Dartmouth Events

Abstract:  Today, because of an ever-growing demand for clean yet reliable electricity and cyber threats arising from global data connectivity, power grids are facing complex and transformative challenges. Addressing these challenges requires unconventional and interdisciplinary research approaches that exploit tools from various disciplines across electrical engineering, computer science, and operations research. In this talk, I will present a summary of my research on analyzing and enhancing power grids resilience and security using tools from combinatorial optimization and machine learning. I will focus on cyber threats against power grids and present computational and analytical methods to improve grids’ readiness and robustness against such threats. First, I will address the problem of power grid topology and state estimation after a cyber-physical attack that may result in line failures as well as measurement data loss. In general, this problem is combinatorial, and I will show that is NP-hard. However, I will demonstrate that under some practical constraints on the location of the uncompromised measurement devices and the failed lines, the grid topology and state can be estimated in polynomial time by leveraging power flow properties and using tools from algebraic graph theory. I will further show that this approach can effectively be extended to the case when the available measurements are noisy using Bayesian regression. Second, I will provide an overview of a geometric approach to power grid intentional islanding using convex embedding of graphs which is an effective approach to limit the consequences of an initial set of failures in the grid—due to an attack or a natural cause. Finally, I will present a new cyber threat against power grids arising from the ubiquity of Internet of Things (IoT) devices. I will demonstrate that an adversary with access to sufficiently many high-wattage IoT devices (e.g., air conditioners and water heaters) can cause a large-scale blackout in the system by manipulating the demands—without any access to the power grid control system. For example, I will illustrate that only a 1% increase in the demands during the summer demand peak in the Polish grid test case can result in a cascading line failure leading to 86% power outage in the system. I will conclude with possible countermeasures against such attacks and future challenges.

Bio. Saleh Soltan is a postdoctoral research associate in the Department of Electrical Engineering at Princeton University. He obtained the Ph.D. degree in Electrical Engineering from Columbia University in 2017. Previously, he received the M.S. degree in Electrical Engineering from Columbia University in 2012 and the B.S. degrees in Electrical Engineering and Mathematics (double major) from Sharif University of Technology, Iran in 2011. He is the recipient of Columbia University Electrical Engineering Jury Award in 2018, the recipient of Columbia University Electrical Engineering Armstrong Memorial Award in 2012, and the Gold Medalist of Iran 23rd National Mathematics Olympiad in 2005. He is broadly interested in network science, graph algorithms, data analysis, and machine learning with a focus on their applications in power grid resilience and security. His papers have been selected twice among the best papers in power systems modeling and analysis at the IEEE PES-GM conferences and have been covered by prominent media outlets including Financial Times, Wired, Fortune, and CNET among many others.