考虑运行状态相似性的风电机组数据异常检测方法

Data Anomaly Detection Method for Wind Turbines Considering Operation State Similarity

文中提出一种利用风电场内运行状态相似的风电机组数据采集与监控(SCADA)数据提升异常检测结果准确性的方法。首先,基于实际数据的分析结果,提出了风电机组运行状态的相似性比较原则,进而提出了基于互信息特征选择算法和迭代自组织数据分析聚类算法的运行状态相似性评估方法。然后,在考虑状态变量短时相依性的基础上,利用待检测风电机组的历史SCADA数据构建了基于支持向量机的确定性估计模型,利用相似风电机组的历史SCADA数据构建了基于核密度估计的组合概率估计模型。进一步,利用确定型估计模型和组合概率估计模型分别对目标变量的异常状态进行自检测和外部检测,通过2次检测结果的互相印证来提升异常检测结果的准确性和可靠性。最后,基于一个实际风电场内所有风电机组的SCADA数据和对比实验验证了所提方法的可行性和准确性。

A method for improving the accuracy of anomaly detection results is proposed by using the supervisory control and data acquisition(SCADA)data of wind turbines with similar operation state in wind farms.Firstly,a similarity comparison principle of operation state for wind turbines is proposed based on the analysis results of actual data,and a similarity evaluation method of operation state is proposed based on the mutual information feature selection algorithm and the iterative self-organizing data analysis clustering algorithm.Secondly,considering the temporal dependence of state variables,a deterministic estimation model based on the support vector machine is constructed by the historical SCADA data of the wind turbines to be tested.A combined probability estimation model based on kernel density estimation is also built by the historical SCADA data of the wind turbines with similar operation state.Furthermore,the deterministic estimation model and the combined probability estimation model are used for self-inspection and external-inspection of the abnormal states of the target variables,respectively.The accuracy and reliability of anomaly detection results can be improved by verifying the two detection results.Finally,the feasibility and accuracy of the proposed method are verified based on the SCADA data of all wind turbines in a wind farm and comparative experiments.