Immune system memetic algorithm for power distribution network design with load evolution uncertainty

Abstract

A recent paper, [1], has proposed a methodology for taking into account uncertainties in the load evolution within the design of electric distribution networks. That paper has presented an immunological algorithm that is used for finding a set of solutions which are sub-optimal under the viewpoint of the “mean scenario” load conditions, and which are submitted to a sensitivity analysis for the load uncertainty. This paper presents a further development of the algorithm presented in [1], employing now a memetic algorithm (an algorithm endowed with local search operators) instead of the original immunological algorithm. The new algorithm is shown to present a better behavior, achieving a better set of candidate solutions, which dominate the solution set of the former algorithm. The solution set of the proposed algorithm is also stable, in the senses that: (i) the same set of solutions is found systematically; and (ii) the merit function values associated to those solutions vary smoothly from one solution to another one. It can be concluded that the design procedure proposed in [1] should be performed preferentially with the algorithm proposed here.