This article addresses the problem of the optimal selection of conductors in asymmetric three-phase distribution networks from a combinatorial optimization perspective, where the problem is represented by a mixed-integer nonlinear programming (MINLP) model that is solved using a master-slave (MS) optimization strategy. In the master stage, an optimization model known as the generalized normal distribution optimization (GNDO) algorithm is proposed with an improvement stage based on the vortex search algorithm (VSA). Both algorithms work with discrete-continuous coding that allows us to represent the locations and gauges of the different conductors in the electrical distribution system. For the slave stage, the backward/forward sweep (BFS) algorithm is adopted. The numerical results obtained in the IEEE 8- and 27-bus systems demonstrate the applicability, efficiency, and robustness of this optimization methodology, which, in comparison with current methodologies such as the Newton metaheuristic algorithm, shows significant improvements in the values of the objective function regarding the balanced demand scenario for the 8- and 27-bus test systems (i.e., 10.30% and 1.40% respectively). On the other hand, for the unbalanced demand scenario, a reduction of 1.43% was obtained in the 27-bus system, whereas no improvement was obtained in the 8-bus grid. An additional simulation scenario associated with the three-phase version of the IEEE33-bus grid under unbalanced operating conditions is analyzed considering three possible load profiles. The first load profile corresponds to the yearly operation under the peak load conduction, the second case is associated with a daily demand profile, and the third operation case discretizes the demand profile in three periods with lengths of 1000 h, 6760 h, and 1000 h with demands of 100%, 60% and 30% of the peak load case. Numerical results show the strong influence of the expected demand behavior on the plan’s total costs, with variations upper than USD/year 260,000.00 between different cases of analysis. All implementations were developed in the MATLAB® programming environment.
