A concept of virtual power plant (VPP) emerges to participate in electricity wholesale market based on aggregated resources combining photovoltaic and wind turbine. False data injection attack (FDIA) can be one of effective and ingenious options of the VPP to avoid the curtailment dispatch and maximize its revenue in the market. Most FDIA studies, which produce the desired outcome for attackers, include a strong assumption that system information is known. We aim to perform FDIA in the view of the VPP without knowledge on the topology of the system under the assumption that the net-load value of each bus can be accessible to the attacker. In addition to performing FDIA without system information, we showed a reinforcement learning approach finding load combinations that avoid curtailment dispatch is a well-performed solution.
DOI: 10.22982/NEXTWP.2022.12.2
Abstract
A concept of virtual power plant (VPP) emerges to participate in electricity wholesale market based on aggregated resources combining photovoltaic and wind turbine. False data injection attack (FDIA) can be one of effective and ingenious options of the VPP to avoid the curtailment dispatch and maximize its revenue in the market. Most FDIA studies, which produce the desired outcome for attackers, include a strong assumption that system information is known. We aim to perform FDIA in the view of the VPP without knowledge on the topology of the system under the assumption that the net-load value of each bus can be accessible to the attacker. In addition to performing FDIA without system information, we showed a reinforcement learning approach finding load combinations that avoid curtailment dispatch is a well-performed solution.