TR2025-133

Zero-Shot Parameter Estimation of Modelica Models using Patch Transformer Networks

    •  Chakrabarty, A., Forgione, M., Piga, D., Bemporad, A., Laughman, C.R., "Zero-Shot Parameter Estimation of Modelica Models using Patch Transformer Networks", International Modelica and FMI Conference, September 2025.
      BibTeX TR2025-133 PDF
      • @inproceedings{Chakrabarty2025sep,
      • author = {Chakrabarty, Ankush and Forgione, Marco and Piga, Dario and Bemporad, Alberto and Laughman, Christopher R.},
      • title = {{Zero-Shot Parameter Estimation of Modelica Models using Patch Transformer Networks}},
      • booktitle = {International Modelica and FMI Conference},
      • year = 2025,
      • month = sep,
      • url = {https://www.merl.com/publications/TR2025-133}
      • }
  • MERL Contacts:
  • Research Areas:

    Machine Learning, Multi-Physical Modeling, Optimization

Abstract:

This paper introduces a transformer-based generative net- work for rapid parameter estimation of Modelica building models using simulation data from a Functional Mock-up Unit (FMU). Utilizing the MixedAirCO2 model from the Modelica Buildings library, we simulate a single-zone mixed-air volume with detailed thermal and CO2 dynamics. By varying eight physical parameters and randomizing occupancy profiles across 100 simulated systems, we generate a comprehensive dataset. The transformer encoder, informed by room temperature and CO2 concentration out- puts, predicts the underlying physical parameters with high accuracy and without re-tuning (hence, “zero-shot”). This approach eliminates the need for iterative optimization or can be used to warm-start such optimization-based approaches, enabling real-time control, monitoring, and fault detection in FMU-based workflows.