ПРЕПРИНТ
О результатах, изложенных в препринтах, не следует сообщать в СМИ как о проверенной информации.
This paper presents a deep learning-based fault detection system for smart grid transmission lines, addressing the critical need for robust and accurate fault identification to ensure grid stability and reliability. The proposed methodology integrates Digital Signal Processing (DSP) techniques for comprehensive feature extraction from raw current and voltage signals. Key DSP features, including instantaneous zero-sequence components, total magnitudes, and maximum instantaneous values, were derived to characterize various fault conditions. A multi-layer perceptron (MLP) deep learning model was then developed and trained on these engineered features.On a clean dataset, the model achieved exceptional performance with a test accuracy of 99.72%, and near-perfect precision, recall, and F1-scores of 1.00 for both fault and no-fault classes, as evidenced by a confusion matrix showing only 5 misclassifications out of 1801 samples. To assess robustness, the system was further evaluated using data augmented with Gaussian noise (mean=0, std=0.1). When retrained and tested on noisy data, the model maintained high performance, achieving a test accuracy of 99.22% and consistent precision, recall, and F1-scores of 0.99 across classes. This demonstrates the model's resilience to typical measurement uncertainties. The findings underscore the efficacy of combining DSP-engineered features with deep learning for highly accurate and robust fault detection in complex smart grid environments.
Rahman Sh. 2026. DSP-Assisted Deep Learning for Transmission Line Fault Detection in Smart Grids Under Clean and Noisy Conditions. PREPRINTS.RU. https://doi.org/10.24108/preprints-3115242
