CDAformer: Central Difference Attention Based Transformer for Synchronisation Detection of GIS Knife Gate Actions

Detecting synchronization in three-phase knife gates within Gas-Insulated Switchgear (GIS) is critical for ensuring operational reliability and safety. The primary challenge in achieving accurate synchronization detection lies in navigating the complexity of the signals and overcoming environmental noise. To address these challenges, we introduce the Central difference Attention (CDA) mechanism and the Central difference Attention-based Transformer (CDAformer) model. The CDA mechanism precisely identifies mutation points in three-phase currents, enhancing the model's noise resilience and feature recognition capabilities. Consequently, the CDAformer model significantly improves synchronization detection in three-phase knife gate operations, surpassing the limitations of existing deep learning approaches. Experimental results on real datasets reveal that the CDAformer model outperforms current models in detection performance, marking asubstantial advancement in the field.