the multi-level time scale structure of the boarding passenger flow in the time domain,wavelet analysis is adopted to study the regular patterns of urban rail transit.Due to the nonlinearity,non-stationarity,randomness and dynamic nature of urban rail transit short-time passenger flow,this paper proposes a combined short-term boarding passenger flow prediction model with complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and long short-term memory network(LSTM).The CEEMDAN is mainly used to decompose short-term boarding passenger flow of urban rail transit,and the sampled entropy method and hierarchical clustering analysis are applied to merge and recombine the components of decomposed passenger flow sequence,which greatly reduces the operation scale and shortens the calculation time.After these procedures,a long short-term memory network predicting the components after recombination is established,which realizes a high-precision prediction of short-term passenger flows of urban rail transit.Finally,the paper analyzes the case of the Beijing West Railway Station.Compared with classic ARIMA,RNN and LSTM?the CEEMDAN-LSTM model has achieved more than 56%relative declines of MAPE,RMSE and MAE,which can meet the requirements of urban rail transit boarding passenger flow forecasting.