This paper introduces a condition-based maintenance method combined with long short-term memory network for offshore wind turbine. According to the ranking of offshore wind turbine components using multiple indicators (failure rate, repair time, and maintenance cost), the optimization object focuses on four critical components, namely, rotor, pitch system, gearbox, and generator. Long short-term memory network is implemented to evaluate system condition and predict potential risks, then the preventive maintenance can be performed on the component that reaches the reliability threshold. The repair activity provides an advance maintenance opportunity for the other components, sharing the fix maintenance costs and the downtime. A maintenance decision process is presented in this paper, aiming to achieve the maximum cost savings. Calculated and comparative results demonstrate that the policy proposed in this article is superior in validity and accuracy.

本文介绍了一种基于状态的维护方法与长短期记忆网络相结合的海上风力发电机组。根据使用多个指标（故障率，维修时间和维护成本）的海上风力涡轮机组件排名，优化对象着重于四个关键组件，即转子，变桨系统，变速箱和发电机。实施长短期内存网络以评估系统状况并预测潜在风险，然后可以对达到可靠性阈值的组件执行预防性维护。维修活动为其他组件提供了提前的维修机会，从而分担了维修费用和停机时间。本文提出了维护决策过程，旨在实现最大的成本节省。