Sandia Labs FY21 LDRD Annual Report


Advanced techniques for optimal power system emergency control. A major disturbance to the electric power grid can lead to a severe power outage, leaving residential, business, or industry customers without power for extended periods of time. An outage can be exacerbated through cascading power outages caused by protective grid devices tripping power connections automatically to try maintaining dynamic stability. This effort, part of the Resilient Energy Systems LDRD Mission Campaign, focused on investigation and application of direct shooting methods for optimal control of power systems after a disturbance. Features of this framework include the options of constant, piecewise linear, and piecewise cubic basis functions, as well as a direct single-shot method using adaptive time for control action updates. Researchers used the Matlab-based nonlinear programming solver fmincon for optimization and used ode15s for numerical solutions of the stiff differential and algebraic equations. The outcome of this project will inform resilient energy systems. (PI: Bryan Arguello)

Moving target defense for space systems. Space Policy Directive-5 Cybersecurity Principles for Space Systems describes both the cyber threat to space systems and the need for these systems to be secure and resilient against cyberattacks. Nation- state adversaries can disrupt critical infrastructure through cyberattacks targeting systems of networked, embedded computers. This project, funded by the Science and Technology Advancing Resilience for Contested Space (STARCS) LDRD Mission Campaign, developed a patented moving target defense (MTD) algorithm that adds cyber resilience to space systems by improving their ability to withstand cyberattacks. MTDs create dynamic, uncertain environments that seek to confuse the attacker and attempt to defeat cyber threats. Most proposed cyber resilience solutions focus on or require detection of threats before mitigations can be implemented, a significant technical challenge. The new MTD approach avoids this requirement while creating informational asymmetry that favors defenders over attackers. Researchers conducted three key experiments: i.e., functional, cyber resilience, and machine learning (ML), which helped quantify the benefit of the LDRD team’s approach to cyber resilience against different types of cyberattacks. Results show a 97% reduction in adversarial knowledge on a MIL-STD-1553 network. A collaboration with Purdue University using ML to defeat the MTD algorithm highlighted the strength of the algorithm by showing a small change in one of the algorithm parameters substantially decreased the success rate of the LSTM machine learning model. Further, the generalizable algorithm led to Sandia working with a small business that plans to use this technology to mitigate ransomware and disseminate MTD technology for the U.S. government. (PI: Chris Jenkins)



Made with FlippingBook Ebook Creator