MVP-ROM: Masked Vulnerability-Preserving ROM for Power Systems

To date, the ability to assess vulnerabilities, resiliency, and priorities for CI protection remains a formidable challenge. Existing CI modeling techniques face several challenges such as data access and collection, incomplete quantification of dependency types, infeasible scalability, challenge of incorporating high-fidelity dynamic behavior to respond to disruptive changes. This project proposes the MVP-ROM software to build reduced order models of high-fidelity simulations capable of capturing nonlinear dynamics efficiently for power systems vulnerability analysis. MVP-ROM is a novel approach that integrates random matrix theory under a port-Hamiltonian modeling paradigm, both well-established in the computer science and control communities. MVP-ROM enforces the port-Hamiltonian form of equations to preserve the structure of the modeled system, while leveraging the efficiency of random methods to capture high-dimensional nonlinear dependencies with >99% probability and tight upper bounds on the error. This objective of this project is to integrate MVP-ROM under the open-source commercial-ready Cortix platform for stakeholder testing and evaluation. MVP-ROM will be test-hardened on a hardware-in-loop simulation and subject to a purple-team hackathon to test resilience against insider attacks. The project will demonstrate three MVP-ROM capabilities - (a) automated reduction of nonlinear power system models, (b) efficient extraction of the ROM degrees of freedom as a cyber-resilient data communication and remediation protocol, and (c) non-weaponizable model masking to prevent insider-assisted reverse engineering. Facilitated by our testbed partners, MVP-ROM finds commercial applications in a $2.4 billion power systems simulation market and scaling opportunities in the high-growth data masking and synthetic data generation markets.