Abstract: Objective: Model for station passage walking time real-time estimation is established to improve urban rail transit station operation efficiency and safety, to enhance the passenger travel experience.Method: By adopting a data collection method based on Wi-Fi probe, information data such as passenger device MAC (medium access control) address, signal strength, distance from sniffer are collected by Wi-Fi probes installed in station passage. The collected data are initially processed by java language, and further cleaned in-depth by using a combination of Mysql and navicat premium database, which verifies the feasibility of using BPR (bureau of public roads) function to establish a station passage walking time real-time estimation model, and the ant colony clustering algorithm is used to classify the station passage passenger flow early-warning level based on delay time.Result & Conclusion: The Wi-Fi probe passenger flow collection principle and original data cleaning method are summarized, station passage walking time real-time estimation model based on BPR function is established, and real-time estimation of passenger walking time through station passage is realized, the model for which reaches an accuracy rate of 92.8%. The passenger flow early-warning is classified into four levels: smooth, generally smooth, crowded and severely crowded. The model applicability and prediction accuracy is proved by the practical case verification and analysis of Shanghai Rail Transit Line 11 Jiangsu Road Station Wi-Fi probe data.