The complexity of taxi operation system grows out of the inherent dynamics and randomness of taxi services. In this paper, we consider a taxi operation system to be a special structure queuing system, and present a simulation model of cruising taxi operation system. It is supposed that vacant taxis cruise in the city to search for passengers, and their search strategies are based on drivers’ experience and available information. With a given OD distribution, the dynamic features of the taxis’ 24-hour available service can be recurred. Through a simulation experiment, it is shown that considering the time-variant effect can help to get more accurate information about taxi services. A new method of determining the system optimal taxi fleet size is developed. These are helpful to taxi supervisor for regulating urban taxi operation system more efficiently and taxi drivers for providing better service.