Train traffic control is characterized by its dynamic, ever changing nature. Simulation is a powerful tool to evaluate the effectiveness and appropriateness of various proposals in train traffic control activities and can be a highly potential tool to support resolving of time and resource conflicts in train traffic control. For previous train traffic simulation models a normal event-driven simulation model and a network-based simulation model have been proposed. The former is a traditional event-based discrete simulation model; the latter is a simulation done by calculating the longest path to all the nodes in a network that is transformed from a normal train traffic diagram. However, a common problem of two models is that they generally lack explicit knowledge representation and are incomprehensible, and this makes it difficult to verify their correctness. For a more efficient and comprehensible simulation, a rule-based train traffic simulation model is proposed. The proposed model attempts to bring together the best ideas of object-oriented simulations and knowledge-based systems to produce a new kind of modeling environment, in which knowledge-based technologies can be more deeply structured within a simulator and the simulation can be entirely embedded in a knowledge-based system. The most important goal of this effort is to improve the comprehensibility of the produced model. In addition, the simulated times of all the trains can be renewed and updated automatically when changes are made to the entities in the simulation. In this way, a reactive simulation can be easily achieved. Validation by experiments shows that the proposed model is workable.