05/09/2023
By Danielle Fretwell

The Francis College of Engineering, Department of Electrical & Computer Engineering, invites you to attend a doctoral dissertation proposal defense by Gang Cheng on “Power System State Estimation under Measurement Uncertainty and Cyber Threat.”

Candidate Name: Gang Cheng
Defense Date: Tuesday, May 16, 2023
Time: 10 to 11:30 a.m.
Location: Ball Hall, Room 302

This defense will also be held via Zoom. Those interested in attending should contact the student (Gang_Cheng@student.uml.edu) or the advisor (Yuzhang_Lin@uml.edu) at least 24 hours prior to the defense to request access to the meeting.

Committee Members:

  • Advisor Yuzhang Lin, Ph.D., Assistant Professor, Electrical & Computer Engineering, University of Massachusetts Lowell
  • Tingshu Hu, Ph.D., Professor, Electrical & Computer Engineering, University of Massachusetts Lowell
  • Tricia Chigan, Ph.D., Professor, Electrical & Computer Engineering, University of Massachusetts Lowell

Abstract:
The performance of power system state estimation (PSSE) is typically vulnerable to both naturally occurring measurement uncertainties and man-made cyber threats. Conventional state estimators are commonly designed either based on heuristic assumptions of measurement error distributions, or to have formulations that are insensitive to a wide variety of possible distributions, yielding suboptimal error filtering performances as they are inherently agnostic to the true distributions of measurement errors. In this proposal, the concept of adaptive state estimation (ASE) is proposed. Specifically, the Gaussian-Laplacian mixture (GLM) model is developed to fit the body and tail of unknown measurement error distributions, and an adaptive estimation framework is proposed based on the expectation-maximization (EM) algorithm. The GLM-based ASE is capable of tracking the actual error statistics online and adjusting the parameters of SE to maintain near- optimality of state estimates under complex measurement error conditions. To further improve the capability of ASE tracking the complicated measurement error distribution, the Gaussian mixture model (GMM) is adopted to fit the unknown, non-Gaussian, and time-varying measurement error statistics. The proposed GMM-based ASE can capture arbitrarily complex measurement error distributions, preserves high computational efficiency, adapts to abrupt gross errors, and also enables a sensor calibration approach for both phasor measurement units (PMUs) and supervisory control and data acquisition (SCADA) systems without the need of field experiments. 

In addition to measurement uncertainties, the vulnerabilities of information and communication technology (ICT) infrastructures leave room for cyber attacks threatening the reliable operations of power systems. As a major type of cyber threats against PSSE, false data injection attacks (FDIAs) can interfere with the decision-making process, leading to risky operating conditions such as biased locational marginal prices (LMPs), inefficient dispatch of generation, or even blackouts of power systems. In this proposal, we investigate both the construction strategy and detection method of FDIAs. A novel attack strategy, namely model-measurement data integrity (MMI) attack, is proposed. Instead of compromising measurements only, we investigate the possibility where network parameters are coordinately manipulated and model cyber adversaries’ possible behavior of co-planning the manipulated measurement channels and parameter attack vectors prior to the launch of FDIAs. To the attacker’s advantage, the revealed MMI attack strategy significantly reduces the efforts required to launch stealth FDIAs. Moreover, existing FDIA detection methods based on the statistical consistency of measurement values have some limitations. These methods may not work effectively when false data do not significantly deviate from historical trends or may mistakenly treat actual power grid events as FDIAs. To address these issues, we developed a highly discriminative FDIA detector named the k-smallest residual similarity (kSRS) test. This method depends on the rationale that perfect FDIAs can hardly be achieved in AC state estimation, and real-world imperfect FDIAs always lead to subtle changes in the probability distributions of measurement residuals. Therefore, the statistical consistency of measurement residuals can be carefully portrayed to detect practical FDIAs in AC state estimation.