Abstract: A deterioration trend perception algorithm considering the operating characteristics of wind turbines was proposed. Firstly, a kinetic deterioration state indicator based on damage accumulation theory was proposed, and the bearing deterioration level (DL) was calculated by Hertz theory. Secondly, a hybrid long short-term memory (LSTM)-CatBoost predictor (HLCP) deterioration prediction method was proposed. The deterioration growth value was predicted by considering the influence of wind turbine operating characteristics and introducing the bearing speed as a time series input. Then, the critical value of bearing damage was calculated based on the simulation model, so as to predict the remaining bearing life. Finally, it was validated using the vibration and supervisory control and data acquisition (SCADA) data from a wind farm in the north of China. Results show that the proposed deterioration state indicator has a similarity of about 74.42% in predicting trend and historical trend, with obvious trend and monotonicity. The error between the remaining life prediction results obtained based on the HLCP model and the actual failure time is only 640 h, which is 16 461 h less compared to the prediction method that does not take into account the turbine's operating characteristics.