ModelOpt

ModelOpt

Mathematical Parameter Optimization of Modelica Models

XRGModelOpt

ModelOpt is a an optimization application for minimizing cost functions or deviations from a design or measurement value. It uses Dymola/Modelica for modeling and target function definition as well as for simulation (from version Dymola 6.1 to the latest version). A simple and intuitive graphical user interface guides the user step by step through an optimization project. Any model depending on float parameters can be optimized by a selection of robust algorithms which are designed for either searching a local or even a global optimum of a target function. The progress of the iterative algorithm can be observed in a plot diagram or from a table indicating the currently best solution. A separate batch-mode allows to investigate one and the same model for different scenarios in a single optimization run.

Optimization of a Modelica model

Optimization

Important Features of ModelOpt

  • Stand-alone application for Windows operating systems (Win XP and newer)
  • Intuitive GUI
  • Requires an installation of Dymola 6.1 or newer
  • Optimization of free float parameters for minimizing any Modelica target function
  • Local optimization algorithms (NELDER-MEAD, COBYLA, NEWUOA-BOUND)
  • Global optimization algorithms (DIRECT, Randomized DIRECT-L)
  • Spezification of stop criteria (tolerance, time elapsed)
  • Limitation of simulation time and parameter values
  • Batch simulation for different data sets
  • Modelica Library ModelOptLib for easy integration in own models

Modelica Integration

ModelOptLib is an auxiliary Modelica library which allows an easy integration of ModelOpt into any Modelica model. Components for the most common mathematical formulation of calibration or minimization problems are provided within this library.

ModelOptLib Example

Acknowledgement

The development of ModelOpt was partly supported by the Bundesministerium für Bildung und Forschung in the research project OPENPROD (number 01IS09029G) which ended in June 2012.