The long-term continuous high-speed rotation of the rotor of the steam turbine generator set inevitably leads to the change of the rotor vibration state. The quality of the rotor vibration state directly affects the safe operation of the whole unit. Therefore, it is imperative to study the early warning method. Firstly, the running law of rotor vibration is analyzed. Secondly, based on the vibration operation law, the calculation method of rotor vibration abnormal threshold is proposed. Thirdly, based on the abnormal threshold and nonlinear state estimation technique（NSET）, the rotor vibration early warning method is proposed. Finally, it is tested and verified by the operation data of a power plant, and compared with the vibration early warning method based on back propagation（BP） neural network. The results show that the proposed method can accurately and effectively realize the rotor vibration early warning.