An Intelligent Fault Classification and Protection Method for Distribution Networks Integrating Temporal and Spatial Features

With the integration of a large number of new energy and power electronic devices into, the fault characteristics of distribution networks have become more complex and diverse. The safety and stability of traditional distribution networks are seriously threatened. To improve the accuracy of fault classification in distribution networks, we designed a dual-modal input neural network architecture to enhance the accuracy and reliability of distribution network fault classification. This architecture consists of two branches: one branch utilizes convolutional neural networks to extract spatial features; the other branch employs a Transformer network to extract temporal information, capturing time-dependent features. To achieve more efficient feature fusion, the fusion module uses a multi-head attention mechanism, comprehensively processing temporal and spatial information, and finally, through the target classification module, achieves precise classification of distribution network faults. To validate the framework, the dualmodal neural network architecture was implemented in a channel-free feeder automation protection device, with experimental verification conducted via the real-time digital simulation platform RT-LAB. The results demonstrate that the proposed method significantly improves the accuracy and reliability of distribution network fault classification, while also enhancing the correctness of automated protection actions under diverse fault conditions.