Data-Driven Adaptive Open-Circuit Fault Localization Method With Reduced Sensors for Magnetic-Network Energy Router

Fault localization is a critical challenge for magnetic-network energy router (MNER) consisting of numerous switches. This letter proposes a data-driven adaptive open-circuit fault localization method with reduced sensors for MNER. Through the extraction of adaptive data ensembles from inter-arm voltage sampling channels, numerous time-sliding ensemble anomaly encoders (EAEs) are constructed to isolate the abnormal channels. Next, by means of buffering and decoding the EAE's output, the faulty switch can be effectively localized. The proposed method not only reduces fault feature dimensionality, thereby minimizing the requirement for extra hardware, but also eliminates the reliance on mathematical model and threshold setting. A 2-kW MNER experimental platform is constructed to verify the effectiveness of the proposed method.