Abstract: Traditional signal processing methods often fail to effectively capture the complex nonlinear characteristics of lightning surge current signals in distribution networks lightning fault identification. To overcome this limitation, a method for lightning surge current identification in distribution networks based on an improved Transformer network with multi-dimensional feature inputs is proposed. Leveraging the self-attention mechanism of the Transformer network, intricate dependencies and patterns are extracted from multi-dimensional features, and a model classification module is incorporated to accurately classify lightning surge currents, significantly enhancing identification accuracy. The simulation model is built on the Simulink platform to validate the proposed algorithm. Experimental results demonstrate that the proposed method achieves an identification accuracy of 99.12% in classifying lightning surge currents, verifying the effectiveness of the approach.

Key words: 10 kV distribution network, lightning surge current identification, multi-dimensional feature extraction, Transformer model