Relief-MLCBF for FDIA Detection in Wind-Solar Grids with Incomplete Data
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
To address incomplete measurements and scarce attack samples in False Data Injection Attacks (FDIA) detection for photovoltaic-wind hybrid grids, this study proposes the Relief-MLCBF framework, integrating data augmentation (representation learning + latent space sampling with contrastive learning), feature fusion (missing measurement coding + topology priors + ASF from 5 unsupervised algorithms), and Relief-based feature selection with a cascaded multi-label classifier (XGBoost, MLRF, MLERF). Validated on the IEEE-39 bus system, it achieves 88.2% AUC under incomplete data, over 91% accuracy, and 90.65% F1 score at 50:1 imbalanced data, outperforming SMOTE, CNN, and autoencoder with superior robustness and generalization.
Fields of Study & Subjects
Smart Grid Security and Resilience
Power System Optimization and Stability
Electricity Theft Detection Techniques
Publication details
- Journal / source
- Journal of Physics Conference Series
- Volume
- 3198
- Issue
- 1
- Pages
- 012006–012006
- Publication date
- 1 March 2026
- Language
- EN