Abstract: In order to establish the Chinese greenhouse system models with the characteristics of continuous input variables and discrete action climate control equipments, the authors proposed a model greenhouse temperature hybrid system with the control of ventilator. Outside environment factors were used as continuous input variables, while the switch of the ventilator was used as the states of hybrid automata and the inside temperature was used as continuous output variable. So the temperature system hybrid automaton model was established. Each sub model was developed by linear auto-regression moving average with extra input model for air temperature system. Statistic hypothesis test and model mechanism were applied together to select the model structure, and gradually extended least squares method was adopted to identify the model parameters on line. Moreover, an intelligent supervisory segment was devised to monitor and fix problems appearing during the on-line modeling process. Experiments were carried out in a greenhouse to validate the model. It is concluded that the temperature of the hybrid automaton model of fitting degree can reach more than 90% and the maximum error of hybrid control system is 2. 5 ℃. The results show the algorithm is an effective measure to predict the temperature system.