An Anomaly Pattern Detection Method Based on Spatial Density of Electric Power Sensor Data
Weijiang Lu, Shuya Lei, Weiwei Liu, Bin Yu · IOP Conference Series Earth and Environmental Science · 2021
Abstract Electric power sensor data is a typical time-series data. Their anomalies are not only abnormal points, but also continuous data fragments named anomaly patterns. Based on the analysis of the abnormal patterns of electric power sensor data, we propose an abnormal pattern detection method to handle the anomaly of spatial density of power sensor data. The proposed method trying to discover abnormal density distribution in the super high-dimensional space. It takes the spatial distribution of the sensor data to build a density model. Moreover, density value of each sensor is computed by matching original sensor data with the density intervals in the model. Furthermore, anomalies are detected according to the density values. In order to verify effectiveness of the method, exclusive experiments are conducted on the real data of power plants. Experimental results show that the proposed method has high recall rate, low false alarm rate and high accuracy while low cost of time when detecting abnormal data patterns