Algorithms Based on Evolution for Multicriteria Design of Electric Distribution Networks

The optimal design of electric distribution networks has been usually carried out by the minimization of a single objective function representing the economic costs of the system planning, including the optimal size and/or location of the feeders and/or substations of the power system. Usually, such objective function is subject to a set of technical constraints (Kirchhoff ́s current law constraints, power capacity limits constraints of feeders and/or substations, etc.). Therefore, generally the optimal design is a large combinatorial non-linear optimization problem. Several optimal design models have been proposed for solving it, frequently mixed-integer design models, that have been applied by using diverse optimization techniques (mathematical programming techniques, expert systems, heuristic techniques, algorithms based on evolution, etc.). On the other hand, very few research works have been dealt with the multicriteria (multiobjective) optimal design of electric distribution networks. This paper presents the applications of algorithms based on evolution (evolutionary algorithms) to solve the design of electric networks considering multiple objectives simultaneously. An evolutionary algorithm works with a population of individuals (solutions), that can evolve by means of the application of several procedures of selection, reproduction, crossover and mutation. In this paper, two evolutionary algorithms have been applied for the multiobjective optimal design of distribution systems that allows for optimizing n objectives simultaneously. Particularly, it has been applied to the optimization of two objectives: an objective function of the distribution system economic costs and other objective function related to the electric network reliability. Furthermore, they are optimized subject to the above mentioned technical constraints of the distribution network. Thus, the achieved computer results have shown a set of non-dominated solutions (Pareto optimality solutions) that represents the best electric distribution network solutions from the point of view of the economic costs and the network reliability, that is, the solutions with the best reliability values and with the lowest economic design costs. Therefore, these non-dominated solutions are the most suitable solutions that can be shown to the planner in order to he/she selects the final best satisfactory distribution network solution from his/her professional experience and own engineer criteria. Also, this allows the planner to have deeper insight on the design process and on the problem itself. Lastly, these evolutionary algorithms have been analysed and compared form its computational performance in distribution systems of significant dimensions, showing that are efficient tools for multiobjective design.