Determination of Feeder Areas for the Design of Large Distribution Networks

One of the methods for setting distribution tariffs is benchmarking competition. In Chile, a yardstick competition scheme is applied. This yardstick consists of a "model company" that is designed in order to quantify what would be the least cost of operating an efficient distribution company serving the same demand as the actual company. Input data for the design of the company's network are load locations, magnitudes, and growth forecast, and feasible feeder routes. The design of this network, considering its high combinatorial degree, size, and operational constraints, is a difficult task and it is known to be an NP hard problem. In this paper, a new approach is presented where the problem is separated by feeder areas. The criterion used for dividing the areas is to maintain an equitable distribution of loads among them. Two methodologies are compared for area dimensioning: 1) Voronoi tessellation and 2) k-means clustering. Once the feeder areas are defined, the network is designed by applying a genetic algorithm based on the generation of spanning trees. Finally, in order to fulfill real operational constraints, it is considered that all of the feeders share the same output from the substation. This is achieved by using an algorithm that identifies shared routes and performs the final step of the design of the system.