Abstract—Sixth Generation (6G)-enabled smart grid communication requires a robust installation and secure framework to manage high-speed and low-speed data communication. Existing models face challenges such as a lack of labeled data, poor temporal modeling, weak interpretability, inadequate behavioral profiling, and difficulties with edge-deployment. Additionally, the massive influx of data increases vulnerabilities to faults, cyberattacks, and zero-day anomalies. This study presents an innovative deep learning architecture that utilizes semantic encoding and self-supervised learning to detect both known and unknown anomalies within a 6G-enabled smart grid communication system. The proposed approach begins with cleaning and normalizing the electrical sensor logs. Features are extracted using a One-Dimensional Convolutional Neural Network (1D CNN), and a Multi-Modal Vision Transformer (MM-ViT) transforms sensor sequences into semantic event tokens. The self-supervised model combines Contrastive Predictive Coding (CPC), Temporal Convolutional Networks (TCN), and Attention mechanisms to enable robust temporal anomaly detection from sensor data. While TrustNet combines TCN and Attention, it models network session behavior and supports dynamic trust evaluation. Compared to the existing anomaly detection methods, the proposed semantic and self-supervised methods perform extremely well across many metrics, achieving a 99.34% accuracy. This demonstrates its effectiveness and suitability for reliable anomaly detection in 6G edge-enabled smart grid infrastructures.
 
Keywords—anomaly detection, attention mechanism, behavioral profiling, self-supervised deep learning, semantic interpretation, smart grid


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).