基于时空联合聚类方法的输变电设备状态异常检测

鉴于传统的输变电设备状态异常检测方法较少考虑到状态数据的空间信息，提出一种基于时空联合聚类方法的设备状态异常检测方法，该方法综合利用大量设备状态、气象环境等历史数据，在实现异常检测的同时将结果形象化。其具体做法为：通过移动时窗将状态数据时间序列划分为多个子序列，并将子序列与空间位置坐标相结合以构成时空联合数据；使用c均值模糊聚类方法对每个时窗中的时空联合数据进行聚类分析，并基于历史状态数据对每一类赋予异常度值，根据异常度值的大小判断该类数据是否异常；通过在每个时窗的类之间建立模糊关系实现异常状态沿连续时间段传播过程的形象化。最后结合实例验证了提出方法的有效性。

Anomaly Detection of State Information of Power Equipment Based on Spatiotemporal Clustering

In view of the fact that the traditional anomaly detecting methods for power equipment do not consider the spatial information of the state data, this paper proposes a method for anomaly detection of state data of power equipment based on spatiotemporal clustering method, which employs historical big data of the equipment state and meteorological environment and makes visualization of the equipment states in process. The detail of the method is as follows: With a sliding window, the time series are divided into a number of subsequences which will be combined with space coordinates to form spatiotemporal data; the available spatiotemporal structure within each time window is discovered using the FCM method, and an anomaly score is assigned to each cluster, whose value determines whether the cluster is anomalous or not ; then the visualization of a propagation of anomalies occurring in consecutive time intervals is realized by using a fuzzy relation formed between revealed structures. At last, the effectiveness of the method is verified by an example.