Short-term Prediction Method for Passenger Density in Urban Rail Transit Based on Non-linear Kalman Filter

Abstract

Objective In order to cope with massive passenger flow incidents, it is necessary to identify accurately the spatial and temporal distribution state and the evolution law of massive passenger flow in urban rail transit, making short-term prediction of passenger flow density in urban rail transit based on EKF(extended Kalman filter) and UKF(unscented Kalman filter). Method From both the station and section levels, AFC(automatic fare collection) data processing method is introduced, and the comfort levels of urban rail transit station and section are classified. By defining the state equation and measurement equation of passenger flow density, the short-term prediction and computation methods of passenger density in urban rail transit using EKF and UKF models are introduced respectively. The prediction accuracy of EKF model and UKF model is compared based on a certain line under the networked operation of rail transit in a Chinese city. Result & Conclusion The results show that both EKF model and UKF model can predict the passenger flow density of the station and the section for the next period by collecting the real-time AFC data in the current period, applicable to the short-term prediction scenario of urban rail transit passenger flow density. Compared to the EKF model, the predicted values of UKF model for different time periods in a day are closer to the real trend of changes, and the scatter distribution of the UKF model predicted value and the real value is more convergent. All the RMSE(root mean square error), MAE (mean absolute error)and MAPE (mean absolute percentage error)of UKF model are relatively lower, indicating that the prediction accuracy of UKF model is relatively high.

Keywords

• urban rail transit
• passenger flow density
• short term prediction
• ekf
• ukf