Water Network Sectorization (WNS) consists of dividing a water system into sectors with an independent water supply. When each district is supplied by its only water source the districts can be defined as i-DMAs (isolated District Meter Areas) because they are completely cut off from the rest of the water network. This isolation of the i-DMAs may decrease hydraulic performance of the water system reducing its topologic (network loops) and hydraulic (diameter sections) redundancy. Traditionally the design of WNS is carried out by empirical or simulation assisted trial-and-error approaches that are difficult to apply to large water distribution systems. In this paper an original methodology for automatic sectorization of water networks is proposed. The methodology is based on Shortest Path techniques that allow defining a tree graph of the network with dissipated power used as weight of the pipes (or links). Once the districts are found, a swapping phase follows which is achieved by using a genetic algorithm (GA) that allows refining the choice of nodes that belong to each district. The objective function of the GA is based on network mean pressure. The methodology was tested, using different performance indices, on two real water supply systems.