Visually inspired power quality disturbances recognition via Gramian Angular Difference Field, swin transformer and temporal–frequency–symmetry attention

Accurate and real-time identification of power quality disturbances (PQDs) remains a pressing challenge in modern power systems, especially with the increased penetration of renewable energy sources and the resulting complexity of electrical networks. This study proposed a novel hybrid framework for PQD recognition, integrating Gramian Angular Difference Field (GADF) image encoding, the Swin Transformer for hierarchical local feature extraction, and a Temporal-Frequency-Symmetry Enhanced Global Attention Mechanism (TFSGAM) for capturing global and domain-specific features. The one-dimensional PQD signals are first converted into two-dimensional images using GADF, effectively preserving temporal dependencies. The Swin Transformer exploits local contextual information, while TFSGAM further enhances feature representation by incorporating temporal position encoding, frequency-domain awareness, and symmetry-based spatial attention. Experimental results on synthetic and real-world datasets demonstrated that the proposed framework achieved classification accuracy exceeding 98 % under most noise conditions, while maintaining strong robustness across 25 PQD types and ensuring real-time applicability with an average inference time of 169 ms/sample. Comparative studies with state-of-the-art methods and extensive ablation analyses confirmed that this approach exhibits strong robustness in noise scenarios with SNR = 20/30/40 dB.