Abstract: In order to avoid excessive concentration of passenger flow in subway carriages and guide passengers on the platform to disperse and board trains, this paper aimed to balance the passenger flow in each carriage of the subway, used the standard deviation of balance degree as an indicator, comprehensively considered the existing number of passengers in the carriage, the number of waiting passengers, and the expected number of alighting passengers, to optimize the subway platform passenger flow guidance method. Through AnyLogic simulation software, the paper simulated the platform conditions of the Pearl River New Town Line 5 of Guangzhou Metro, and compared the influence of different guidance modes on the uneven distribution of passengers in the carriage under the conditions of passenger flow ranging from 120 000 people/day to 600 000 people/day. The results show that when the passenger flow of the subway station reaches 240 000 people/day, the standard deviation of the passenger flow balance of the carriages under the platform passenger flow guidance method considering the expected number of passengers getting off significantly improves. Compared with the guidance method that does not consider the number of passengers getting off at the station, the standard deviation of passenger flow balance has increased by a maximum of 25.6%, and compared to the manual guidance method, the improvement has reached 53.6%. The optimized guidance method effectively improves the refined service capacity of the subway.