Aiming at the problems of large error and complicated calculation in the existing blade remaining useful life (RUL) prediction methods, an RUL prediction method considering blade fatigue failure mode is proposed. Firstly, the blade damage model is established, and the crack extension process is observed under basic operating conditions. Then, for the nonlinearity and uncertainty of crack expansion, unscented Kalman filter algorithm is used for prediction. The tracking accuracy of the crack expansion trend is improved by fusing the fatigue failure mode with the crack observation data. Finally, simulation tests are conducted using a National Renewable Energy Laboratory (NREL) 5 MW wind turbine. Different average wind speeds and initial crack lengths are set by test. The simulation results show that the crack extension prediction result agrees with the observation data by 98.75%, and the RUL prediction error is greatly reduced. Comparative analysis results show that the proposed method can effectively suppress the interference of nonlinear factors and reduce the calculation time, which can be effectively applied to crack prediction. The research can provide theoretical support and engineering application reference for wind turbine blade health management.