District energy systems have the potential to decrease the CO2 emissions linked to energy services (heating, hot water, cooling and electricity), thanks to the implementation of large polygeneration energy conversion technologies, connected to a group of buildings over a network. The synthesis of district energy systems requires a large number of integer and continuous variables involved in non linear models, resulting in a mixed integer non linear programming problem (MINLP). A new method is being developed to design district energy systems, by decomposing the multi-objective optimization problem into two sub-problems: a master optimization problem and a slave optimization problem. In this paper, the method developed as well as the first results of the complete resolution procedure are presented.