【摘要】 在风电机组数据采集与监视控制（SCADA）系统数据中，若噪声数据密度过高，则会在预处理过程中误清洗额定功率数据。使用基于密度的噪声应用空间聚类（DBSCAN）算法剔除额定功率数据附近的噪声数据点，确保仅保留正常的额定功率数据，然后在“风速-功率”曲线上找到额定功率数据与其他数据的分界线，将上半部分暂存，对下半部分采用肖维勒准则与Box＿Cox变换相结合的方式处理，最后将两部分数据合并，可有效减少风电机组SCADA数据预处理时，因噪声数据密度过高而误清洗额定功率数据的问题。更多还原

【Abstract】 In the SCADA system data of wind turbines, if the density of noise data is too high, it may mistakenly clean the rated power data during the preprocessing process. To address this issue, the DBSCAN clustering algorithm can be used to remove noise data points near the rated power data, ensuring that only normal rated power data is retained. Then, on the wind speed-power curve, identify the boundary between the rated power data and other data, and temporarily store the upper part. For the lower part, apply a combination of Chauvenet′s criterion and Box-Cox transformation to handle it. Finally, merge the two parts of the data. This approach can effectively reduce the problem of mistakenly cleaning rated power data due to high noise data density during the preprocessing of wind turbine SCADA data.更多还原