Abstract A multi-turbine condition monitoring method using supervisory control and data acquisition (SCADA) data for large-scale wind farm is proposed. The method takes the difference between the SCADA data of each turbine with the median of other remaining turbines, and establishes condition vector consisting of the differences. Considering the autocorrelation of turbine SCADA data, vector autoregression (VAR) model is used to remove the autocorrelation in the condition vector of wind farm. Hotelling and multivariate exponentially weighted moving average (MEWMA) control chart are applied to monitor the residual vector. An industrial wind farm example is given to illustrate the proposed method. Compared with the existing turbine condition monitoring charts, the false alarm of proposed method is reduced for considering the autocorrelation of operation data, and monitoring strategy using MEWMA improves detected rate and expedites alarm time compared with Hotelling. The proposed method realizes monitoring multiple turbines simultaneously in farm by a fault indicator, which has important theoretical and engineering significance to the practical operations and maintenance activities in large-scale wind farm.

中文翻译：

---

一种基于VAR模型的农场级多汽轮机状态监测方法

摘要 提出了一种利用监控和数据采集（SCADA）数据的大型风电场多汽轮机状态监测方法。该方法取每台涡轮机的SCADA数据与其他剩余涡轮机的中值之间的差异，并建立由差异组成的条件向量。考虑到风机SCADA数据的自相关性，采用向量自回归（VAR）模型去除风电场条件向量中的自相关性。应用 Hotelling 和多元指数加权移动平均 (MEWMA) 控制图来监控残差向量。给出了一个工业风电场示例来说明所提出的方法。与现有的汽轮机状态监测图相比，考虑到操作数据的自相关性，减少了所提出方法的误报，与Hotelling相比，使用MEWMA的监控策略提高了检测率并加快了报警时间。该方法通过故障指示器实现了对场内多台风机的同时监测，对大型风场实际运维活动具有重要的理论和工程意义。