Abstract: Aiming at the problems of incomplete passenger flow forecasting indicators, coarse temporal and spatial granularity, and insufficient applicability of multiple scenarios in most passenger flow forecasting systems, this paper took the fine passenger flow forecasting demand of urban rail transit with large-scale network operation as the research object, analyzed the implementation method of railway network passenger flow forecasting suitable for multiple scenarios, and used Hadoop, Spark & Hive, Redis, micro service, H5 and other advanced technologies to build a big data platform for passenger flow forecasting, so as to implement the refined passenger flow function of railway network passenger flow OD, provide passenger flow forecasting data support with full indicators such as inbound, outbound, transfer, cross-sectional flow, refined space-time granularity for dispatching command and passenger transport management, and improve the pertinence of rail transit dispatching command, rationality of passenger flow organization and passenger transport service level.