CAREER: Time-Synchronized Estimation in Power Systems: Unique Challenges and Innovative Solutions

Arizona State University, Scottsdale AZ

Investigators

Abstract

Over the next two decades, the need for high-speed, high-precision monitoring, protection, and control of the electric power infrastructure will increase considerably as more renewable energy resources are added, electric vehicles become abundant, and frequency and intensity of extreme weather events rise. Time-synchronized measurements can satisfy this need and ensure resilience of this critical infrastructure only if the fundamental concerns regarding limited sensor coverage, lack of data and model interpretability, and heavy online computational burden are successfully addressed. This NSF CAREER project aims to alleviate these concerns by combining recent advances in robust statistics and machine learning to enable accurate and fast time-synchronized estimation in power systems. The project will bring transformative change by bridging the gap between physics-based and data-driven modeling and ensuring creation of algorithms that adapt to the needs of the data and not vice-versa. The intellectual merits of the project lie in creating new mathematical techniques in convex programming, interval-theoretic learning, and distributed optimization. The broader impacts of the project include engaging high school students in intellectually stimulating yet fun problem-solving projects that will expose them to STEM concepts as well as creating a power system workforce that is knowledgeable about data-driven methods in science and engineering. The goal of this project is to explore a new class of optimization problems that are fundamental to time-synchronized parameter, tracking, and dynamic estimation, respectively, in power systems. The proposed research will make new discoveries in two areas: (1) Linear estimation: by creating robust techniques that account for unknown noise characteristics and/or bounded perturbations in both dependent and independent variables. (2) Severely ill-structured estimation: by producing fast, valid, physics-compliant solutions using Bayesian inference and machine learning for problems in which classical methods fail to provide a consistent answer. The methodological and theoretical outcomes of this project achieved through the cross-pollination of ideas from power systems, data science, information theory, and statistics, will significantly boost the use of time-synchronized measurements for operational decision-making. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.