数据驱动的风电机组变桨系统状态监测

CONDITION MONITORING OF WIND TURBINE PITCH SYSTEM USING DATA-DRIVEN APPROACH

针对数据采集与监视控制(SCADA)系统存在误报、故障报警滞后等问题,提出一种基于单分类模型的风电机组变桨系统在线状态监测方法。首先,从SCADA数据中提取出与变桨系统相关的特征参数并进行特征重构以进一步提取出更值得关注的桨叶之间的差异化信息。其次,基于单分类支持向量机对历史数据的分析确定变桨系统运行数据的健康边界,进而通过判断实时运行数据是否位于健康边界内部来辨别变桨系统当前的运行状态。最后,以变桨系统的实际工程案例分析验证所提方法的有效性。

To address the problems of a high false alarm rate and delayed fault alarm in supervisory control and data acquisition(SCADA)systems,an online condition monitoring method for wind turbine pitch system is proposed.Firstly,the parameters related to the pitch system are extracted from the SCADA data,and these parameters are reconstructed to extract the differential information between the blades.Secondly,the health boundary of the working status of pitch system is found based on the analysis of the historical data by using one-class support vector machine,and then the online health status of the pitch system can be assessed by judging whether the real-time data is within the health boundary.Finally,the effectiveness of the proposed method is verified by two actual engineering cases of the wind turbine’s pitch system.