A self-data-driven method for remaining useful life prediction of wind turbines considering continuously varying speeds

Predictive maintenance is one of the most promising ways to reduce the operation and maintenance (O&M) costs of wind turbines (WTs). Remaining useful life (RUL) prediction is the basis for predictive maintenance decision. Self-data-driven methods predict the RUL of a WT driven by its own condition monitoring data without depending on failure event data. Therefore, they are applicable in industrial cases where no sufficient failure event data is available. One challenging issue for RUL prediction of WTs is that they generally suffer from varying rotating speeds. The speed variation has serious impact on the degradation rates as well as the amplitudes of state observations. To deal with this issue, this paper proposes a self-data-driven RUL prediction method for WTs considering continuously varying speeds. In the method, a generalized cumulative degradation model is constructed to describe the degradation process of WTs under continuously varying speeds. A baseline transformation algorithm is developed to transform health state observations under varying speeds into a baseline speed. A continuous trigging algorithm is employed to determine the first degradation time (FDT) for degradation modeling and the first predicting time (FPT) for RUL prediction. The best fitting model is selected adaptively to keep in line with the degradation trend of interest. The effectiveness of the method is demonstrated using a simulation case study and an industrial case study.

Automated summary

Predictive maintenance is crucial for reducing the operation and maintenance costs of wind turbines, and self-data-driven RUL prediction methods can be applied in industrial cases without sufficient failure event data. A method has been proposed for RUL prediction of WTs considering continuously varying speeds, showing effectiveness through simulation and industrial case studies.