Data-Driven Algorithm for Disturbance Classification in Transmission Systems: Design, Implementation, and Experimental Evaluation

This paper investigates event detection and classification in power transmission systems based on real-world disturbance records from a part of the French 110 kV transmission grid. The study relies on measured voltage and current signals rather than simulated data. A novel signal-based algorithm is proposed for detecting abrupt changes in electrical signals and classifying power system events, enabling reliable identification of energizing events, no-load conditions, load changes, and single-phase, phase-to-phase, and three-phase faults. In addition to event classification, key transient characteristics such as fault inception time, event duration, peak short-circuit currents, and inception angle are extracted. By transforming raw disturbance records into structured and interpretable event information, the proposed approach supports knowledge management practices. It facilitates the conversion of operational experience into reusable insights for analysis, training, and continuous system improvement. The proposed framework is benchmarked against conventional symmetrical component–based methods and demonstrates improved detection speed and classification accuracy, particularly for shortduration events. By employing appropriate signal segmentation and RMS-based analysis, the algorithm maintains robust performance in the presence of measurement noise. Overall, the approach supports real-time monitoring, disturbance analysis, and operator decision support in large-scale transmission networks, contributing to enhanced system resilience and reliability.