Thesis
Deep learning based abnormal event detection using PMU data: Building a better mousetrap
Washington State University
Master of Science (MS), Washington State University
08/2020
DOI:
https://doi.org/10.7273/000004143
Handle:
https://hdl.handle.net/2376/125058
Abstract
The steadily increasing utilization of Phasor measurement units (PMUs) to obtain synchronized and accurate data such as frequency, voltage, and current phasors has opened a window into a wide variety of risks associated with the cyberattacks on power systems. A malicious attack on PMUs, like injecting manipulated data, may lead to unintended activities that could endanger the reliability of the power system. The attackers often disguise their malicious activities by injecting spoofed signal to the system. This paper proposes a deep learning-based technique to detect and identify the anomalous data introduced by the attackers. Our system uses Long Short-Term Memory (LSTM) networks and several other machine learning techniques to distinguish between "normal" data and "spoofed" data. LSTM networks are a type of Recurrent Neural Network and are capable of learning order dependence in sequence prediction problems. Unlike standard feed-forward neural networks, LSTM has feedback connections and a memory component which helps in learning the sequence. In this research, anomalies are modeled by manipulating the normal data using various spoofing strategies. The PMU data streams used in this work are obtained from a large provider's PMU network and an inter-university PMU network. Experimental results presented in this work show that the developed approach is efficient in classifying spoofed data. We also compare the accuracy of our system with previous works that aimed at detecting the outliers in multivariate time-series data. The experimental results show that using LSTM is more efficient in detecting anomalous data with higher performance.
Metrics
9 File views/ downloads
32 Record Views
Details
- Title
- Deep learning based abnormal event detection using PMU data
- Creators
- Priya Narayana Subramanian
- Contributors
- Scott Wallace (Advisor) - Washington State University, Engineering and Computer Science (VANC), School of
- Awarding Institution
- Washington State University
- Academic Unit
- Engineering and Computer Science (VANC), School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Identifiers
- 99900890779801842
- Language
- English
- Resource Type
- Thesis