Intelligent algorithm-driven data fusion and accurate operational state perception for feeders in new distribution networks

Le résumé fourni par la source

The integration of high-proportion renewable energy sources intensifies complexity and uncertainty in distribution network operations, necessitating precise feeder state perception for reliability assurance. This paper proposes an intelligent algorithm-driven framework for multi-source data fusion and operational state identification in modern distribution networks. The methodology integrates rigorous data preprocessing, comprehensive feature engineering capturing temporal dynamics and physical characteristics, and a fused machine learning approach combining ensemble classifiers (Random Forest, Gradient Boosting), neural networks (MLP), and unsupervised anomaly detectors (Isolation Forest, One-Class SVM) for robust state classification. DBSCAN clustering further identifies latent operational patterns. Validated on simulated feeder data encompassing normal, voltage anomaly, voltage sag, and overcurrent states, the framework achieves 95.2% anomaly detection accuracy and a 93.8% F1-score for state identification. Multi-dimensional visualizations demonstrate the system's capacity to deliver precise, interpretable state awareness, highlighting its significant potential for intelligent maintenance and proactive management in evolving distribution grids.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.