Abstract: In order to accurately identify states of traffic flow, and support real-time traffic flow guidance system, the traffic flow states identification method is put forward based on dynamic Bayesian networks, combining three kinds of traffic flow parameters (speed, volume and occupancy). The method is validated through simulation experiments making use of the date of the city of Southampton, UK. The results show that the traffic flow identification method based on dynamic Bayesian networks can determine more accurately states of traffic flow. It provides a theoretical support for intelligent transportation systems, particularly real-time traffic flow guidance system.