Demand Generation for Simulation-based Section Flow Prediction System in Urban Rail Transit

Due to the inconsistency between the time range of input origin-destination (OD) matrix and the time interval for generating travel demand in the real-time simulation-based forecast system, a combined strategy integrated with the station inflow prediction was proposed. The time span corresponding to the OD matrix was discretized into the equal-length sub-periods, and the short-term station inflow was predicted by using the data from automatic fare collection system. Then the flow ratio in each sub-period was gained which made the demand generation process follow the real passenger arrival distribution. A station of Beijing metro was applied to verify the method. Compared to the normal generation approach, the relative error of the proposed strategy can be reduced by 22.89%. Compared with various short-term prediction models, it demonstrates that the Kalman filter model can meet the requirements of time and accuracy for on-line predictions.