【摘要】 为了解决原始的风电机组数据采集与监控系统(SCADA)中包含大量异常记录的数据、难以准确反映机组运行状态的问题，提出了一种带噪声基于密度的空间聚类(DBSCAN)模型的风电机组SCADA异常数据识别方法。该方法从分析风速-功率曲线的特点出发，采用预测误差和分类准确度来选取关键聚类参数邻域半径和邻域最小样本点数，避免了人工确定聚类参数的主观性，且参数选择过程可以完全自动化，实现了风电机组SCADA异常数据的有效识别。通过某风场中风电机组的监测数据进行实例验证，结果表明：所提方法能够在保证异常数据被剔除的前提下，保留尽可能多的正常数据，异常识别效果好于现有的k-dist图法和基于k-平均最近邻算法的改进算法(KANN-DBSCAN)。该研究可为开展风电机组状态分析提供参考。更多还原

【Abstract】 To address the issue that wind turbines’ supervisory control and data acquisition(SCADA) system contains a significant amount of data about abnormal records, which affects the accurate representation of the turbines’ operational status, a method for identifying abnormal data based on density-based spatial clustering of applications with noise(DBSCAN) is proposed. Based on the characteristics of the wind speed-power scatter curve, this method involves the use of prediction errors and classification accuracy to select the key clustering parameters: neighborhood radius and minimum number of sample points in the neighborhood. It avoids the subjectivity of manually determining the clustering parameters, allowing for a fully automated parameter selection process. As a result, it achieves effective identification of abnormal data in a wind turbine’s SCADA system. The proposed method is validated using monitoring data from wind turbines in a specific wind farm. The results demonstrate that the method helps to retain as much normal data as possible while ensuring the removal of abnormal data. It also shows superior anomaly identification performance compared to k-distance graph and KANN-DBSCAN, an improved algorithm based on k-nearest neighbors. This study provides valuable insights for the status analysis of wind turbines.更多还原