In this paper, the method of system identification and predictive control is used to control the stress of steam turbine rotor. Firstly, the three-dimensional model of turbine rotor is established by using finite element simulation software, and then the results are identified by using ANSYS finite element analysis software. The transfer function is obtained. Firstly, the temperature model is identified with an accuracy of 99.73%, and then the stress model is identified with an accuracy of 98.2% $. Then the transfer function is discretized, and the data is input into the discrete transfer function to verify the accuracy of system identification. After the accurate verification of the transfer function, the transfer function is transformed into the dynamic matrix control model, and then the stress feedback controller is designed to realize the model predictive control (MPC) of the rotor stress. At the same time, the output response of the system before and after the addition of the controller is compared. Finally, the rationality of the control model is verified. Future work will concentrate on more complex working conditions.