ABSTRACT  

In electrical engineering, designing a heterogeneous system must be 
processed by means of a global approach, also called “systemic approach”. 
Indeed, the existence of interactions between the different sub-systems 
and the multi-field nature the device strongly influence the system 
performances. The optimisation of the whole system with regard to several 
objectives has to be achieved by including all these couplings. The work 
done in this thesis deals with multiobjective evolutionary optimisation 
applied to the optimal design. Optimisation issues resulting of this 
process are complex. For that purpose, multiobjective Pareto evolutionary 
algorithms have been chosen. To select genetic operators, we have developed 
a self-adaptation which increases the exploration robustness. Through the 
design of several electrical engineering systems (example of an electrical 
vehicle,…), global design oriented models have been proposed from which 
algorithms are used to seek the best trade-offs versus optimisation criteria 
(losses, mass,…). Exploiting optimisation results shows that this process 
constitutes a very efficient design tool if performance and coupling analysis 
issues are concerned.

Department:	Laboratoire d'Electrotechnique et d'Electronique Industrielle (LEEI)
Thesis Advisors:	Roboam, Xavier and Sareni, Bruno

KEYWORDS : Optimal design - System approach - Couplings - System modelling - 
Optimisation - Multiobjective methods - Evolutionary methods - Genetic algorithms
Subjects:	Electrical engineering