This paper describes the algorithm for optimal distribution network reconfiguration using the combination of a heuristic approach and genetic algorithms. Although similar approaches have been developed so far, they usually had issues with poor convergence rate and long computational time, and were often applicable only to the small scale distribution networks. Unlike these approaches, the algorithm described in this paper brings a number of uniqueness and improvements that allow its application to the distribution networks of real size with a high degree of topology complexity. The optimal distribution network reconfiguration is formulated for the two different objective functions: minimization of total power/energy losses and minimization of network loading index. In doing so, the algorithm maintains the radial structure of the distribution network through the entire process and assures the fulfilment of various physical and operational network constraints. With a few minor modifications in the heuristic part of the algorithm, it can be adapted to the problem of determining the distribution network optimal structure in order to equalize the network voltage profile. The proposed algorithm was applied to a variety of standard distribution network test cases, and the results show the high quality and accuracy of the proposed approach, together with a remarkably short execution time.

No abstract

This paper presents novel cycle-break (spanning tree generation) algorithms which can be used to find the optimal distribution network topology. These algorithms (adjacency matrix/top-down/bottom-up cycle break) represent a novel way of obtaining radial network topology by cycle regrouping using adjacency matrix or elementary cycle information. Proposed methods assure connected radial network topology and can be used in combination with genetic algorithms to obtain optimal distribution network structure under minimum active power loss or network loading index framework. The cycle-break algorithms are used in initial population generation, crossover and mutation process to enhance the performance of the genetic algorithms in terms of convergence rate. These modifications make the proposed approach suitable for the use on realistic distribution networks without concern of its complexity. The algorithms are tested on a several standard test networks and the results are compared with the other existing approaches.

No abstract

No abstract