A Simulated Annealing Genetic Algorithm for Urban Power Grid Partitioning Based on Load Characteristics

Abstract

Power network partitioning is an old but still challenging and meaningful problem. A large power grid is divided into several zones so that buses within zones are electrically close. Urban power grid partitioning is usually ignored due to its scale is smaller than large power grid partitioning. This paper proposes a method that considers the characteristics of urban loads and a hybridization of genetic algorithm with simulated annealing algorithm is applied to get the best scheme for network partitioning problems. The optimal number of partitions is determined by related theories of matrix analysis and the candidate solutions are evaluated by a multi-index fitness function based on the electrical distance matrix. This method is applied to a real urban power grid with 134 buses. The short circuit currents before and after partitioning are compared to test the validity and practicability of this method.