Evolving non-dominated solutions in multiobjective service restoration for automated distribution networks

Abstract The problem here dealt with is that of Service Restoration (SR) in automated distribution networks. In such networks, configuration and compensation level as well as loads insertion status can be remotely controlled. The considered SR problem should be handled using Multiobjective Optimization, MO, techniques since its solution requires a compromise between different criteria. In the adopted formulation, these criteria are the supply of the highest number of loads and the minimum power losses. The Authors propose a new MO approach, the Non-dominated Sorting Fuzzy Evolution Strategy, NS_FES, which uses part of the Non-dominated Sorting Genetic Algorithm, NSGA, proposed by K. Deb. The ability of NSGA to divide a population of solutions in classes of dominance allows a fruitful application of another efficient MO strategy already proposed and tested by the Authors (FES, Fuzzy Evolution Strategy). In this way, diversity and high quality of solutions is possible. After a brief description of the SR problem and a review of the approaches recently proposed in literature, the NS_FES solution strategy is presented in detail. Finally, test results using the three approaches (NSGA, FES, NS_FES) are carried out and compared.