基于INNER-DBSCAN和功率曲线模型的风机异常状态检测

Condition Monitoring for Wind Turbines Based on INNER-DBSCAN and Power Curve Pattern

针对目前风电机组异常运行状态无法快速检测问题,提出一种基于INNER-DBSCAN算法和功率曲线模型的数据驱动实时检测方法。该方法先利用贝茨理论和RC模型构造一个新的约束来进行数据预处理,剔除机组极端异常运行数据;再基于提出的区间DBSCAN算法对数据进行聚类,得到正常数据和异常数据;最后利用区间邻域最值对正常数据进行边缘识别,构造风电机组正常运行时的功率曲线模型,并通过模式图的上下临界值识别风机异常运行状态。利用8台风电机组SCADA数据进行实验,结果表明,该方法能有效实时检测风机异常运行状态。

In the light of the problem of rarely an easy and effective way to monitor the abnormal operation status of wind turbine, a data driven abnormal real-time monitoring method for operation status based on INNER-DBSCAN algo- rithm and power curve pattern is put forward in this paper. A new constraint is developed by Betz＇ theory and RC mod- el to conduct data preprocessing so that the extremely abnormal operation data can be ruled out. Then interval DBSCAN algorithm is proposed for data clustering and hence both normal data and abnormal data could be obtained. Also, the maximum value of the interval neighborhood is utilized to recognize the edge normal operation data of wind turbine, and the power curve pattern is built based on it. Finally, the abnormal operation status is monitored by the upper and lower limit value of the pattern. The performance of the presented method is evaluated using SCADA data sets of eight wind turbines, and the results show this method could effectively monitor the abnormal operation status in real time.