A Semi-Supervised Learning-Based Method for Classifying the Status of DC Charging Piles

Latent faults in DC charging piles are difficult to reproduce, and manual fault diagnosis is time-consuming, hindering efficient operation and maintenance. This paper proposes a semi-supervised learning-based approach for classifying the operational status of DC charging piles. A comprehensive status assessment framework is developed, integrating electrical, operational, and safety data. Key health factors are selected through correlation analysis, and a semi-supervised learning method is employed to train a classification model, combining multi-feature extraction with artificial intelligence techniques. Specifically, statistical features derived from weekly data are used to construct health factors, which are then fused using an artificial neural network for accurate classification of the charging pile's status. Extensive validation on real-world data demonstrates that the proposed method achieves a classification accuracy exceeding 91%.