A new concept of capacity flexibility of the subway station system to deal with demand changes is proposed in this paper and the computation framework of which is also established. Two approaches of the framework to calculate capacity flexibility are developed using queuing network capacity models. One method is to preserve the existing demand pattern to estimate the station capacity, and it shows that the flexibility will just vary in relation to the variation of demand volume. The other allows the demand pattern and demand volume to vary to obtain the entire changes of demand. In this method, two models are applied to consider two different types of capacity flexibility: (i) total capacity flexibility and (ii) individual capacity flexibility. Besides, a simulation-based solution to these models is proposed. It combines genetic algorithm (GA) with data envelopment analysis (DEA) to evaluate the simulation results and to guide the search direction. Case study is provided to demonstrate those different concepts of capacity flexibility for the subway station system under demand changes.