Skip to content

Calibration

Introduction🔗

This library is designed to facilitate parameter calibration of Modelica models within Modelon Impact.

Calibration is a specific type of optimization problem. The objective is to minimize the difference between simulation output and given reference values that the model is supposed to reproduce. The degrees of freedom to be optimized are a given set of model parameters. The problem is solved by finding the set of model parameters that minimizes the difference between the given values for the reference variables and the model output.

The Calibration library is able to find the set of parameters that minimizes the error between the output of a specific model and experiment and a given reference output.

The implementation is limited to the Nelder-Mead method which is suitable for smaller, e.g., single heat-exchanger, calibration problems.

Note

The library is currently in beta-stage, which means that it is still being developed and some parts of the library are subject to change for the final release. We welcome user feedback via our support channels.

Documentation🔗

In addition to the general information about the library in this page, there are a additionally a set of articles that documents the Calibration library:

Features🔗

The main components and functions in this library are:

  • Custom Function: Calibrate - Analysis that performs the model parameter calibration.

  • Custom Web App: Reference file loader - Can be utilized to upload data files to serve as reference results in the calibration process, accessible via the "Apps" menu.

  • Custom Function: Generate reference result - A custom function to generate a reference result from a data file uploaded to a resource folder.

  • Custom Function: Steady state with derivatives - A custom function that uses non-linear algebraic equations solver to find stationary solution for a dynamic models, trying to find conditions where all time derivatives are zero. If applicable, can significantly accelerate the calibration procedure.

  • Modelica Library: Calibration - A collection of Modelica examples that demonstrate how to use the calibration tools effectively.

Link to main article with full information: Library content

Limitations🔗

  • Currently the calibration toolbox does not support trajectory based (time-dependent) calibration. This means that when dynamic simulations are calibrated, only the final result point will be compared against the reference result.