Fast Quench Detection in SFCL Pancake Using Optical Fibre Sensing and Machine Learning

Abstract

A fast hotspot detection technique has been implemented and patented at École Polytechnique Fédérale de Lausanne (EPFL) under the European Union project FastGrid. The optical fibre sensing based technique uses the Mach-Zehnder Interferometer (MZI) and can detect even singular hotspots in the superconductor within 15ms. The MZI setup is very sensitive to external perturbations and mechanical stresses, which can manifest in the output along with the response to hotspots. The disturbance the setup is subject to, varies with the sample being tested and the environment it is being tested in. For example, the sample length and configuration, in addition to the routing of the optical fibre can determine the extent of the stresses on the optical fibre. Previous publications, showcased experiment results on relatively simple sample configurations which were straight HTS tapes with short lengths (0.3m or 1m). This manuscript will investigate the technique feasibility for a more complicated sample type: a pancake prototype for a superconducting fault current limiter (SFCL), comprising 12m of HTS tape in a bifilar winding, with taped optical fibre along the length of the conductor. The manuscript will also give a brief overview of a machine learning based post processing technique for the experimental data, developed at EPFL to supplement the MZI based quench detection for easy and reliable quench detection in large scale applications.