@misc{9246901,
  abstract     = {{Control systems at European Spallation Source (ESS) are large-scale, complex and difficult to model using traditional approaches. Data-driven methods are therefore required to model systems such as the Cryogenic Moderator System (CMS) and Target Helium Cooling System (THCS). This work presents a control-system modeling approach based on the Transformer and Long Short-Term Memory (LSTM), enabling long-range multivariate predictions using control variables as exogenous inputs. The proposed framework integrates these models into a closed-loop setting, enabling iterative, autoregressive, predictions during both training and deployment. The results were promising for the THCS and artificial systems, while performance on the CMS was more limited. This discrepancy was likely caused by insufficient system excitation in the available datasets, owing to the absence of beam-on-target operations at ESS to date. While the proposed data-driven framework demonstrates considerable potential, its overall reliability is currently constrained by the limited diversity of operational data and a lack of out-of-domain testing.}},
  author       = {{Olofsson, Eskil}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Data-Driven Modeling of Large-Scale Control Systems at the European Spallation Source (ESS)}},
  year         = {{2026}},
}