Sofia Gedda
Centre for Mathematical Sciences, Lund University/Modelon AB, Sweden
Christian Andersson
Centre for Mathematical Sciences, Lund University/Modelon AB, Sweden
Johan Åkesson
Department of Automatic Control, Lund University/Modelon AB, Sweden
Stefan Diehl
Centre for Mathematical Sciences, Lund University/Modelon AB, Sweden
Download article
http://dx.doi.org/10.3384/ecp12076819Published in: Proceedings of the 9th International MODELICA Conference; September 3-5; 2012; Munich; Germany
Linköping Electronic Conference Proceedings 76:84, p. 819-828
Published: 2012-11-19
ISBN: 978-91-7519-826-2
ISSN: 1650-3686 (print), 1650-3740 (online)
Representing a physical system with a mathematical model requires knowledge not only about the physical laws governing the dynamics but also about the parameter values of the system. The parameters can sometimes be measured or calculated; however some of them are often difficult or impossible to obtain in these ways. Finding accurate parameter values is crucial for the accuracy of the mathematical model.
Estimating the parameters using optimization algorithms which attempt to minimize the error between the response from the mathematical model and the physical system is a common approach for improving the accuracy of the model.
Optimization algorithms usually require information about the derivatives which may not always be easily available or which may be difficult to compute due to; e.g.; hybrid dynamics. In such cases; derivative-free optimization algorithms offer an alternative for design and parameter optimization.
In this paper; we present an implementation of derivative-free optimization algorithms for parameter estimation in the JModelica.org platform. The implementation allows the underlying dynamic system to be represented as a Functional Mock-up Unit (FMU); thus enables parameter estimation of models designed in modeling tools following the standardized interface; the Functional Mock-up Interface (FMI); such as Dymola.
