A dynamic partitioning scheme for parallel transient stability analysis

Abstract

A dynamic partitioning scheme for a large power system is used to speed up the transient stability solution by utilizing different step sizes for the different portions. The partitioning and the different step sizes naturally lend themselves to a parallel implementation. Results are obtained by utilizing the successive overrelaxation (SOR) Newton parallel algorithm on the Alliant FX/8, an eight-processor parallel computer. For a 662 bus system, a total speedup, by partitioning and parallelizing, of 33.5 is obtained. Since parallelization by itself with 8 processors can produce a speedup of about 7, the partitioning scheme provides a speedup factor of about 5. The partitioning scheme is not specific to either the SOR Newton algorithm or the Alliant machine, and its application to any other parallel method should result in a similar speedup.<>