An Unsupervised Approach to Enhance Cyber Resiliency of Power Systems Against False Data Injection Attacks on Voltage Stability

The digital transformation of power system introduces False Data Injection Attacks (FDIAs) on voltage stability that compromises the operational integrity of power grids. Existing detection mechanisms for FDIAs often fall short as they overlook the complexities of cyberattacks targeting voltage stability and rely on outdated models that do not capture the dynamic interplay between power system operations and potential threats. In response to these gaps, this paper proposes a novel FDIA detection method designed specifically for voltage regulation vulnerabilities, aiming to enhance the voltage stability index. The proposed method utilizes an unsupervised learning framework capable of identifying cyberattacks targeting voltage regulation. A bi-level optimization approach is put forward to concurrently optimize the objectives of both attackers and defenders in the context of voltage regulation. The effectiveness of this approach is validated through comprehensive training and testing on a variety of attack scenarios, demonstrating superior generalization across different conditions. Extensive simulations on the Iberian power system topology, with 486 buses, show that the proposed model achieves more than 93% detection rate. These results highlight the robustness and efficacy of the proposed strategy in strengthening the cyber resilience of power systems against sophisticated FDIA threats on voltage stability.