Speaker
Description
The performance of magnet power supplies is critical for a stable operation in storage-ring accelerators. Thus, anticipating any issues in their performance can reduce any downtime associated with their repairs. In this work, we evaluate recurrent Long Short-Term Memory (LSTM) networks and Transformer-based architectures for predicting power-supply temperature and detecting anomalous behavior from multivariate time-series data. We present a comparative analysis of prediction error, computational efficiency, and model interpretability, and discuss implications for real-time deployment in the Advanced Photon Source Upgrade (APS-U) machine-protection and operations workflows
Funding Agency
Work supported by the U.S. Department of Energy, Office of Science, under Contract No. DE-ACO2-O6CH11357.
| In which format do you inted to submit your paper? | LaTeX |
|---|