Date of Award

Summer 2014

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Cyberspace Engineering

First Advisor

Travis Atkison

Abstract

As the level of sophistication in power system technologies increases, the amount of system state parameters being recorded also increases. This data not only provides an opportunity for monitoring and diagnostics of a power system, but it also creates an environment wherein security can be maintained. Being able to extract relevant information from this pool of data is one of the key challenges still yet to be obtained in the smart grid. The potential exists for the creation of innovative power grid cybersecurity applications, which harness the information gained from advanced analytics. Such analytics can be based on the extraction of key features from statistical measures of reported and contingency power system state parameters. These applications, once perfected, will be able to alert upon potential cyber intrusions providing a framework for the creation of power system intrusion detection schemes derived from the cyber-physical perspective. With the power grid having a growing cyber dependency, these systems are becoming increasingly the target of attacks. The current power grid is undergoing a state of transition where new monitoring and control devices are being constantly added. These newly connected devices, by means of the cyber infrastructure, are capable of executing remote control decisions along with reporting sensor data back to a centralized location.

This dissertation is an examination of advanced data mining and data analytic techniques for the development of a framework for detecting malicious cyber activity in the power grid based solely on reported power system state parameters. Through this examination, results indicate the successful development of a cyber-event detection framework capable of detecting and localizing 92% of the simulated cyber-events. In focusing on specific types of intrusions, this work describes the utilization of machine learning techniques to examine key features of multiple power systems for the detection of said intrusions. System analysis is preformed using the Newton-Raphson method to solve the nonlinear power system partial differential power flow equations for a 5-Bus and 14-Bus power system. This examination offers the theory and simulated implementation examples behind a context specific detection approach for securing the current and next generation's critical infrastructure power grid.

Download

Share

COinS