GA-based feature selection for faults identification in electrical distribution systems

This paper develops a novel diagnostic tool for the identification and location of faults in radial distribution networks. In this aim, the problem of minimizing the number of electrical features allowing the identification of a set of faults has been dealt with. An overview of the fault diagnosis problem in medium voltage distribution systems is first presented. Then the diagnostic problem in electrical distribution systems is described and a new diagnostic model is presented. Finally the paper suggests an evolutionary algorithm for the electrical features selection problem, based on an advanced gene expression based genetic algorithm. In the last section the results for an application carried out for a MV radial network are presented and discussed.