Waveform abnormality detection method for distribution system equipment condition monitoring

Abstract

In recent years, high quality sensors and meters have enabled equipment condition monitoring. In comparison with power quality monitoring, which has been well-studied in the past three decades, equipment condition monitoring is a relatively new field that focuses on the identification of incipient faults or abnormal operations. In current literature, equipment condition monitoring is carried out through the abnormality detection of waveform-type disturbance data. However, most waveform abnormality detection methods focus on events associated with specific equipment, with only a few methods developed for generic detection. Generic detection is crucial when engineers have no knowledge of the exact equipment that is causing abnormal operations while monitoring a complex system, such as a substation. To this end, this paper summarizes the theories behind abnormality detection and develops three different generic methods in the process, namely the Direct Comparison method, the KLD method and the RSV method. It is proven in this paper that the KLD method and the RSV method are fundamentally inferior in performance compared to the Direct Comparison method due to their assumptions being too ideal. Finally, the Direct Comparison method is proved to be reliable in terms of detection rate and fast enough for online applications.