Abstract

Distribution network Greenfield planning is recognized as an important task of the planning studies. This importance lies in its use as a regulatory tool to determine the distribution charges applied to final users. However, the Greenfield planning requires solving a combinatorial optimization problem classified as a NP-hard. Because of its complexity, different models based on heuristic strategies have been proposed to solve this problem. In this work, an evolutionary strategy -as a global search methodology-is proposed. The corresponding Greenfield problem is formulated considering the network capacity constraints and the voltage drop. The proposed algorithm assumes that the system elements costs and its optimal assignment are provided beforehand. Next, the individuals' generation is based on a heuristic biased construction of random spanning trees. Then, the fitness function penalizes the unfeasible solutions with a dynamic function. Finally, four crossover methodologies are proposed and evaluated on a test case. A comparison of the solutions and the algorithm performance is performed on a test case system