PI: Parv Venkitasubramaniam, Lehigh University
Cyber-enabled Demand Side Management (DSM) plays a crucial role in smart grid by providing automated decision-making capabilities that selectively schedule loads on these local grids to improve power balance and grid stability. Typically relying on Advanced Metering Infrastructure for two way communication capabilities, such DSM methods consider current and forecasted power generation, as well as load types and requirements, availability of storage, and flexible pricing incentives and tariff structures to either dispatch load (either directly via controllable loads or indirectly via pricing signals) in response to inadequate and/or uncontrollable power supply.
This project investigates the detection and mitigation of data falsification attacks in the DSM feedback loop that targets specific components or operations of the distribution grid. We utilize supervised learning models for the spatio temporally dependent demand processes and a non linear stochastic pricing mechanism to capture the interdependence within the loop, and develop statistical inference methodologies to detect, localize and identify the potential targets for data falsification attacks. We further investigate game theoretically a robust defense mechanism to protect against such attacks.