Collocation methods in JModelica.org

Abstract: In this thesis, we use CasADi to implement a new optimization algorithm in the open-source platform JModelica.org. CasADi is a tool for computing derivatives using automatic differentiation, which is tailored for optimal control. JModelica.org is a platform for simulation and optimization of physical systems created using the Modelica modeling language. The implemented optimization algorithm is based on direct collocation using Radau or Gauss collocation schemes. We provide a thorough presentation of how a dynamic optimization problem described by Modelica and Optimica code is transcribed into a nonlinear programming problem using direct collocation. This nonlinear programming problem is then solved using CasADi’s interface to Ipopt, a numerical solver for optimization problems. The implemented algorithm is compared to a similar and already existing optimization algorithm in JModelica.org in five different benchmark problems, including a distillation column and a combined cycle power plant. The new algorithm compares favorably to the other JModelica.org algorithm in a majority of the cases.