07/23/2024
By Shamsun Nahar Edib
The Francis College of Engineering, Department of Electrical & Computer Engineering, invites you to attend a doctoral dissertation defense by Shamsun Nahar Edib on “Cyber-Physical Power System Resilience: Innovations in Design and Restoration Strategies.”
Candidate Name: Shamsun Nahar Edib
Degree: Doctoral
Defense Date: Monday, Aug. 5, 2024
Time: 9:30 to 11:30 a.m.
Location: Ball Hall 302; This defense will also be held via Zoom. Those interested in attending should contact ShamsunNahar_Edib@student.uml.edu or committee advisor Yuzhang_Lin@uml.edu at least 24 hours prior to the defense to request access to the meeting.
Dissertation Title: Cyber-Physical Power System Resilience: Innovations in Design and Restoration Strategies
Committee:
- Advisor: Yuzhang Lin, Ph.D., Assistant Professor, Department of Electrical & Computer Engineering, New York University
- Vinod M. Vokkarane, Ph.D., Professor, Department of Electrical & Computer Engineering, University of Massachusetts Lowell
- Chunxiao (Tricia) Chigan, Ph.D., Professor, Department of Electrical & Computer Engineering, University of Massachusetts Lowell
The evolution of cyber-physical power systems (CPPS) has transformed the conventional electricity delivery infrastructure by integrating advanced information infrastructure for monitoring, communication, control, and computation. Precise measurement, transmission and processing of operational data are indispensable for reliable and efficient CPPS operation. Phasor measurement units (PMUs) play a pivotal role in real-time monitoring and control, which relies on a communication network to transmit data to the phasor data concentrator (PDC) at the control center for further processing and analysis.
Existing literature lacks coordinated PMU and communication link placement techniques, often employing separate and sequential approaches that ignore cross-domain knowledge. This dissertation presents a resilient cross-domain PMU and communication link placement method to minimize the overall installation cost of the wide-area measurement system (WAMS). By breaking down the barrier between the power grid and communication network domains, the proposed WAMS design can withstand any single component failure in the power or communication domain.
Conventional WAMS design frameworks address normal conditions or a small and definite number of failures, whereas high-impact events, such as natural disasters or cyber attacks, may lead to extensive failures. This dissertation proposes an observability-risk-aware WAMS architecture design framework that maximizes statistical observability under multi-component multi-domain failures. Numerical studies demonstrate that the proposed framework achieves higher statistical grid observability under numerous multi-domain failures compared to existing methods.
While a robust WAMS enhances CPPS resilience, the system remains vulnerable to extensive cyber blackouts. Malicious cyber-attacks can compromise PMUs, communication network, and PDCs, jeopardizing grid observability. This dissertation presents an optimal cyber restoration strategy to swiftly recover observability post-disruptions. The restoration problem is formulated considering three condition (PMU measurability, communication network connectivity, and PDC processability) and limited cyber restoration resources. Numerical results show that the proposed method achieves faster observability restoration than heuristic methods, highlighting the need for systematic cyber restoration planning.
This dissertation addresses the challenge of developing a secure yet speedy load restoration strategy admits uncertainties from intermittent renewable energy sources and dynamic loads. By leveraging real-time measurements of restored loads, a Kalman filter-based estimator refines statistical knowledge about unrestored loads. Using this updated knowledge, a rolling-horizon load restoration problem is formulated with specified confidence levels under frequency, voltage, and power flow constraints. Numerical results demonstrate the method's efficacy in achieving faster and more secure load restoration than conventional methods.
During high-impact events, physical and cyber domains may experience simultaneous failures due to their functional and geographical interdependence. This dissertation proposes an integrated CPPS restoration framework that considers the interdependence between these domains. The proposed framework maximizes the energy and information delivery capabilities through coordinated restoration actions, enhancing CPPS resilience.
The CPPS restoration problem is inherently stochastic, with uncertainties in initial outage conditions and restoration action failures. Traditional optimization-based methods are time-consuming and lack adaptability to dynamic conditions. To address these challenges, this dissertation formulates the observability recovery problem as a Markov decision process and uses deep reinforcement learning to solve it. Numerical simulations demonstrate that proposed approach outperforms optimization-based methods in both performance and computational efficiency.