Implementation of Grey-Box Identification in JModelica.org

Abstract: Grey-box identification is a tool to identify and improve nonlinear system models by estimating parameters. The estimation is done by optimizing a cost function using measurement data. The robustness of the estimations can then be analyzed with statistics. JModelica.org is a platform for modeling and optimization of dynamical models. In order to do grey-box identification one need models and be able to optimize. JModelica.org supports modeling and optimization so it has a huge potential to support grey-box identification. So far there is no complete solution for grey-box identification in JModelica.org. This work is focusing on how to implement greybox identification in JModelica.org in order to estimate parameters for nonlinear models. The theory of grey-box identification has been investigated as well as the possibilities with JModelica.org. Finally, an interactive method to estimate model parameters and a method to calculate the confidence intervals for the estimates have been implemented. The implementation has been tested for nonlinear models and works as expected.