ПРЕПРИНТ
О результатах, изложенных в препринтах, не следует сообщать в СМИ как о проверенной информации.
Accurate fault diagnosis plays a crucial role in ensuring the stability and reliability of smart grid transmission systems. In this study, a state-of-the-art deep learning approach is developed and fine-tuned for both binary fault detection and multi-class fault classification tasks. Using current and voltage datasets, a robust preprocessing pipeline is implemented, including standardization and label encoding for six distinct fault categories. Two tailored deep learning architectures are designed using sequential models with dense and dropout layers to effectively capture the nonlinear characteristics of grid data. Experimental results demonstrate the effectiveness of the proposed framework in anomaly detection, where the fault detection model achieves 100% test accuracy with an almost ideal AUC of 0.9999. The fault classification model achieves an overall accuracy of 86%, showing strong performance across most classes, particularly for single line-to-ground faults. However, detailed analysis using confusion matrices reveals challenges in correctly identifying more complex fault combinations, such as “LLL fault (three-phase fault)” and “LLLG fault (three-phase symmetrical fault).” Supported by comprehensive evaluation metrics and visualizations, the proposed framework demonstrates strong potential for real-time grid monitoring while also highlighting key areas for improvement in handling minority and complex fault classes.
Rahman Sh. 2026. Proposing A Fine-Tuned Deep Learning Framework for Accurate Fault Detection and Classification in Smart Grid Transmission Line. PREPRINTS.RU. https://doi.org/10.24108/preprints-3115324
