A Unified System Theoretic Framework for Cyber Attack-Resilient Power Grid

Abstract:

We  developed  an  analytical  and  computational  framework  for  vulnerability analysis of dynamical power networks to link and node-based attacks.   For stochastic link failure attacks, we provided a systematic procedure for ranking links  (transmission  lines)  of  power  network  in  the  order  of  their  critical importance to maintain synchronization among network generators. The computational procedure for the identification of critical links relies on solving a Linear Matrix Inequality (LMI). A novel metric employing tools from system theory is developed to analyze the vulnerability of power networks to simultaneous attacks on network nodes and links. The developed metric essentially captures the impact of an attack and the relative difficulty of detecting the attack, where the attack is deemed most vulnerable if it is not only impactful but also difficult to detect. A convex optimization-based framework is developed for identifying links in dynamical networks such that the removal of these links lead to maximum degradation of the controllability of the network. A game-theoretic framework has been developed to model Cyber-Physical Security of Wide Area Monitoring Protection and Control (WAMPAC) applications. This framework captures both the cyber and physical costs incurred by the attacker and the defender and their inherent interdependencies. We have created a federated CPS security testbed environment linking Iowa State's PowerCyber testbed with University of Southern California Information Sciences Institute's DETER testbed. We have successfully demonstrated  realistic  cyber-attack  defense  experiments  on  Wide-Area Protection (Remedial Action Scheme) involving coordinated cyber-attack and defense  capabilities  over  the  federated  testbed.  This  attack  defense demonstration was showcased at the Smart America Challenge Expo in Washington, D.C. in June 2014. A real-time data-driven voltage and angle stability monitoring algorithm developed in our prior work is implemented and tested on real time power system simulator (RTDS). The goal will be to make use of the algorithm in the development of a data-driven approach for attack impact characterization.