Date of Award

Spring 2013

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)

Department

Electrical and Computer Engineering

First Advisor

Nirwan Ansari

Second Advisor

Ali Abdi

Third Advisor

Edwin Hou

Fourth Advisor

Ali Mili

Fifth Advisor

Roberto Rojas-Cessa

Abstract

Smart grid modernization enables integration of computing, information and communications capabilities into the legacy electric power grid system, especially the low voltage distribution networks where various consumers are located. The evolutionary paradigm has initiated worldwide deployment of an enormous number of smart meters as well as renewable energy sources at end-user levels. The future distribution networks as part of advanced metering infrastructure (AMI) will involve decentralized power control operations under associated smart grid communications networks. This dissertation addresses three potential problems anticipated in the future distribution networks of smart grid: 1) local power congestion due to power surpluses produced by PV solar units in a neighborhood that demands disconnection/reconnection mechanisms to alleviate power overflow, 2) power balance associated with renewable energy utilization as well as data traffic across a multi-layered distribution network that requires decentralized designs to facilitate power control as well as communications, and 3) a breach of data integrity attributed to a typical false data injection attack in a smart metering network that calls for a hybrid intrusion detection system to detect anomalous/malicious activities.

In the first problem, a model for the disconnection process via smart metering communications between smart meters and the utility control center is proposed. By modeling the power surplus congestion issue as a knapsack problem, greedy solutions for solving such problem are proposed. Simulation results and analysis show that computation time and data traffic under a disconnection stage in the network can be reduced.

In the second problem, autonomous distribution networks are designed that take scalability into account by dividing the legacy distribution network into a set of subnetworks. A power-control method is proposed to tackle the power flow and power balance issues. Meanwhile, an overlay multi-tier communications infrastructure for the underlying power network is proposed to analyze the traffic of data information and control messages required for the associated power flow operations. Simulation results and analysis show that utilization of renewable energy production can be improved, and at the same time data traffic reduction under decentralized operations can be achieved as compared to legacy centralized management.

In the third problem, an attack model is proposed that aims to minimize the number of compromised meters subject to the equality of an aggregated power load in order to bypass detection under the conventionally radial tree-like distribution network. A hybrid anomaly detection framework is developed, which incorporates the proposed grid sensor placement algorithm with the observability attribute. Simulation results and analysis show that the network observability as well as detection accuracy can be improved by utilizing grid-placed sensors.

Conclusively, a number of future works have also been identified to furthering the associated problems and proposed solutions.

Share

COinS