In this letter, a data-driven cyber-attack detection method for islanded dc microgrids is proposed. Data are collected by monitoring the behavior of an intelligent attacker who is able to bypass… Click to show full abstract
In this letter, a data-driven cyber-attack detection method for islanded dc microgrids is proposed. Data are collected by monitoring the behavior of an intelligent attacker who is able to bypass the conventional cyber-attack detection algorithms and disrupt the operation of the system. The reinforcement learning algorithm emulates the actions of such intelligent attacker, who exploits the vulnerability of index-based cyber-attack detection methods, such as discordant detection algorithm. The data are then used to train a neural-network-based detector that complements the conventional method with additional capability to detect a larger set of possible attacks. Through experiments, the effectiveness of the proposed method is validated.
               
Click one of the above tabs to view related content.