Probabilistic RUL calibration and risk-constrained rolling optimization for wind turbine gearbox maintenance in harsh environments

Abstract

This paper addresses the challenges associated with predicting the remaining useful life (RUL) of wind turbine gearboxes operating in harsh environments, such as deserts and the Gobi region. These challenges include significant time-varying operational conditions, high maintenance costs, and limited accessibility for maintenance. To address these issues, a method is proposed for optimizing maintenance intervals and modeling time-varying costs by incorporating RUL information. A RUL calibration mechanism using the Dual-Path Calibration with Survival Consistency Network (DPCS-Net) is developed to correct prediction bias and quantify uncertainty. Based on calibrated RUL distribution, a failure risk constraint is set to obtain the maximum safe maintenance interval. A time-varying cost model with opportunity maintenance discount is built, and a rolling optimization strategy is used to dynamically determine the lowest-cost maintenance timing. This approach enables adaptive adjustment of maintenance intervals according to the degradation state and operating environment. Simulation results demonstrate that the proposed method effectively reduces long-term operation and maintenance costs while satisfying reliability constraints. Moreover, it exhibits superior robustness and cost-effectiveness compared with fixed-interval or point-estimate RUL-based strategies. This study innovatively integrates RUL calibration with maintenance optimization, introducing the DPCS-Net to ensure prediction accuracy and uncertainty quantification under survival consistency. It constructs a time-varying cost model incorporating environmental constraints and opportunity maintenance discounts, and adopts a rolling optimization strategy to dynamically adjust maintenance intervals, filling gaps in adaptive maintenance for wind turbine gearboxes in harsh environments.