A CGAN-based adversarial attack method for data-driven state estimation

With the development of data science and technology, the data-driven algorithms under the new power system architecture provide support for real-time and accurate state estimation (SE). However, the adoption of data-driven algorithms also introduces new security risks to power systems due to their i...

Full description

Saved in:
Bibliographic Details
Main Authors: Qi Wang, Jing Zhang, Jianxiong Hu, Shutan Wu, Shiyi Hou, Yi Tang
Format: Article
Language:English
Published: Elsevier 2025-09-01
Series:International Journal of Electrical Power & Energy Systems
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S0142061525004260
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the development of data science and technology, the data-driven algorithms under the new power system architecture provide support for real-time and accurate state estimation (SE). However, the adoption of data-driven algorithms also introduces new security risks to power systems due to their inherent limitations, such as lack of interpretability and poor robustness. Their sensitivity to small perturbations and the limitation of generalization ability make them vulnerable to adversarial attacks. Therefore, to reveal the security threats of adversarial attacks, a conditional generative adversarial network (CGAN) is innovatively adapted to the attack vector generation scenario of the power system in this paper. The feature decoupling and parameter-guided generation capabilities of the model are leveraged to produce adversarial samples that are both covert and threatening. In the proposed CGAN-based adversarial attack method, attackers can execute effective attacks by accessing grid measurements, without requiring any prior knowledge or deep understanding of the system. Simulation results demonstrate that the proposed method could compromise the state estimation process while evading bad data detection, all with reduced attack costs. This enables attackers to manipulate the output of the state estimation model according to their intentions, which represents an improvement over previous work. Comparative analysis with other traditional adversarial attack methods further indicates its effectiveness, thereby uncovering critical security vulnerabilities in data-driven state estimation algorithms.
ISSN:0142-0615