IAP DYSCO Study Day : Dynamical systems, control and optimization: Program and abstracts pages:15
IAP DYSCO Study Day : Dynamical systems, control and optimization location:Mons, Belgium date:24 May 2013
Having the possibility to systematically evaluate objectives of different nature at the same time is becoming more and more crucial to operate processes, plants and production sites under more sustainable conditions. Most often, this multi-objective nature is tackled by combining all individual objectives into a global weighted sum. However, selecting appropriate weights for the individual objectives is a non-trivial task, especially when no price information is available and/or the different individual objectives are incommensurable. Unfortunately, systematically varying the weights often does not always yield an accurate indication of the different alternative solutions and the trade-offs between them. This research introduces a strategy that exploits alternative scalarisation based multi-objective optimisation methods (e.g., Normal Boundary Intersection (NBI), Enhanced Normalized Normal Constraint (ENNC)) to efficiently and accurately generate the set of trade-off or Pareto optimal solutions. The decision maker can pick a solution from this set according to his/her preferences. This selected solution can then explicitly be linked via analytical relations to a set of weights to be used in a classical weighted sum approach. Hence, the procedure is particularly attractive for the systematic tuning of (Nonlinear) Model Predictive Controllers ((N)MPC) which often involve optimisation problems with a weighted sum as objective function. The proposed procedure is illustrated on several (bio)chemical case studies. To efficiently solve the resulting multi-objective dynamic optimisation problems, these scalarisation based methods must be integrated with deterministic Single and Multiple Shooting approaches allowing the exploitation of fast deterministic solvers. To this end, ACADO Multi-Objective (www.acadotoolkit.org) has been employed, which is a flexible toolkit for solving dynamic optimisation or optimal control problems with multiple and conflicting objectives.