Detection of False Data Injection Attacks (FDIA) Targeting Distribution Transformers in Smart Grids Using Hybrid Artificial Intelligence Methods

Communication technologies in modern power systems have rendered energy grids vulnerable to False Data Injection Attacks (FDIA). This study evaluates FDIA detection on the IEEE 118-bus test system using pandapower. A synthetic dataset of 2000 samples was partitioned into 75%/25% training/test splits. Three models — LSTM networks, Isolation Forest and One-Class SVM — were comparatively assessed. LSTM achieved the highest performance with 97.0% recall, F1-score of 0.965 and AUC of 0.987, outperforming Isolation Forest (93.0%) and One-Class SVM (90.0%). These results confirm that deep learning approaches are highly effective for smart grid cybersecurity and compatible with IEC 61850-based infrastructures.