Graduation Year


Document Type




Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Electrical Engineering

Major Professor

Mia Naeini, Ph.D.

Committee Member

Yasin Yilmaz, Ph.D.

Committee Member

Nasir Ghani, Ph.D.

Committee Member

Xinming Ou, Ph.D.

Committee Member

Kaiqi Xiong, Ph.D.


Cyber-Physical Stresses, Information Sharing, Kalman Filters, Multi-Region


Wide area situational awareness (WASA) in smart grids includes automatic monitoring, perception and detection of anomalies in these systems. The goal of WASA is to make smart grids aware of their physical and operational state for more effective operational decisions and control. As such, tracking the system's state or state estimation is one of the key objectives of WASA. The extensive integration of cyber elements into smart grids, such as large deployment of various monitoring and measurement devices, provides new opportunities to improve WASA. However, the tight coupling of power grids with cyber components introduces vulnerabilities to cyber and physical stresses.

State estimation is one of the key functions in WASA. The conventional state estimators have been widely deployed in utility control centers to help with monitoring the state of the system. However, traditional model-based state estimation methods do not adequately meet the real-time monitoring and accuracy requirements for smart grids. Many of the model-based state estimation techniques are based on steady-state analysis, which cannot be accurate for modern power systems due to highly dynamic and stochastic variations introduced by, for instance, distributed energy generations and fast-changing loads. The availability of large volume of measurement data in smart grids has opened new directions to complement the traditional state estimation techniques using data-driven state estimation methods. In this dissertation, data-driven state estimation techniques are developed to support the WASA functions, such as monitoring the state of the system and detecting cyber and physical stresses in the system. The presented data-driven and machine learning models include linear Minimum Mean Square Error (MMSE) estimation, Bayesian Multivariate Linear Regression (BMLR) combined with Auto-Regressive AR(p) process, and Kalman filters and Temporal Graph Convolution Neural Networks (T-GCNNs). In addition to the measurement data, the T-GCNN can learn the features in the non-Euclidean domain of the system’s topology, which can capture the structures and interactions among the components of power grids. The performance of the proposed techniques are evaluated using simulated power system measurement data under various normal and stressed scenarios.

Moreover, low latency data processing is important for real-time WASA in smart grids. Distributed and local processing of data is a promising strategy that can improve system monitoring tasks, as it satisfies the low latency requirements while avoiding the enormous overhead of transferring a huge volume of time-sensitive data to central processing units. Distributed data processing may improve the efficiency of many tasks and one such task is state estimation in power systems. In this dissertation, multi-region distributed state estimation is modeled and analyzed under various information sharing techniques among the regions. The regions in the system are defined based on physical distance and the correlation among the state of the components. Several data-driven and machine learning models for centralized and distributed state estimation are evaluated for the system with respect to the various ways of information sharing techniques. It is discussed that the multi-region distributed state estimation can achieve comparable performance to centralized techniques with reduced communication and computation cost.