Detecting False Data Injection and Unstable Operation in Smart Grid via System-Aware Graph Boundary Learning
StarGNN flags smart-grid instability and unseen false-data injection using only stable training data.
StarGNN learns the stable operating region of decentralized smart-grid control from clean stable configurations and uses one abnormality score to flag both genuine instability and false data injection. Each configuration is a producer-consumer star graph processed by a role-aware graph neural network, with physics-constrained pseudo-negatives replacing unavailable unstable and attack labels. Across nine unseen FDI scenarios, detection on stable cases ranged from 0.780 to 0.973, post-attack instability recall stayed between 0.972 and 0.999, and recall against a stronger adaptive attacker was 0.899.
- Learns the stable region only from clean stable data.
- One score covers instability and unseen false-data injection.
- Attack detection on stable cases was 0.780 to 0.973.
- Adaptive attacker recall was 0.899.
From the original article
1 min read at arxiv.orgCyber-physical power systems increasingly rely on data-driven tools to detect instability and support reliable grid operation. However, reliable stability prediction is difficult when unstable operating configurations are rare, sensitive, or unavailable during model development, since collecting such data safely and at scale is often impractical.
By Emad Efatinasab · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.