Advanced intelligent quench diagnostics for high temperature superconducting coils based on principal component analysis of voltage harmonic ratios and Support Vector Machine

Abstract

The utilization of superconductors in modern aviation, power and energy, space, healthcare, and quantum sectors offers substantial advantages, including significant energy savings, higher reliability, lower CO${}_2$ emissions, higher efficiencies, and increased power densities. However, the quench phenomenon – transition from superconducting state to normal state – presents a major challenge, particularly in high temperature superconductors (HTSs), and especially in sensitive applications including fusion energy and aviation sectors. Most of the existing quench detection systems, i.e., conventional techniques, simply work based on threshold, and therefore, they have challenges being inflexible, non-intelligent, slow, and with the potential of missing quench or raising false alarm. This study proposes an advanced quench diagnostic technique aiming at enhancing reliability of HTS devices. This diagnostic technique quantifies quench events by analyzing the ratio of a selected harmonic component’s amplitude, including the second to ninth harmonics, to that of the fundamental frequency in voltage measurement. Fast Fourier Transform (FFT) is employed to analyze the frequency-domain characteristics of voltage signals in both quench and non-quench stages. Amplitudes of the main harmonic frequencies were measured in the spectrum and calculated the harmonic amplitude ratio. Principal Component Analysis (PCA) was applied to reduce dimensionality and extract the most relevant features associated with quench events. The original 17 harmonic ratio features were transformed into three principal components per label, preserving essential information of the dataset for classification. The Support Vector Machine (SVM) was used as an automatic intelligent decision-making technique to discriminate new cases of quench from non-quench condition. Finally, after continuous optimization of the PCA and SVM model, this novel approach, based on experimental data, demonstrated to reach an accuracy of 100% and promised a fast, more flexible, and reliable method for quench diagnostics. In addition, to validate and demonstrate the generalizability of the proposed quench detection method, this diagnostic technique was tested for a totally new HTS coil sample with different type, configuration, and turns number, which high discrimination accuracy was observed. This is a real testimony to the effectiveness of the proposed technique which is robust and generalized. However, further tests on larger and more diverse datasets are underway to accelerate its future applicability in fusion, electric transportation, and renewable energy sectors.