Genetic algorithms and Voronoi polygons applied to decision making in the distribution systems expansion problem

Some expansion plans consider a lower demand scenario, producing an increase in the network losses and the risk of a failure on the protection devices. The solution of this problem comprises addition of new feeders to get the system back to an adequate operating level. The formulation of this problem results in a optimization problem classified as of NP- hard. Additionally, there are operational constraints that must be considered, such as the maximum current per feeder and the consideration that some feeders are forced to share routes in urban areas. Given the real demand data and the planning period, a new load forecast is performed making possible to determine the number of feeders to supply the loads. Due to the size of the problem, a ldquodivide and conquerrdquo strategy is applied by means of Voronoi polygons, making possible to establish coverage areas for each feeder. The final assigned areas are determined using a genetic algorithm (GA) that balances the load among feeders. Thus, the expansion problem is divided into coverage areas and for each of them a GA based on the generation of random spanning trees where three different types of costs are considered (expansion costs, reinforcement costs and operational costs) is applied. This paper presents results for artificially simulated system.