Improved Niche-based Genetic Algorithm For Distribution Network Reconfiguration And Analysis

Distribution Network Reconfiguration is defined as an adjustment method of network structure for some purpose, which is not only an operating and controlling method of distribution system, but also an important part of distribution management system (DMS). In order to strengthen reliability and reduce loss of the network under normal condition, new operating mode through on and off switch is needed termly to reconstruct the distribution network, so that the loads can be distributed reasonably by mutual transfers among different feeders. On the other hand, corresponding network adjustment is also required to restore power supply for users and reduce power loss when faults occur.Distribution Network Reconfiguration algorithm is used to search for an optimal network structure which satisfies a certain objective function or multiple ones. Considering that there are often a great number of lines and switches and different constraints in the network, the distribution network reconfiguration is namely a NP problem. Traditional network reconstruction algorithms can be classified into three categories: optimal flow algorithms, heuristic algorithms, and artificial intelligence algorithms. Artificial intelligence algorithms include artificial neural networks, genetic algorithms, simulated annealing algorithms and so on.Niche algorithm selects a number of individuals with higher fitness as outstanding representatives of the classes divided from every generation to form species, and then creates a new generation of individuals through adaptive crossover and mutation in and among different species, and finally complete the selecting operation by the preselect mechanism, crowding mechanism or sharing mechanism. This paper presents an improved genetic algorithm based on Niche technology for distribution network reconfiguration. In the initial stage of evolution, Niche algorithm is used to maintain the diversity of species evolution and avoid falling into local optimum. In the later stage of evolution, elite strategy is introduced to make the algorithm converge to the global optimal solution with probability 1 theoretically.This thesis uses the forward and backward substitution method to calculate power flow, and takes minimal network loss as the optimization objective of the evolution. In the evolution process, a new search which is based on the node connections and status is used to verify the connectivity of network topology, and the random spanning tree approach is introduced to solve the problems of ring network and isolated island. Finally, the network-equivalent method is applied to reliability analysis of distribution network. The feasibility of the algorithm proposed is verified by numerical simulations.