LTH-image

Efficient symbolical and numerical algorithms for nonlinear model predictive control with OpenModelica

Bernhard Bachmann, Fachhochschule Bielefeld University of Applied Sciences

Abstract:

During the last decade nonlinear model predictive control (NMPC) has become increasingly important for today’s control engineers. In order to apply NMPC a nonlinear optimal control problem (NOCP) must be solved, which needs a very high computational effort. Nowadays, corresponding modeling of the system dynamics and formulation of the optimization problem can be done in Modelica and Optimica, respectively.

State-of-the-art NOCP solution algorithms are based on multiple shooting and/or collocation algorithms. Only parallelizing these time-consuming algorithms can give reasonable performance appropriate for online-applications. In addition, efficient symbolical and numerical treatment of the underlying model formulation (e.g. matching, sorting, and tearing) are necessary, when solving NOCP involving complex system. Furthermore, for performance and stability reasons NOCP corresponding symbolically derived Jacobian and Hessian matrices and their efficient computation (e.g. identify and utilize the sparsity pattern of Jacobian matrices) are needed.

This talk will discuss these mathematical aspects of NMPC as well as the current and future implementation of efficient, partly parallelized symbolical and numerical algorithms available in and with OpenModelica.

Presentation Slides

Biography:06.05.1964 Born in Waldshut (Germany)
1984 – 1989 Study courses in mathematics and computer science at the University of
             Karlsruhe. Diploma in Mathematics
1989 – 1993 Doctoral thesis (Dr. phil.) at the Institute of Mathematics, University Zürich
              Switzerland (Prof. Dr. Rüdiger Verfürth)
1989 – 1998 Scientist at the Theoretical Group of ABB Corporate Research Center,
             Baden-Dättwil, Switzerland
1994 – 1995 Research Stay at ABB Transmission Technology Institute, Raleigh, North
  and 1997 Carolina, USA
1998 Research Stay at Dynasim AB, Lund, Sweden
1998 – 1999 Senior Scientist at ABB Network Partner, Baden, Switzerland
1999 – current Professor (C3), Applied Mathematics, University of Applied Sciences
                Bielefeld
2004 and 2007 Visiting Professor at Universit Linköping, Sweden
1996 – current Founding member of Modelica Association
2004 – current Member of Open Source Modelica Consortium
2008 Chairman and Editor of the 6th International Modelica Conference,
    Bielefeld
2008 – current Founding member of the Research Group Applied Mathematical
                Modeling and Optimization (AMMO), University of Applied Sciences
               Bielefeld
2005 – 2008 Vice-Dean of Department of Mathematics and Engineering
2005 – current Head of Bachelor-program Applied Mathematics
2008 – current Head of Master-program Optimization and Simulation
Patent holder of two worldwide patents and one European patent
Expert on mathematical aspects of object-oriented modeling and simulation, see given tutorials at the last five International Modelica Conferences 2005-2011.