Paper Title
A New Artificial Intelligence Method for Distribution System Service Restoration

Service restoration is an important task for any power utility, which can be achieved via distribution system reconfiguration. This paper addresses this multi-objective, nonlinear problem by using a new, fast, non-dominated sorting genetic algorithm to satisfy all objectives simultaneously with a relatively small initial population size and number of generations. The objectives are to restore as much load as possible with fewest switching operations, minimize real power losses, maximize the minimum operating voltage, and minimize branch loading. Instead of generating several ranks from the non-dominated set of solutions, this algorithm deals with only one rank; then the most suitable solution is chosen according to the operator’s wishes. If there is no preference, the best solution is determined by considering the sum of normalized objective values. Also, a guided mutation operation is applied to speed up convergence. Radial topology is satisfied by formulating the branch-bus incidence matrix and checking the rank of each topology. Finally, a load shedding strategy is introduced based on minimizing the out-of-service load and considering priority customers. To test the algorithm, it is applied to two different cases of a 32-bus test system. Keywords - Distribution System Reconfiguration, Service Restoration, Guided Mutation, Load Shedding, Multi-Objective Optimization, Non-Dominated Sorting Genetic Algorithm.