Dynamic game-based defensive primary frequency control system considering intelligent attackers. (December 2021)

• Record Type: Journal Article
• Authors: Ding, Zhetong; Chen, Chunyu; Cui, Mingjian; Bi, Wenjun; Chen, Yang; Li, Fangxing
• Is Part Of: Reliability engineering & system safety. Volume 216(2021)
• Publication Date: 2021-12
• Subjects: Cyber–physical system -- Cyber attack-tolerant control -- Dynamic game -- Reinforcement learning; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2021.107966
• Languages: English
• ISSNs: 0951-8320

Abstract

Abstract: Due to cyber security vulnerabilities of electrical cyber physical systems, cyber attacks against frequency-active power controllers become an emerging problem. Cyber attack-tolerant frequency control strategies are thus presented to reduce abnormal frequency excursions. These strategies mainly treat attackers as 'dumb' agents. However, attackers in some situations are more like 'intelligent' agents with initiative, who can actively and purposefully respond by considering actions of defenders. In this paper, this actively confrontational interaction is modeled by a dynamic game. We propose two types of game scenarios in the primary frequency control system. Improved Nash equilibrium-based defending strategies are studied using reinforcement learning algorithms. By testing the proposed dynamic game-based strategies on the benchmark system, the validity is verified. The proposed analytical procedure concerning dynamic games can also be extended to the study of other industrial control applications. Highlights: A dynamic game-based defense framework is proposed for the primary frequency control system. Strictly and loosely competitive game scenarios are used to characterize different strategic interactions between defenders and attackers. Equilibrium-based defense and attack strategies are validated using the benchmark system.