from the 6th to 9th of September 2016
|Topic||design of energy systems, local and global optimization methods, efficiency of optimization processes, robust optimization, analysis of optimization results|
|Structure||Crash course, one week, 30 h, half lectures half lab work at the computer using the software modeFRONTIER|
|Number of participants||8 -16|
|Is remote access possible and how?||No, Physical presence is needed|
|Registration, Admission criteria, Deadlines||
Registrations will open on June 1st
Deadline to register will be July 1st
|Cost||Free for KIC InnoEnergy PhD Students|
When designing an energy system, the engineer / researcher often limits its exploration of the design space to a basic parametric studies, not necessarily for lack of available computational power to perform a full optimization process but rather for lack of experience in the use of optimization techniques.
The course will provide the useful knowledge for selecting an optimizer adapted to the nature of the design problem: single or multiple-objective, constrained or unconstrained, with a limited or large number of design parameters. The focus will be on optimizing systems and energy components involving in particular the use of the software.
Some selected optimization techniques will be described in detail in order to develop an in-depth knowledge of the user-defined parameters upon which the engineer can play in order to maximize the performance of these techniques (faster convergence to a global optimum ideally).
Since the course is oriented towards engineering and not mathematical optimization the cost of the optimization process will be carefully analyzed and practical means to reduce this cost by the use of surrogate models will be proposed and described. Techniques for robust design optimization will be reviewed in order to avoid selecting an over-sensitive optimal design.
Selected optimization techniques will be finally applied to an engineering problem (design of a Die Press Model).
The lectures hours will alternate systematically with lab work sessions where engineering optimization problems will be solved using the commercial software Got-I so as to build a solid knowledge of the methods and tools described in the lectures. Knowledge gained using Got-It can be easily applied with other (free) optimization tools such as Dakota for instance. It should be noted that Got-It has a free version (with fewer options).
Intended Learning Outcomes
At the end of the course, the participant will be able to:
- Formulate an optimization problem (mono-objective and multi-objective multi-parameter engineering optimization problem).
- Select an appropriate optimizer (local vs global methods) for single or multi-objective problems.
- Ensure the robustness of an optimal design: key ideas for designing under uncertainty.
- Use methods to increase the efficiency of an optimization / reduce the computational cost: basics of surrogate modelling.
The course is oriented towards the practical solution of optimization problems arising when designing an energy system or component. The general methodologies are provided, useful for a wide range of problem scales, with a focus on multi-variable mono-objective and multi-objective problems. The key concepts are applied to a set of examples using commercial software.