基于支持向量数据描述和XGBoost的风电机组异常工况预警研究

Abnormal State Early Warning of Wind Turbine Generator Based on Support Vector Data Description and XGBoost

该文提出一种基于支持向量数据描述(SVDD)和XGBoost模型的风电机组异常工况预警方法。从机组监控与数据采集系统(SCADA)数据中选择与转速和发电功率密切相关的特征变量,利用SVDD算法对建模数据进行预处理,采用XGBoost建立风机正常性能预测模型。以所建预测模型为基础,构建时间滑动窗计算性能评价指标,并根据统计学的区间估计理论合理确定风机性能异常预警指标阈值。采用某风电场1.5MW风电机组SCADA系统记录的若干真实故障案例,开展异常工况预警仿真试验。结果表明:基于SVDD和XGBoost的风机异常工况预警方法,可以有效地清洗数据,及时识别风电机组异常状态,对于提高风电机组运行的安全性具有较好的工程实用意义。

An abnormal state early warning method for wind generating units is proposed based on support vector data description(SVDD) and XGBoost(eXtreme gradient boosting)model. The feature variables closely related to the generator speed and output power are selected. Then SVDD algorithm is employed to preprocess the SCADA historical data and the XGBoost-based normal performance prediction model is set up. The time-sliding window model is constructed to calculate the performance evaluation index on the basis of the developed model, and the threshold value of which is determined in accordance with the interval estimation theory of statistics. The abnormal state warning tests are carried out using several true historical fault cases recorded in the SCADA system of a 1.5MW wind power unit. It is shown that the abnormal state warning method based on SVDD and XGBoost can clean the original data effectively, and identify the wind turbine abnormal state timely. The proposed method has practical engineering significance for improving the operation safety of wind turbine generator system.