©2000 - 2002 Dipl.-Ing. Tobias Jockenhövel. All rights reserved.

[OptControlCentre] [Offline Menu] [Edit Menu] [Online Menu] [Sim. Annealing] [System Requirements]

There are a lot of excellent dynamic simulation models available in the industry. It is often desired to find optimal control trajectories and control parameters without intensive modifications or rewriting of these models. To accomplish this task, a stochastic approach using Simulated Annealing was developed for the OptControlCentre by Richard Faber and Tobias Jockenhövel.

The Simulated Annealing algorithm generates random control trajectories and dynamic MATLAB/Simulink models are used as “Black Boxes” to determine the corresponding objective value. The Simulated Annealing algorithm allows up-hill steps of the objective function and it has the potential to find the global optimum. The algorithm was modified to allow simultaneous optimization of control profiles and time-invariant control parameters. Only offline-optimization can be conducted due to the low calculation efficiency (hours to days) of Simulated Annealing. The (global) offline solutions provide excellent starting values for the online-optimization.

The screen shot shows the Simulated Annealing menu of the OCC. In the upper right corner, the values of the accepted simulations over number of total simulations are plotted (dotted cloud). An early stage of the optimization is shown here with “just” 600 dynamic simulations. The lower right axis shows the current values of the system variables if the best simulation found so far would be chosen. The user can see all system and Simulated Annealing parameters at run-time and can change most of them at run-time as well.