Anomaly Detection for HTS Cable Using Stacked Autoencoder and Reflectometry

Since a high–temperature superconducting (HTS) cable has a cooling process and the possibility of a quench, a real–time monitoring technique is essential than other conventional power cable. Conventional reflectometry–based monitoring parameters have inherent disadvantages in that the boundary between normal and abnormal cannot be determined and multiple baselines cannot be considered during the monitoring process. In this paper, a stacked autoencoder–based anomaly detection technique for HTS cable is newly proposed. The autoencoder monitors the condition of HTS cable in real–time by calculating an anomaly score from the input data. The proposed method can automatically set the boundary between normal and abnormal and has the advantage of using multiple baselines. The performance of the proposed method is verified by emulating a local quench in a testbed containing a commercial 22.9 kV HTS cable and compared with that of the conventional monitoring parameter. Based on the advantage of being able to set multiple baselines, it is expected that the proposed technique can be used for real–time monitoring of HTS cable to enhance the reliability of the systems.