Novel spectral partitioning method for large-scale distribution system feeder reconfiguration

This paper proposes a novel spectral partitioning algorithm, recursive spectral bisection, for weighted graph partitioning that models large-scale distribution feeder reconfiguration problem. Since the large-scale network may not be partitioned efficiently, we present a multi-level method to speed up. Firstly we replace the graph by a coarse approximation and partition it instead in order to get a rough partitioning, and then uncoarsen to get the final partitions. This algorithm can overcome the weakness of many of the previously described reconfiguration algorithms, which require a load flow solution to be performed in each iteration of the reconfiguration. Besides, the result is independent of the initial topology of networks. Hence it can achieve the global or near-global optimum solution in shorter time than other methods and makes it possible to achieve on-line distribution feeder reconfiguration for loss reduction and load balancing.