Date of Award

2016

Document Type

Campus Access Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering (PhD)

Administrative Home Department

Department of Electrical and Computer Engineering

Advisor 1

Shiyan Hu

Committee Member 1

Xin Li

Committee Member 2

Sumit Paudyal

Committee Member 3

Zhaohui Wang

Abstract

The concept of smart home cyber-physical system has recently gained significant popularity. It controls the residential energy usage to reduce the electricity bills of customers and peak energy load of power grid. The implementation of smart home is enabled by advanced metering infrastructure (AMI), which has become indispensable in a smart grid to support the real time and reliable information exchange between utilities and customers. Yet, due to the vulnerabilities of advanced metering infrastructure and smart meters, the interaction between the two parties is under the threats of cyberattacks with different attacking strategies, among which the prevailing ones are pricing cyberattacks and energy theft. Since smart home schedules the energy usage according to the predictive electricity price published by the utility, pricing cyberattack manipulates the predictive price to mislead customers. This can increase the electricity bill of customers and peak energy load in the power grid. It is possible for hackers to launch cyberattacks at the community level while causing a large area power system blackout through cascading effects. On the other hand, the total electricity bill is computed based on the community wide energy consumption inside a community of networked smart homes. Thus, in energy theft, a customer can hack his/her own smart meter and decrease the metering data. This can save electricity bill of the hacker at the cost of increasing those of others. Pricing cyberattack and energy theft can be coordinated to enhance the effect and increase the difficulty to be detected.

In this dissertation, the vulnerability of the smart home cyber-physical system is assessed. The impact of pricing cyberatttacks and energy theft are analyzed. Subsequently, the detection techniques for these cyberattacks are proposed. For both pricing cyberattacks and energy theft, long term detection techniques are developed using partially observable Markov decision process (POMDP). In addition to detecting the anomaly at a single time slot, the proposed detection technique predicts the potential cyberattacks and accounts the cumulative impact. The simulation results demonstrate that the proposed detection technique can effectively reduce the peak energy load and electricity bills induced by the cyberattacks without imposing high labor cost. Furthermore, a detection framework targeting coordinated cyberattacks is developed based on continuous POMDP. The solving of continuous POMDP is limited by high complexity and difficulties in updating of belief state. To resolve these problems, this dissertation proposes a cross entropy state sampling method to approximate the solution of POMDP. Subsequently, the belief state is approximated using Fourier belief state approximation and updated using particle filtering. It is demonstrated by the simulation results that the proposed detection technique can achieve superior performance in terms of mitigating the impacts of cyberattacks compared to the conventional detection methods.

Available for download on Monday, July 24, 2017

Share

COinS