2019/09/27 by Kostas Hatalis, Hatalis, Kostas, Parv Venkitasubramaniam +3
Computer Science · Engineering · #Computer science #Computer security #Critical infrastructure #Cyber-attack #Electric power system #Energy (signal processing) #Energy management #Energy management system #Engineering #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Machine learning #Network Security and Intrusion Detection #Power (physics) #Power System Optimization and Stability #Real-time computing #Reliability (semiconductor) #Signal Processing (eess.SP) #Smart Grid Energy Management #Smart Grid Security and Resilience #Smart grid #Stability (learning theory) #cs.LG #eess.SP #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1909.12894
arxiv created 2019/09/27 · openalex publication_date 2019/09/27 · arxiv updated 2019/10/01 · openalex created_date 2022/07/28 · openalex updated_date 2026/08/05
Demand-Side Management (DSM) is a vital tool that can be used to ensure power\nsystem reliability and stability. In future smart grids, certain portions of a\ncustomers load usage could be under automatic control with a cyber-enabled DSM\nprogram which selectively schedules loads as a function of electricity prices\nto improve power balance and grid stability. In such a case, the security of\nDSM cyberinfrastructure will be critical as advanced metering infrastructure,\nand communication systems are susceptible to hacking, cyber-attacks. Such\nattacks, in the form of data injection, can manipulate customer load profiles\nand cause metering chaos and energy losses in the grid. These attacks are also\nexacerbated by the feedback mechanism between load management on the consumer\nside and dynamic price schemes by independent system operators. This work\nprovides a novel methodology for modeling and simulating the nonlinear\nrelationship between load management and real-time pricing. We then investigate\nthe behavior of such a feedback loop under intentional cyber-attacks using our\nfeedback model. We simulate and examine load-price data under different levels\nof DSM participation with three types of additive attacks: ramp, sudden, and\npoint attacks. We apply change point and supervised learning methods for\ndetection of DSM attacks. Results conclude that while higher levels of DSM\nparticipation can exacerbate attacks they also lead to better detection of such\nattacks. Further analysis of results shows that point attacks are the hardest\nto detect and supervised learning methods produce results on par or better than\nsequential detectors.\n