Dynamic preventive maintenance model for offshore wind turbine bearings based on remaining useful life prediction

YH Du and XL Geng and QC Zhou and S Cheng and HL Zhang, SUSTAINABLE ENERGY TECHNOLOGIES AND ASSESSMENTS, 84, 104750 (2025).

DOI: 10.1016/j.seta.2025.104750

To improve the availability of offshore wind turbine bearing and reduce maintenance costs, research on bearing remaining useful life (RUL) prediction and preventive maintenance is of significant importance. Although some studies have explored predictive maintenance strategies based on RUL prediction, most have failed to integrate condition monitoring within a multi-objective optimization framework, resulting in inadequate dynamic adaptability of maintenance strategies under complex operational conditions. To address this, this study proposes an innovative dynamic maintenance framework that integrates condition monitoring and RULbased prognostics within a multi-objective optimization framework. First, to accurately predict the bearing RUL, a deep echo state network with information entropy (IEDESN) is designed, which uses information entropy to measure the uncertainty of the system's information, fully exploring the complex models of bearing degradation data. Second, a multi-objective optimization model considering cost and availability is constructed for bearing maintenance in offshore wind turbines, incorporating RUL prediction results that exceed specified thresholds into the optimization process. To further optimize the multi-objective maintenance model, a nonlinear adaptive grey wolf optimization (NAGWO) is proposed. Finally, the results show that the IEDESN model improves the R2 value by 1.98% to 5.45% over comparison methods; the availability of the maintenance model is 0.9964, and the maintenance cost is reduced by approximately 7.88%.

Return to Publications page