The Cyber-Physical Challenges of Transient Stability and Security in Power Grids
Abstract:
Motivation:
Power transmission networks underpin our way of life and are at the center of the transformation of the US energy system. “Keeping the lights on” as we transform an already large and complicated power system is a fundamental challenge for cyber- physical engineering. Future power networks will be instrumented with synchrophasor measurement units, communication infrastructures and distributed computing, and are therefore prototypical examples of cyber-physical systems with tightly coupled computational and physical resources. Our team integrates expertise in power networks, fault detection, cyber security, control systems, distributed systems, and network science.
Our focus:
To develop new principles that integrate physical and cyber aspects into a unified theory, we focus on the problem of detecting and avoiding fast instabilities of power networks that can cause blackouts. We propose novel approaches to analyze these dynamic instabilities and to design cyber-physical control methods to mitigate them. The controls must perform robustly in the presence of variability and uncertainty in generation, loads, communications, and the operating state, and during abnormal states caused by natural faults or malicious attacks. A general insight is that we learned the importance of finding simple conditions summarizing intricate physics and engineering that can give actionable information to enable cyberphysical controls.
Highlights:
Transient stability is the ability of the entire power grid to stay synchronized together at 60 Hz. Almost exact conditions for transient stability have been obtained in terms of the spread of natural frequencies of generators (oscillators) and the coupling provided by the power grid, and this has been published in the PNAS journal. Electromechanical grid oscillations occur, for example, when voltages in Arizona slowly swing at about 1 Hz relative to voltages Canada. We have published an analytic for-mula for the damping of these harmful oscillations by redispatching generators based on measurements of the patterns of oscillations and power flows. (Some other researchers thought this impossible.) We are formulating a practical algorithm based on synchropha- sor measurements to suppress the oscillations. Due to an insight into obtaining consistent data sets describing both the power grid and line outages, we can now track real outages on a real grid model from standard industry outage data. This new capability disprove hypotheses in the literature about cascading, and should lead to new insights as we now can explore real cascading data. Power grid operation relies on voluminous sensor data and is robust to errors to some extent but not to cyber attacks. We designed an anomaly detector which minimizes the “worst-case” probability of error against all possible manipulations of up to n sensor measurements. We proved a necessary condition for the detector to be optimal and derived a heuristic detector, which is asymptotically optimal. Florian Dorfler graduated with a PhD from UCSB and has now joined the faculty at ETH. Iowa State visitor Sarai Mendoza graduated with PhD in Physics from Universidad Michoacana, Mexico, and continues her work on the project as a postdoc.