For the thermal process is of special features, and it could be disturbed by uncertain factors easily, a novel modeling method based on least mean square (LMS) adaptive filter is presented. Based on the internal model control (IMC) principle, the sub-loop system and the inert area of the superheated steam system are considered as a generalized control object, an internal model-P cascade control system is designed. The models are built by the adaptive LMS filter, the corresponding internal controller is designed, and the selfadaptive control of the full operating condition is realized. The simulation result under typical operating condition in the superheated steam process of a supercritical once-through 600MW boiler show that the method has a good regulating performance. The method is based on the online identification; it is adapted to the complex industrial process whose dynamic characters varied with different operating conditions.