The reduction of CO2 emissions is a challenge for the coming decade, especially with the implementation of the Kyoto protocol. Since energy services (mainly heating and cooling of buildings) contribute to over 40% of the final energy consumption in a country like Switzerland, it is essential to find ways to improve the efficiency of energy conversion technologies. This can be done by combining these energy conversion technologies into polygeneration systems for instance. However, to ensure that polygeneration systems operate as often as possible at or near their optimal load, it is meaningful to implement systems that meet the requirements of more than just one building, in order to take advantage of the various load profiles of the buildings by compensating the fluctuations and having therefore a smoother operation. Besides, because these systems are complex and defacto difficult to operate, there are usually not justified in an individual building where no continuous professional control can be guaranteed. It is much more advantageous to implement them in a small plant that serves several buildings, and that is managed by an energy service company. This means that a network needs to be designed, that optimally connects the buildings and the energy conversion technologies together. A new methodology to design and optimize district energy systems is therefore being developed. The method (see figure) is based on the combination of an evolutionary algorithm, a network design and optimization algorithm, and several thermo- economic models for the energy conversion technologies. The first step is to select a district for which an energy system has to be developed or modified. The available renewable energy sources existing near or in the district, and thus the possible energy conversion technologies, are identified. Besides, all the relevant information regarding the district have to be structured: the geographical coordinates of the buildings, the load profiles of the buildings and finally the constraints (legal regulations, topology, existing networks,…). Once this information structuring phase is completed, the method for the design of the network and the energy conversion technologies can be applied, resulting in a number of different configurations. The costs and CO2-emissions are computed for each configuration on a Pareto-curve and the results compared. In this presentation, we present the first results of the implementation of this method.