基于运行规律和TICC算法的风电SCADA高维时序数据聚类方法

Clustering Method of High-dimensional Time Series SCADA Data from Wind Turbines Based on Operational Laws and TICC Algorithm

针对大型风力发电机组高维SCADA时序数据的工况识别问题,结合风电机组运行规律和TICC算法,提出一种自动分割聚类方法。从高维的SCADA数据中选取风速、转速和桨距角等少量特定参数作为初始分割聚类对象,分析特定参数的运行规律,确定风电机组理论的运行工况。选取一段特定参数的历史数据,利用TICC算法进行离线聚类分割,获得聚类的最优特征参数。将最优特征参数作为TICC算法的输入,对新的特定参数时间序列数据进行分类。最后根据特定参数时间序列的聚类结果,对未进行分割的SCADA时序数据进行聚类处理。选取某2.5 MW双馈风电机组的SCADA时间序列数据对方法进行验证,同时将所提出的方法与FCM算法、GMM算法、K-Means算法进行对比研究。实例验证和对比研究表明,所提的聚类方法充分融合理论知识和TICC算法的优点,可高效处理高维SCADA聚类分割问题,同时保证聚类结果与理论分析结果一致性。

Aiming at the working condition identification of high-dimensional SCADA time series data from wind turbines, a new time series clustering method is proposed by combining the operational laws of wind turbines and the TICC algorithm. A small set of key parameters such as wind velocity, rotational speed and pitch angle are selected from the high-dimensional SCADA data as the cluster datasets. The repeated patterns in temporal data of the key parameters and the theoretical operational conditions of wind turbines are analyzed. Some historical data of key parameters are used to obtain the optimal feature parameters by the TICC algorithm.The optimal feature parameters are then taken as the input of the TICC algorithm which clusters new time series data of the key parameters. Finally, the SCADA time series data of non-key parameters are segmented and clustered based on the clustering results of the key parameters. The SCADA data of 2.5 MW doubly-fed induction generator wind turbine are used to validate the present method.Performance comparison of TICC algorithm, FCM algorithm, GMM algorithm and K-Means algorithm with respect to SCADA data cluster are analyzed. The validation example and comparative study show that the proposed method takes full advantage of the theoretical knowledge and the TICC algorithm. It can efficiently segment and cluster high-dimensional time series SCADA data while the clustering results are consistent with the theoretical results.