Fast dynamic density outlier detection algorithm for power quality disturbance data

Abstract

The existing anomaly detection methods have the problems of low accuracy and slow identification speed for outliers detection of high-dimensional power quality disturbance data with time sequence characteristics and large fluctuations. In order to solve the above problems, a fast dynamic density anomaly detection method for power quality disturbance data is proposed. The data set is divided into different time slices according to time, and only the changed data in different time slices are dynamically clustered, so that the data on the next time slice can get accurate clustering results with less time cost. The experimental results show that the proposed method not only ensures the accuracy of anomaly detection results, but also improves the time efficiency.