Abstract:
Ce memoire traite de l foptimisation du dispatching economique dans les reseaux electriques, en
particulier a travers l fintegration des pertes techniques et l futilisation des algorithmes genetiques
(AG). Face aux enjeux croissants lies a la gestion efficace de l fenergie electrique, le travail propose
une approche innovante basee sur les metaheuristiques, mise en oeuvre a l faide du langage Python et
de la bibliotheque Pymoo. Deux reseaux standards ont ete etudies : IEEE 30 noeuds et IEEE 59
noeuds. Pour chacun, plusieurs simulations ont ete realisees afin de comparer les couts totaux avec et
sans prise en compte des pertes, en fonction d fun facteur de ponderation ƒ¿. Les resultats ont mis en
evidence l fimportance strategique d fintegrer les pertes dans les modeles economiques, notamment
dans les regimes intermediaires ou leur impact est maximal. Le modele developpe demontre
egalement la performance des AG dans la resolution de problemes multi-objectifs complexes, avec
des resultats precis, fiables et adaptes aux contraintes des reseaux modernes. Ce travail ouvre la voie
a de futures extensions vers des reseaux plus vastes et integrants des energies renouvelables ou des
contraintes en temps reel.
This thesis addresses the optimization of economic dispatch in electrical networks, particularly
through the integration of technical losses and the use of genetic algorithms (GA). In response to the
growing challenges related to efficient energy management, the work proposes an innovative approach
based on metaheuristics, implemented using the Python language and the Pymoo library. Two standard
networks were studied: IEEE 30-bus and IEEE 59-bus. For each, several simulations were conducted
to compare total costs with and without considering losses, depending on a weighting factor á. The
results highlighted the strategic importance of integrating losses into economic models, especially in
intermediate regimes where their impact is greatest. The developed model also demonstrates the
performance of GAs in solving complex multi-objective problems, delivering accurate, reliable results
adapted to the constraints of modern networks. This work paves the way for future extensions to larger
networks incorporating renewable energy or real-time constraints