Thesis
One-class learning on PMU data for outlier detection
Washington State University
Master of Science (MS), Washington State University
2016
Handle:
https://hdl.handle.net/2376/100190
Abstract
The large-scale electric power system has been widely used to distribute power throughout the world. An unhealthy power system can negatively impact not only household electricity but also the operations of business and industry. As a result, the ability to monitor the health of the power system is important. In this work, we use data recorded by phasor measurement units to detect abnormal events. Although several works have focused on outlier detection in a large-scale power system, one-class classification based methods are rarely used. In this work, we use two one-class classification methods, probability density and OCSVMs, to detect outliers in the system. We also combine time series forecasting methods to characterize and preprocess the data to provide more insights than only classification results, such as whether a method can identify the location of an electric fault. Our results show that our methods allow us to identify anomalies in the data while still classifying normal data as normal in most cases. Some methods allow us to identify possible electric faults, and at times, some methods allow us to identify possible PMU device failure
Metrics
9 File views/ downloads
47 Record Views
Details
- Title
- One-class learning on PMU data for outlier detection
- Creators
- Kuei-Ti Lu
- Contributors
- Scott Andrew Wallace (Degree Supervisor)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University; [Pullman, Washington] :
- Identifiers
- 99900524875901842
- Language
- English
- Resource Type
- Thesis