TR2020-031

Nonlinear State Estimation with FMI: Tutorial and Applications


    •  Laughman, C.R., Bortoff, S.A., "Nonlinear State Estimation with FMI: Tutorial and Applications", American Modelica Conference 2020, Tiller, M. and Tummescheit, H. and Vanfretti, L. and Laughman, C. and Wetter, M., Eds., DOI: 10.3384/​ECP20169186, March 2020, pp. 186-195.
      BibTeX TR2020-031 PDF
      • @inproceedings{Laughman2020mar,
      • author = {Laughman, Christopher R. and Bortoff, Scott A.},
      • title = {Nonlinear State Estimation with FMI: Tutorial and Applications},
      • booktitle = {American Modelica Conference 2020},
      • year = 2020,
      • editor = {Tiller, M. and Tummescheit, H. and Vanfretti, L. and Laughman, C. and Wetter, M.},
      • pages = {186--195},
      • month = mar,
      • publisher = {Linköping University Electronic Press},
      • doi = {10.3384/ECP20169186},
      • issn = {1650-3686},
      • isbn = {978-91-7929-900-2},
      • url = {https://www.merl.com/publications/TR2020-031}
      • }
  • MERL Contacts:
  • Research Areas:

    Control, Dynamical Systems, Multi-Physical Modeling

Abstract:

One of the key uses enabled by the functional mockup interface (FMI) standard is the ability to combine Modelica models governed by differential-algebraic equations with measurement data to systematically estimate unmeasured quantities in physical systems. While it is clear how this might be done in theory, many implementation challenges can make this difficult in practice. This paper provides a tutorial connecting the mathematical formulation of two different estimators, the extended Kalman filter (EKF) and the ensemble Kalman filter (EnKF), to an FMIbased Modelica implementation of these estimators. The efficacy of these methods are demonstrated on an example of a small motor model and a larger thermodynamic model of a building, and some of the advantages and disadvantages of this FMI-based approach to estimation are discussed, as well as limitations of FMI associated with constraint management for these estimation methods. The code for the motor example is publicly available and is attached to this publication.