People in major cities are increasingly choosing urban rail transit as their primary mode of public transportation. Excessive passengers on trains and platforms represent a high risk and low comfort for passengers, so it is critical to implement a passenger flow control strategy to improve passenger safety and comfort on trains and platforms. This study proposes a collaborative optimization of train timetable and train formation plan combined with a passenger flow control strategy. To characterize the problem mathematically, a mixed-integer nonlinear programming model is developed to trade-off the total number of train units and the congestion insider stations and carriages with an appropriate passenger flow control strategy. This model is then reformulated into an equivalent mixed-integer linear programming model using linearization methods and solved by an existing solver, i.e., CPLEX. The computational results show that in comparison to the fixed train formation, the multiple train formation can reduce the total number of train units without increasing the number of passengers stranded, and also show that adopting a passenger flow control strategy can prevent the number of passengers on the platform from exceeding the capacity of the platform, thereby improving the metro lines' operation efficiency and safety.