DSpace Repository

Unsupervised Power Line Fault Segmentation and Classification using Periodic Time Series Analysis Techniques

Show simple item record

dc.contributor.advisor Dailey, Matthew N.
dc.contributor.author Singha, Abhinav
dc.contributor.other Anutariya, Chutiporn
dc.contributor.other Taparugssanagorn, Attaphongse
dc.date.accessioned 2021-07-22T07:01:53Z
dc.date.available 2021-07-22T07:01:53Z
dc.date.issued 2021-07
dc.identifier.other AIT
dc.identifier.uri http://www.cs.ait.ac.th/xmlui/handle/123456789/1009
dc.description.abstract Automated anomaly detection has the potential to increase the efficiency of human work in electrical grids. Early detection and accurate classification of faults would help operators avoid escalation of an issue and recover quickly, and would enable efficient and effective analysis of problems across the entire grid. In this research study, I develop methods for detecting anomalies in time series data from digital fault recorders and classifying those anomalies. For this purpose I use a variant of the Periodic Curve Anomaly Detection (PCAD) algorithm, which is an unsupervised learning algorithm for anomaly detection in asynchronous periodic time series data. Taking inspiration from the method used by Rebbapragada et al. (2009), I devised a method that uses phase shift and k-means clustering with a Euclidean distance measure to segregate anomalies in the data. The segregated anomalies are then aggregate and mapped to clusters for classification. The method is able to detect anomalous segments in the time series recorded by digital fault recorders with an accuracy of 98.56%. The classification method however, is not accurate enough for production use in identifying fault signatures. I recommend implementing the PCAD method for fault segmentation followed by a supervised classifier or an unsupervised method with a more sophisticated distance metric for classification.
dc.language.iso en_US en_US
dc.publisher AIT en_US
dc.subject anomaly detection in time series en_US
dc.title Unsupervised Power Line Fault Segmentation and Classification using Periodic Time Series Analysis Techniques en_US
dc.type Research report en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account